Browse the corpus

Panel Research in context Evidence before this study The primary progression-free survival analysis of the ICON7 trial reported significantly improved progression-free survival when bevacizumab was added to standard chemotherapy in newly diagnosed ovarian cancer. The effect was greatest in patients at high risk of disease progression. Similar progression-free survival findings were reported in the GOG-218 trial. Added value of this study In a planned mature analysis of overall survival, no difference in overall survival was noted between those patients who received bevacizumab plus chemotherapy and those who received chemotherapy alone. However, in subgroup analyses, improved overall survival was noted in patients at high risk of disease progression who received bevacizumab compared with those who did not. Implications of all the available evidence Bevacizumab may have a role in the treatment of patients with poor-prognosis ovarian cancer. Future work should address questions of treatment duration, targeting, timing, re-challenge, and dose fractionation. Introduction Ovarian cancer is the seventh most common cancer worldwide, with 238 700 new cases and 151 900 deaths in 2012.1 The prognosis of the disease remains poor: the European mean age-standardised 5-year survival was only 37·6% for women diagnosed between 2000 and 2007.2

Bevacizumab may have a role in the treatment of patients with poor-prognosis ovarian cancer. Future work should address questions of treatment duration, targeting, timing, re-challenge, and dose fractionation. Introduction Ovarian cancer is the seventh most common cancer worldwide, with 238 700 new cases and 151 900 deaths in 2012.1 The prognosis of the disease remains poor: the European mean age-standardised 5-year survival was only 37·6% for women diagnosed between 2000 and 2007.2 Until 2011, the international standard of care for women with advanced or poor-prognosis early-stage ovarian cancer mainly consisted of debulking surgery followed by chemotherapy with carboplatin and paclitaxel.3 Since then, modulation of VEGF has moved from a theoretical concept4 to a key component of treatment. Two large-scale phase 3 randomised trials, GOG-218 and ICON7, both undertaken in first-line settings, showed that the addition of the anti-VEGF monoclonal antibody, bevacizumab, to conventionally administered carboplatin and paclitaxel chemotherapy significantly improved progression-free survival.5, 6 Two further randomised trials in recurrent ovarian cancer have shown significant improvement in progression-free survival through the addition of bevacizumab to conventionally administered carboplatin and gemcitabine chemotherapy both in the platinum-sensitive and platinum-resistant relapsed setting.7, 8 These data were used to support bevacizumab licensing through the European Medicines Agency for use in the first-line setting for patients with at least International Federation of Gynecology and Obstetrics (FIGO) stage IIIB disease (according to 1988 staging criteria), at first recurrence for patients with platinum-sensitive disease not previously treated with bevacizumab or other VEGF-targeted drugs, and in the setting of platinum-resistant recurrence combined with paclitaxel, topotecan, or pegylated liposomal doxorubicin.9 Bevacizumab has also been approved in the platinum-refractory setting by the US Food and Drug Administration.10

tive disease not previously treated with bevacizumab or other VEGF-targeted drugs, and in the setting of platinum-resistant recurrence combined with paclitaxel, topotecan, or pegylated liposomal doxorubicin.9 Bevacizumab has also been approved in the platinum-refractory setting by the US Food and Drug Administration.10 The GOG-218 trial,5 which enrolled 1873 patients with FIGO stage III–IV ovarian cancer with macroscopic residual disease after primary surgery, showed a significant improvement in progression-free survival with the addition of bevacizumab (hazard ratio [HR] 0·72, 95% CI 0·63–0·82; p<0·001). Patients received treatment with carboplatin and paclitaxel chemotherapy, and either concurrent bevacizumab 15 mg per kg every 3 weeks followed by up to 16 cycles of maintenance bevacizumab (at the same dose) or placebo for the same duration, all administered intravenously. The ICON7 trial,6 which was done in a patient population that included those with poor-prognosis, early-stage disease and those with optimally or suboptimally debulked advanced disease, showed improved progression-free survival in patients receiving bevacizumab, with restricted mean progression-free survival over 36 months of 20·3 months on standard chemotherapy and 21·8 months with bevacizumab (HR 0·81, 95% CI 0·70–0·94; p=0·004). An increased effect was noted in patients at high risk of disease progression, a similar group to the GOG-218 study population, with restricted mean progression-free survival after 42 months' follow-up of 14·5 months in the standard chemotherapy group and 18·1 months in the bevacizumab group (HR 0·73, 95% CI 0·60–0·93; p=0·002).

ed effect was noted in patients at high risk of disease progression, a similar group to the GOG-218 study population, with restricted mean progression-free survival after 42 months' follow-up of 14·5 months in the standard chemotherapy group and 18·1 months in the bevacizumab group (HR 0·73, 95% CI 0·60–0·93; p=0·002). To fully assess a new treatment, the effect on overall survival as well as progression-free survival must be known. ICON7 was designed with a progression-free survival primary endpoint but was also powered to detect an overall survival improvement. At the time of the primary progression-free survival analysis, overall survival data were immature in both ICON7 and GOG-218, and preliminary analyses showed no difference between treatment groups in either study. Here, we present the final analysis of mature overall survival data from ICON7, together with detailed data about the effect of bevacizumab according to stage and extent of residual disease after primary debulking surgery.

GOG-218, and preliminary analyses showed no difference between treatment groups in either study. Here, we present the final analysis of mature overall survival data from ICON7, together with detailed data about the effect of bevacizumab according to stage and extent of residual disease after primary debulking surgery. Methods Study design and participants The design of this trial has previously been described in detail.6 In brief, eligible women with ovarian cancer were recruited from 263 centres (a mixture of general hospitals and specialist centres) in 11 countries across Europe, Canada, Australia, and New Zealand (appendix pp 1–5) and were randomly assigned 1:1 in an open-label study to receive standard chemotherapy or standard chemotherapy with bevacizumab. Eligible patients were aged 18 years or older; with newly diagnosed epithelial ovarian, fallopian tube, or primary peritoneal cancer; an Eastern Cooperative Oncology Group (ECOG) performance status of 0–2; FIGO 1988 stage IIb–IV or high-risk (grade 3 or clear cell histology) stage I–IIa disease; had undergone debulking cytoreductive surgery or, in advanced disease, had a biopsy with no further surgery planned; and had adequate coagulation parameters and liver, renal, and bone marrow function. The exclusion criteria were having other tumour types, previous systemic therapy, planned surgery, and uncontrolled hypertension.

one debulking cytoreductive surgery or, in advanced disease, had a biopsy with no further surgery planned; and had adequate coagulation parameters and liver, renal, and bone marrow function. The exclusion criteria were having other tumour types, previous systemic therapy, planned surgery, and uncontrolled hypertension. The study protocol was compliant with good clinical practice guidelines and the Declaration of Helsinki. Ethics approval was obtained in all participating countries and where required in all participating centres. All patients provided written informed consent. Randomisation and masking Randomisation was done centrally by a computer system based at the Medical Research Council Clinical Trials Unit (London, UK) accessed via the web or telephone. Randomisation was done using 1:1 allocation and a minimisation algorithm, and was stratified by Gynecologic Cancer Intergroup (GCIG) group, a combination of FIGO stage and residual disease (stage I–III and ≤1 cm residual disease vs stage I–III and >1 cm residual disease vs inoperable stage III and stage IV disease) and planned interval between surgery and chemotherapy (≤4 weeks or >4 weeks). Neither patients nor physicians were masked to treatment allocation.

ombination of FIGO stage and residual disease (stage I–III and ≤1 cm residual disease vs stage I–III and >1 cm residual disease vs inoperable stage III and stage IV disease) and planned interval between surgery and chemotherapy (≤4 weeks or >4 weeks). Neither patients nor physicians were masked to treatment allocation. Procedures Patients received either six 3-weekly cycles of intravenous carboplatin (AUC 5 or 6) and paclitaxel (175 mg/m2 of body surface area), or the same regimen with intravenous bevacizumab (7·5 mg/kg of bodyweight) given concurrently and continued for 12 further 3-weekly cycles (with a duration of bevacizumab exposure of about 1 year), or until disease progression. To avoid delayed wound healing, bevacizumab was omitted at cycle 1 if chemotherapy was started within 4 weeks of surgery. Bevacizumab cycles omitted for any reason were not replaced.

tly and continued for 12 further 3-weekly cycles (with a duration of bevacizumab exposure of about 1 year), or until disease progression. To avoid delayed wound healing, bevacizumab was omitted at cycle 1 if chemotherapy was started within 4 weeks of surgery. Bevacizumab cycles omitted for any reason were not replaced. CT scans were done after treatment cycles 3 and 6, and then 9 and 12 months after randomisation. Following treatment, women were seen every 3 months until the end of year 3, every 6 months during years 4 and 5, and annually thereafter. Scans continued every 6 months until the end of year 3, then as clinically indicated. Disease progression was assessed by investigators according to RECIST 200011 guidelines, and needed radiological or clinical evidence of progression. Cancer antigen 125 (CA125) progression alone was insufficient to define progressive disease. Following disease progression, women were seen every 6 months up until year 5, then annually. Quality of life was assessed with the European Organisation for Research and Treatment of Cancer QLQ-C30 and QLQ-OV28 questionnaires. Outcomes The primary outcome of ICON7 was progression-free survival, which has been previously reported.6 Secondary outcomes were overall survival and safety outcomes of adverse events, laboratory results, and worsened ECOG performance status. Exploratory outcome measures were quality of life, health economics, and translational research.

outcome of ICON7 was progression-free survival, which has been previously reported.6 Secondary outcomes were overall survival and safety outcomes of adverse events, laboratory results, and worsened ECOG performance status. Exploratory outcome measures were quality of life, health economics, and translational research. Risk groups were defined at the time of the primary progression-free survival analysis to enable comparison with the GOG-218 study population. They were refined slightly before database lock for the present analysis in accordance with current clinical practice and defined prospectively in the statistical analysis plan. High risk of progression was defined as stage IV disease, inoperable stage III disease, or suboptimally debulked (>1 cm) stage III disease (appendix p 10). To enable comparison with previous analyses, results are also presented for two other high-risk definitions: exclusion of inoperable stage III–IV patients, to match the previous ICON7 high-risk group; and inclusion of patients with 0–1 cm residual tumour, to match the GOG-218 population (appendix p 8). Between the primary progression-free survival analysis and the present analysis, recruiting centres were asked (following the GCIG fourth ovarian cancer consensus conference statement3) to reclassify patients with up to 1 cm of residual disease into those with no macroscopic residuum or those with macroscopic residuum measuring 1 cm or less. The non-high-risk patients were defined as those who did not meet the criteria for high-risk disease.

GCIG fourth ovarian cancer consensus conference statement3) to reclassify patients with up to 1 cm of residual disease into those with no macroscopic residuum or those with macroscopic residuum measuring 1 cm or less. The non-high-risk patients were defined as those who did not meet the criteria for high-risk disease. Three further patient subgroups of particular interest were defined: patients with clear cell carcinoma (roughly 10% of patients because of enriched enrolment—ie, the use of less restrictive staging criteria, which meant that patients with clear cell carcinoma of any stage were eligible); high-risk low-stage patients, with stage I–IIA clear cell or grade 3 carcinoma; and low-grade serous carcinomas (grade 1). Statistical analysis This study was designed and powered to detect differences in progression-free and overall survival between the treatment groups. The analysis of overall survival needed 715 deaths to detect a 10-month improvement in median survival from 43 to 53 months (HR 0·81), with 80% power at a two-sided 5% significance level. The progression-free survival analysis needed 684 disease progression events to show a 5-month progression-free survival increase from 18 to 23 months, with 90% power and two-sided 5% significance level.

improvement in median survival from 43 to 53 months (HR 0·81), with 80% power at a two-sided 5% significance level. The progression-free survival analysis needed 684 disease progression events to show a 5-month progression-free survival increase from 18 to 23 months, with 90% power and two-sided 5% significance level. Analysis followed the principle of intention-to-treat and included all patients randomly assigned to treatment. The primary analysis used an unstratified log-rank test to compare overall survival between randomised groups. Treatment effects were estimated from Cox regression analyses when proportional hazards could be assumed. With evidence of non-proportionality, flexible parametric survival models12 were used to smooth survival curves and estimate survival differences during a 5-year period, which is the approximate follow-up if patients were enrolled midway through the recruitment period and remained in follow-up at the study end. Stata version 13.1 was used for all analyses. This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN 91273375.

Analysis followed the principle of intention-to-treat and included all patients randomly assigned to treatment. The primary analysis used an unstratified log-rank test to compare overall survival between randomised groups. Treatment effects were estimated from Cox regression analyses when proportional hazards could be assumed. With evidence of non-proportionality, flexible parametric survival models12 were used to smooth survival curves and estimate survival differences during a 5-year period, which is the approximate follow-up if patients were enrolled midway through the recruitment period and remained in follow-up at the study end. Stata version 13.1 was used for all analyses. This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN 91273375. Role of the funding source This study was led and funded by the UK Medical Research Council Clinical Trials Group. The trial was designed by members of the trial management group who reviewed and approved the protocol. The trial management group included representatives of the GCIG participating groups and the funding sources. Final decisions about trial conduct were the responsibility of chief investigators and funder. The trial management group were invited to comment on draft versions of this report, but responsibility for the report remained with the authors. ADC and AE had full access to all the raw data, and ADC had final responsibility for the decision to submit for publication.

responsibility of chief investigators and funder. The trial management group were invited to comment on draft versions of this report, but responsibility for the report remained with the authors. ADC and AE had full access to all the raw data, and ADC had final responsibility for the decision to submit for publication. Results Between Dec 18, 2006, and Feb 16, 2009, seven GCIG groups recruited 1528 women with ovarian cancer from 263 centres across Europe, Canada, Australia, and New Zealand; these women were enrolled and randomly assigned to receive standard carboplatin and paclitaxel chemotherapy (n=764) or standard chemotherapy plus bevacizumab (n=764; figure 1). Median follow-up was 48·9 months (IQR 26·6–56·2), and follow-up ended on March 31, 2013. Patient baseline characteristics are summarised in appendix p 8; the median age of the patients was 57 years (IQR 50–64); 1415 (94%) of 1501 (excluding those with unknown performance status) had an ECOG performance status of 0 or 1; 1340 (89%) of 1502 (excluding those with cancer originating from multiple sites) had cancer of ovarian origin; 1054 (69%) had disease of serous histology; 175 (11%) had FIGO stage III, IIIA, or IIIB disease, and 1071 (70%) stage IIIC or IV disease. Following primary surgery, 395 (26%) patients had residual disease larger than 1 cm, 369 (24%) had visible disease up to 1 cm in diameter, and 734 (48%) had no visible residual disease. All randomised patients were included in analyses, and patient characteristics were well balanced between the groups (appendix p 8). Median follow-up was 48·6 months (IQR 24·3–56·0) in the standard chemotherapy group and 48·8 months (28·2–56·4) in the bevacizumab group, with shorter follow-up durations of 29·0 months (14·1–50·7) and 38·9 months (21·1–52·5), respectively, for high-risk patients.

s were well balanced between the groups (appendix p 8). Median follow-up was 48·6 months (IQR 24·3–56·0) in the standard chemotherapy group and 48·8 months (28·2–56·4) in the bevacizumab group, with shorter follow-up durations of 29·0 months (14·1–50·7) and 38·9 months (21·1–52·5), respectively, for high-risk patients. The patient subgroup at high risk of progression consisted of 502 (33%) of 1528 patients. Their median age was 60 years (IQR 52–66) and 60 (12%) had cancer of primary peritoneal origin (compared with 106 [7%] of all enrolled patients overall). Most of the high-risk patients (381 [76%]) had serous-type ovarian cancer, and the group included 30 (6%) patients who did not undergo debulking surgery.

28 patients. Their median age was 60 years (IQR 52–66) and 60 (12%) had cancer of primary peritoneal origin (compared with 106 [7%] of all enrolled patients overall). Most of the high-risk patients (381 [76%]) had serous-type ovarian cancer, and the group included 30 (6%) patients who did not undergo debulking surgery. In total, 714 patients (47%) died during the study: 352 (46%) of those in the chemotherapy group and 362 (47%) of those in the chemotherapy plus bevacizumab group (table 1). The difference in overall survival between randomised groups was neither clinically nor statistically significant (log-rank test p=0·85), although non-proportionality was evident (p=0·02). Figure 2A shows the Kaplan-Meier survival curves for the two groups. Over time, the largest absolute difference in survival was less than 5%, occurring around 2 years after enrolment and favouring patients who received bevacizumab (figure 2B). Because the evidence of non-proportionality renders a hazard ratio difficult to interpret, we estimated restricted mean survival in each group. Restricted mean survival was 44·6 months (95% CI 43·2–45·9) for women in the chemotherapy group and 45·5 months (44·2–46·7) in the chemotherapy plus bevacizumab group (table 1).

ecause the evidence of non-proportionality renders a hazard ratio difficult to interpret, we estimated restricted mean survival in each group. Restricted mean survival was 44·6 months (95% CI 43·2–45·9) for women in the chemotherapy group and 45·5 months (44·2–46·7) in the chemotherapy plus bevacizumab group (table 1). Of the 502 high-risk patients, 332 (66%) died, including 174 (69%) of 254 in the chemotherapy group and 158 (54%) of 248 in the bevacizumab group (table 1). Evidence suggested longer overall survival in those who had received bevacizumab (p=0·03, figure 2C) but evidence of non-proportional hazards (p=0·01) meant that the hazard ratio was difficult to interpret. Restricted mean overall survival time was 39·3 months (95% CI 37·0–41·7) for the bevacizumab group and 34·5 months (32·0–37·0) for the chemotherapy group (log-rank p=0·03; table 1). The absolute difference in survival exceeded 10% after 2 years, and remained at 4·4% (95% CI −4·1 to 12·9) at 5 years (figure 2D). However, in non-high-risk patients, the restricted mean survival time did not differ significantly between the two treatment groups (49·7 months [95% CI 48·3–51·1]) in the standard chemotherapy group vs 48·4 months [47·0–49·9] in the bevacizumab group; p=0·20). Further analyses of survival by stage, residual disease burden, and risk of recurrence showed a benefit from bevacizumab with worsening prognostic factors (figure 3). Similar patterns were also noted for progression-free survival (p=0·014 for stage, p=0·005 for high risk; appendix p 12).

Of the 502 high-risk patients, 332 (66%) died, including 174 (69%) of 254 in the chemotherapy group and 158 (54%) of 248 in the bevacizumab group (table 1). Evidence suggested longer overall survival in those who had received bevacizumab (p=0·03, figure 2C) but evidence of non-proportional hazards (p=0·01) meant that the hazard ratio was difficult to interpret. Restricted mean overall survival time was 39·3 months (95% CI 37·0–41·7) for the bevacizumab group and 34·5 months (32·0–37·0) for the chemotherapy group (log-rank p=0·03; table 1). The absolute difference in survival exceeded 10% after 2 years, and remained at 4·4% (95% CI −4·1 to 12·9) at 5 years (figure 2D). However, in non-high-risk patients, the restricted mean survival time did not differ significantly between the two treatment groups (49·7 months [95% CI 48·3–51·1]) in the standard chemotherapy group vs 48·4 months [47·0–49·9] in the bevacizumab group; p=0·20). Further analyses of survival by stage, residual disease burden, and risk of recurrence showed a benefit from bevacizumab with worsening prognostic factors (figure 3). Similar patterns were also noted for progression-free survival (p=0·014 for stage, p=0·005 for high risk; appendix p 12). No benefit of bevacizumab was reported for other predefined poor-prognosis tumour types (table 2). Some baseline imbalance was recorded within subgroups, which is most likely a consequence of small numbers and is unlikely to have affected the results (appendix p 9). The mortality rate in all three subgroups was lower than in the overall trial population, especially in patients with low-stage high-grade disease (table 2). No evidence of difference between treatment groups was recorded within these subgroups (table 2).

ers and is unlikely to have affected the results (appendix p 9). The mortality rate in all three subgroups was lower than in the overall trial population, especially in patients with low-stage high-grade disease (table 2). No evidence of difference between treatment groups was recorded within these subgroups (table 2). In our extension of the previously reported quality-of-life analysis, now including data up to the predefined timepoint of week 76, in patients without disease progression, global quality of life did not differ between those who received standard chemotherapy and those receiving bevacizumab at week 76 (p=0·43, appendix p 9). In further exploratory analyses, a clinically small difference was recorded in patients receiving bevacizumab who were in the non-high-risk group relative to patients not receiving bevacizumab (−5·1 points, 95% CI −9·4 to −0·7; p=0·02), whereas a small and insignificant benefit (+4·3 points, 95% CI −4·9 to 13·4; p=0·36) relative to patients not receiving bevacizumab was noted in high-risk patients. A sensitivity analysis of missing data suggested that these findings were robust (appendix p 9).

ng bevacizumab (−5·1 points, 95% CI −9·4 to −0·7; p=0·02), whereas a small and insignificant benefit (+4·3 points, 95% CI −4·9 to 13·4; p=0·36) relative to patients not receiving bevacizumab was noted in high-risk patients. A sensitivity analysis of missing data suggested that these findings were robust (appendix p 9). The primary analysis of progression-free survival was previously reported when 759 patients had experienced disease progression or died (392 in the standard chemotherapy group and 367 in the bevacizumab group).6 Since these primary analyses, a further 321 patients have subsequently progressed or died without progression (134 in the standard chemotherapy group and 187 in the bevacizumab group) for a total of 1080 progression events or deaths. The overall difference in progression-free survival between randomised groups was no longer statistically significant (table 1; appendix p 11). However, in high-risk patients, a significant benefit remains (p=0·001) with strong evidence of non-proportional hazards (p<0·0001), and longer mean restricted progression-free survival in the bevacizumab group than in the chemotherapy group (table 1).

s was no longer statistically significant (table 1; appendix p 11). However, in high-risk patients, a significant benefit remains (p=0·001) with strong evidence of non-proportional hazards (p<0·0001), and longer mean restricted progression-free survival in the bevacizumab group than in the chemotherapy group (table 1). Most adverse events that occurred in the trial have been previously reported: bevacizumab was associated with an increase in grade 1–2 mucocutaneous bleeding (271 [36%] patients in the bevacizumab group vs 55 [7%] patients in the standard chemotherapy group), grade 2 or worse hypertension (136 [18%] vs 16 [2%]), grade 3 or worse thromboembolic events (51 [7%] vs 23 [3%]), and grade 3 or worse gastrointestinal perforations (ten [1%] vs three [<1%]).6 During extended follow-up of overall survival, one further treatment-related grade 3 event (gastrointestinal fistula in a bevacizumab-treated patient), three grade 2 treatment-related events (cardiac failure, sarcoidosis, and foot fracture, all in bevacizumab-treated patients), and one grade 1 treatment-related event (vaginal haemorrhage, in a patient treated with standard chemotherapy) were also reported.

t (gastrointestinal fistula in a bevacizumab-treated patient), three grade 2 treatment-related events (cardiac failure, sarcoidosis, and foot fracture, all in bevacizumab-treated patients), and one grade 1 treatment-related event (vaginal haemorrhage, in a patient treated with standard chemotherapy) were also reported. Discussion The results of ICON7 show that bevacizumab did not improve overall survival in the intention-to-treat population of women randomly assigned to receive it in conjunction with chemotherapy, although heterogeneity of benefit was observed dependent on residual disease burden before treatment. Although the overall difference between treatment groups was not statistically significant, non-proportionality was recorded, despite the fact that the magnitude of the change over time was not clinically meaningful. However, in a preplanned analysis, women at high risk of disease progression had a significant improvement in overall survival with the addition of bevacizumab to standard chemotherapy. A similar improvement in overall survival was also reported in high-risk patients (>1 cm residual tumour) in GOG-218 (HR 0·73 in ICON7, HR 0·86 in GOG-218), with some differences as expected because of varying post-progression treatment strategies, in particular greater use of bevacizumab (ie, more participants receiving the drug) in GOG-218 patients. These findings are relevant since, in practice, bevacizumab use has become focused on the high-risk patient group.

OG-218), with some differences as expected because of varying post-progression treatment strategies, in particular greater use of bevacizumab (ie, more participants receiving the drug) in GOG-218 patients. These findings are relevant since, in practice, bevacizumab use has become focused on the high-risk patient group. In addition to the contrast between high-risk and non-high-risk patients, there was a clear association between increasing disease severity and a stronger beneficial effect of bevacizumab (figure 3). To our knowledge, this is the first study with bevacizumab to show this trend in a single trial. These observations provide a clinical framework for the appropriate use of bevacizumab in ovarian cancer that is consistent with the biological requirement for angiogenesis in growing tumours, and a hypothetical framework to explain this effect biologically. Our data suggest that a residual physical tumour burden, presumably producing VEGF, is necessary to enable bevacizumab to exert its effect on the tumour microenvironment. Other trials in ovarian cancer have also reported an overt benefit in women with a measurable (higher) disease burden following recurrence, in both platinum-sensitive7 and platinum-resistant settings.8

den, presumably producing VEGF, is necessary to enable bevacizumab to exert its effect on the tumour microenvironment. Other trials in ovarian cancer have also reported an overt benefit in women with a measurable (higher) disease burden following recurrence, in both platinum-sensitive7 and platinum-resistant settings.8 These final results complement the findings of the earlier primary progression-free survival analysis,6 with a benefit of bevacizumab recorded in women with advanced-stage suboptimally debulked disease. The magnitude and duration of benefit in this group is both clinically and statistically significant. The primary progression-free survival analyses also showed similar outcomes in GOG-218 patients and the high-risk patient group of ICON7.5, 6 The updated progression-free survival analysis reported here showed a reduced overall effect, with greater elapsed time from treatment with bevacizumab, but the earlier results remain the primary pre-specified analysis.

lyses also showed similar outcomes in GOG-218 patients and the high-risk patient group of ICON7.5, 6 The updated progression-free survival analysis reported here showed a reduced overall effect, with greater elapsed time from treatment with bevacizumab, but the earlier results remain the primary pre-specified analysis. Our data also identify patients who might not benefit from bevacizumab in the first-line setting. Women with early-stage (FIGO stage I/II) disease, even if judged to be high risk on the basis of grade or clear cell histology, do not seem to benefit. Women with optimally debulked (<1 cm) stage III disease also had no benefit. Furthermore, a small reduction in overall quality of life was recorded in non-high-risk patients treated with bevacizumab. Our trial included patients with all stages of newly diagnosed epithelial ovarian, fallopian tube, and primary peritoneal cancer for whom postoperative chemotherapy would usually be indicated. It also included 30 patients in whom primary, and subsequent, debulking surgery was regarded as unlikely to be in the patient's best interests.

ed patients with all stages of newly diagnosed epithelial ovarian, fallopian tube, and primary peritoneal cancer for whom postoperative chemotherapy would usually be indicated. It also included 30 patients in whom primary, and subsequent, debulking surgery was regarded as unlikely to be in the patient's best interests. Three subgroups based on tumour type were also predefined: clear cell, low-grade serous, and high-risk low-stage cancer. Women with clear cell carcinoma comprised 10% of the study population. This histology was previously thought to confer a substantially worse outcome than other tumour subtypes but these patients did surprisingly well in our trial, with a 72% survival after a median of 51 months' follow-up, with no benefit from bevacizumab. Patients with low-grade serous cancer also did not benefit from the addition of bevacizumab, although a major limitation of this assessment is the absence of central pathology review for these tumours. Patients with low-stage high-risk tumours also did not benefit from the addition of bevacizumab. For all three subgroup analyses, the numbers of patients were small and statistical power to detect differences was low.

gh a major limitation of this assessment is the absence of central pathology review for these tumours. Patients with low-stage high-risk tumours also did not benefit from the addition of bevacizumab. For all three subgroup analyses, the numbers of patients were small and statistical power to detect differences was low. To reflect on the choice of outcome measures for the ICON7 trial is pertinent. The primary outcome measure was progression-free survival, but the trial was also a-priori structured to assess overall survival. To assess this outcome measure needed a further 3 years of follow-up after the primary progression-free survival analysis, but also simplified aspects of trial design, such as placebo control and independent masked radiology review, which are essential when progression-free survival is the only outcome measure. The primary analysis in 2011 showed a significant progression-free survival benefit in the intention-to-treat population, which was most pronounced in women at high risk of progression.6 Over time, the effect closely followed bevacizumab treatment, with the maximum benefit coinciding precisely with duration of treatment (appendix p 11). Notably, the overall survival difference outlasts the duration of bevacizumab exposure and points to a durable benefit in the high-risk group (figure 2D), raising the possibility of additional benefit to high-risk patients from further extension of treatment duration, which is the subject of ongoing research in the BOOST trial (NCT01462890). Following disease progression in ICON7, the pattern of further treatment was similar in both randomised groups, with little use of further bevacizumab in either group.

l benefit to high-risk patients from further extension of treatment duration, which is the subject of ongoing research in the BOOST trial (NCT01462890). Following disease progression in ICON7, the pattern of further treatment was similar in both randomised groups, with little use of further bevacizumab in either group. Questions about the optimum timing of bevacizumab therapy in the trajectory of a woman's disease remain. Should it be considered at initial presentation, time of platinum-sensitive recurrence, or after the development of platinum resistance? Bevacizumab has shown a significant progression-free survival benefit in all these settings,5, 6, 7, 8 and an overall survival benefit in some settings too.6, 13, 14 Our data strongly support early use of bevacizumab, based on risk and disease burden. Whether or not bevacizumab can be used beyond progression in this indication, and whether or not treatment can be repeated, remains an intriguing question that is being addressed in the MITO16MANGO2b trial (NCT01802749); with studies in colorectal cancer15 and breast cancer16 having already provided evidence of benefit.

hether or not bevacizumab can be used beyond progression in this indication, and whether or not treatment can be repeated, remains an intriguing question that is being addressed in the MITO16MANGO2b trial (NCT01802749); with studies in colorectal cancer15 and breast cancer16 having already provided evidence of benefit. From a societal perspective, there are strong and sometimes opposing views about the costs and cost–benefit ratio of bevacizumab. The JGOG-316 trial17 reported a similar overall survival benefit without bevacizumab from the use of weekly paclitaxel, whereas GOG-26218 suggests that the effect of bevacizumab may be attenuated by such a strategy. Full economic analyses related to our trial are ongoing and will be reported separately. Interestingly, the small reduction in quality of life associated with bevacizumab that was reported after 54 weeks19 was smaller still by week 76 and was not statistically significant. The ability to predict which patients will benefit most is clearly important and could have the power to change the cost-effectiveness of treatment substantially by not treating patients with little chance of benefit. Our data suggest a simple and pragmatic clinical algorithm based on residual disease. Many studies are underway to identify a biomarker signature of response or resistance in patients in our trial. Collinson and colleagues20 and Backen and colleagues21 have presented biomarker strategies with potentially predictive approaches, whereas Gourley and colleagues22 have reported that bevacizumab might disadvantage women with an immunologically active subtype and Winterhoff and coworkers have reported benefit for women with mesenchymal-subtype disease.23 These findings all need to be validated in independent datasets.

ntially predictive approaches, whereas Gourley and colleagues22 have reported that bevacizumab might disadvantage women with an immunologically active subtype and Winterhoff and coworkers have reported benefit for women with mesenchymal-subtype disease.23 These findings all need to be validated in independent datasets. The addition of bevacizumab to chemotherapy is an important step forward in integrating biological agents with conventional chemotherapy in ovarian cancer. This trial provides evidence of a benefit in poor-prognosis patients. Future studies will refine important questions of biological prediction, duration, and rechallenge. Supplementary Material Supplementary appendix Acknowledgments We thank the women who participated in the trial and their families. The study was supported by Roche and the National Institute for Health Research, through the National Cancer Research Network. Roche also undertook regulatory filing of bevacizumab with progression-free survival data, leading to its widespread international use. Contributors AMO, TJP, MKBP, AMS, GCJ (translational research), and DS (quality of life) led and coordinated the study design. JP, JAL, EP-L, GK, MSC, PB, AC, T-WP-S, GR, FJ, MRM, MP, MQ, AP, and LF led recruitment and data collection. ADC, AE, and MKBP analysed the data, which was interpreted by all coauthors. AMO, TJP, ADC, MKBP, and RK drafted the report with input from all coauthors. All authors have seen and approved the final report.

gn. JP, JAL, EP-L, GK, MSC, PB, AC, T-WP-S, GR, FJ, MRM, MP, MQ, AP, and LF led recruitment and data collection. ADC, AE, and MKBP analysed the data, which was interpreted by all coauthors. AMO, TJP, ADC, MKBP, and RK drafted the report with input from all coauthors. All authors have seen and approved the final report. Declaration of interests GCJ is chair of the Medical Research Council translational research group for ICON7 which received Roche funding. EP-L, PB, AC, T-WP-S, FJ, and MQ have received personal fees from Roche outside the submitted work. GR and TJP have received personal fees from Roche during the conduct of the study. PB has received grant funding from Roche outside the submitted work. RK, MKBP, and TJP have received grant funding from Roche during the study period. ADC and T-WP-S have received non-financial support from Roche, outside the submitted work. MKBP and TJP have received non-financial support from Roche during the study period. T-WP-S has received personal fees from Novartis and AstraZeneca outside the submitted work. GR has received personal fees from Oxigene, Amgen, and Novartis outside the submitted work. MQ has received personal fees from Tesaro outside the submitted work. TJP has received personal fees from Novartis and AstraZeneca outside the submitted work, and has received other fees from several other pharmaceutical companies and research organisations in connection with the costs of running a broad research portfolio, outside the submitted work. AMO, JP, AE, JAL, GK, MSC, MRM, MP, AP, DS, AMS, and LF declare no competing interests.

traZeneca outside the submitted work, and has received other fees from several other pharmaceutical companies and research organisations in connection with the costs of running a broad research portfolio, outside the submitted work. AMO, JP, AE, JAL, GK, MSC, MRM, MP, AP, DS, AMS, and LF declare no competing interests. Figure 1 Trial profile *Includes 16 patients last seen more than 6 months before the end of the study. †Includes 11 patients last seen more than 6 months before the end of the study. Figure 2 Overall survival (A) Overall survival in all patients. (B) Difference in overall survival between all patients in the two groups. (C) Overall survival in high-risk patients. (D) Difference in overall survival between high-risk patients in the two groups. Figure 3 Treatment effect on overall survival by disease status at enrolment Table 1 Primary analysis of overall survival and updated analysis of progression-free survival

(A) Overall survival in all patients. (B) Difference in overall survival between all patients in the two groups. (C) Overall survival in high-risk patients. (D) Difference in overall survival between high-risk patients in the two groups. Figure 3 Treatment effect on overall survival by disease status at enrolment Table 1 Primary analysis of overall survival and updated analysis of progression-free survival All patients High-risk patients Standard therapy (n=764) Bevacizumab (n=764) Standard therapy (n=254) Bevacizumab (n=248) Overall survival Follow-up duration (months) 48·6 (24·3–56·0) 48·8 (28·2–56·4) 29·0 (14·1–50·7) 38·9 (21·1–52·5) Deaths 352 (46%) 362 (47%) 174 (69%) 158 (64%) Median overall survival (months; 95% CI) 58·6 (53·5–67·5) 58·0 (52·4–66·9) 30·2 (27·0–34·3) 39·7 (36·0–44·2) Log-rank test p value p=0·85 p=0·03 HR (95% CI) 0·99 (0·85–1·14) 0·78 (0·63–0·97) Non-proportionality p value* p=0·02 p=0·01 (Restricted) mean survival time (months; 95% CI)† 44·6 (43·2–45·9) 45·5 (44·2–46·7) 34·5 (32·0–37·0) 39·3 (37·0–41·7) Restricted mean survival time difference (95% CI) 0·9 (−0·8 to 2·6) 4·8 (1·5–8·1) Progression-free survival Follow-up duration (months) 16·3 (8·8–48·4) 19·4 (12·7–45·3) 10·1 (7·7–18·2) 15·6 (9·9–21·7) Disease progression 526 (74%) 554 (73%) 228 (90%) 223 (90%) Median progression-free survival (months; 95% CI) 17·5 (15·7–18·7) 19·9 (19·1–22·0) 10·5 (9·3–12·0) 16·0 (14·2–17·8) Log-rank test p value p=0·25 p=0·001 HR (95% CI) 0·93 (0·83–1·05) 0·73 (0·61–0·88) Non-proportionality p value* p<0·0001 p<0·0001 (Restricted) mean survival time (months; 95% CI)† 27·7 (26·1–29·2) 29·2 (27·7–30·7) 15·9 (14·1–17·7) 20·0 (18·1–21·8) Restricted mean survival time difference (95% CI) 1·6 (−0·6 to 3·7) 4·1 (1·4–6·7) Data are median (IQR) or n (%), unless otherwise indicated. HRs, p values, and survival time differences are for differences between the standard therapy and bevacizumab groups. HR=hazard ratio.

7·7–30·7) 15·9 (14·1–17·7) 20·0 (18·1–21·8) Restricted mean survival time difference (95% CI) 1·6 (−0·6 to 3·7) 4·1 (1·4–6·7) Data are median (IQR) or n (%), unless otherwise indicated. HRs, p values, and survival time differences are for differences between the standard therapy and bevacizumab groups. HR=hazard ratio. * Grambsch-Therneau test. † Restricted at 5 years. Table 2 Overall survival in predefined subgroups Clear cell tumours* Low-stage high-grade tumours Low-grade serous tumours Standard therapy (n=77) Bevacizumab (n=82) Standard therapy (n=75) Bevacizumab (n=67) Standard therapy (n=49) Bevacizumab (n=31) Follow-up duration (months) 52·5 (29·0–57·5) 50·7 (28·2–57·9) 55·3 (49·1–60·6) 55·4 (51·2–61·6) 50·5 (28·2–55·1) 55·3 (47·9–62·0) Deaths 20 (26%) 24 (29%) 6 (8%) 9 (13%) 13 (27%) 7 (23%) Log-rank test p value p=0·74 p=0·44 p=0·60 HR (95% CI) 1·09 (0·64–1·88) 1·49 (0·53–4·20) 0·78 (0·31–1·97) Non-proportionality p value† p=0·58 p=0·002 p=0·07 (Restricted) mean survival time (months; 95% CI)‡ 48·0 (43·9–52·2) 47·6 (43·6–51·6) 56·2 (51·5–60·9) 57·5 (55·7–59·4) 50·4 (45·6–55·2) 50·5 (43·9–57·0) Restricted mean survival time difference (95% CI) −0·4 (−6·1 to 5·3) 1·3 (−3·7 to 6·4) 0·1 (−7·9 to 8·0) Data are median (IQR) or n (%), unless otherwise indicated. HRs, p values, and survival time differences are for differences between the standard therapy and bevacizumab groups. * The clear cell tumour group includes some patients with mixed histology. † Grambsch-Therneau test. ‡ Restricted at 5 years.

Research in context Evidence before this study Randomised controlled trials of moderately hypofractionated radiotherapy schedules versus conventionally fractionated radiotherapy for localised prostate cancer, using both older and more modern radiotherapy techniques, have shown inconsistent results for both efficacy and side-effects. These studies have not usually included health-related quality of life or patient-reported outcomes (PROs), which detect more side-effects than do clinician-reported outcomes. We searched PubMed using the terms “patient-reported outcomes” OR “quality of life” AND “hypofractionated” OR “hypofractionation” AND “prostate” up to Oct 1, 2002, and retrieved eight articles. Of these articles, none reported PROs from randomised trials of conventional versus hypofractionated radiotherapy. Added value of this study To our knowledge, this study is the largest randomised trial of moderately hypofractionated versus conventionally fractionated radiotherapy using modern radiotherapy techniques, and the first to report PROs up to 2 years after treatment, showing both early and late developing side-effects. Other randomised trials that have included PROs and have been reported since this study began used older radiotherapy techniques, or only included follow-up to 3 months after radiotherapy, therefore assessing early rather than late treatment effects, which are usually dose-limiting. Implications of all the available evidence

To our knowledge, this study is the largest randomised trial of moderately hypofractionated versus conventionally fractionated radiotherapy using modern radiotherapy techniques, and the first to report PROs up to 2 years after treatment, showing both early and late developing side-effects. Other randomised trials that have included PROs and have been reported since this study began used older radiotherapy techniques, or only included follow-up to 3 months after radiotherapy, therefore assessing early rather than late treatment effects, which are usually dose-limiting. Implications of all the available evidence If efficacy outcomes from CHHiP show non-inferiority for hypofractionated treatments, the absence of any difference in PROs between trial groups adds to the growing evidence for moderately hypofractionated radiotherapy schedules becoming the standard treatment for localised prostate cancer. Introduction Prostate cancer is the most common cancer in men in the UK, with 41 700 patients diagnosed in 2011.1 For patients diagnosed with localised disease, external beam radiotherapy, radical prostatectomy, and brachytherapy are conventional treatments with similar control rates for organ-confined tumours. Management choices are therefore often affected by potential treatment-related toxic effects. Patient-reported outcomes (PROs) detect treatment side-effects more reliably than do clinician-reported measures and might better guide treatment decisions.2, 3

treatments with similar control rates for organ-confined tumours. Management choices are therefore often affected by potential treatment-related toxic effects. Patient-reported outcomes (PROs) detect treatment side-effects more reliably than do clinician-reported measures and might better guide treatment decisions.2, 3 The Conventional or Hypofractionated High Dose Intensity Modulated Radiotherapy in Prostate Cancer (CHHiP) trial (CRUK/06/016) randomly assigned men with localised prostate cancer who were undergoing radiotherapy to a standard fractionation schedule or to one of two hypofractionated regimens. The main aims of the trial were to compare the efficacy and toxic effects of conventional and hypofractionated radiotherapy. Quality of life (QoL) was assessed in a substudy within the main trial, in which we aimed to assess whether PROs differed between patients receiving conventionally fractionated versus hypofractionated radiotherapy up to 24 months after radiotherapy. Methods Study design and participants CHHiP was a randomised, non-inferiority phase 3 trial done in three seamless stages. Participation in the QoL substudy was open to all UK centres participating in the main trial. Because it recruited ahead of schedule, the QoL substudy closed to accrual before the main trial closed.

Methods Study design and participants CHHiP was a randomised, non-inferiority phase 3 trial done in three seamless stages. Participation in the QoL substudy was open to all UK centres participating in the main trial. Because it recruited ahead of schedule, the QoL substudy closed to accrual before the main trial closed. Men older than 16 years who had histologically confirmed T1b–T3aN0M0 prostate cancer and a WHO performance status 0 or 1 were eligible for participation. Initially, men with a prostate-specific antigen (PSA) concentration of less than 40 ng/mL and risk of lymph node involvement less than 30% were eligible; on Aug 1, 2006, these criteria were revised and a PSA concentration less than 30 ng/mL and a risk of seminal vesicle involvement less than 30% were needed. Patients were ineligible if they had T3 tumours and a Gleason score of 8 or higher, or a life expectancy of less than 10 years. Full details of trial design, eligibility, and treatment have been reported previously.4 The study was approved by the London Multi-centre Research Ethics Committee (04/MRE02/10). It was sponsored by the Institute of Cancer Research and was done in accordance with the principles of good clinical practice. All patients provided written informed consent. The Institute of Cancer Research Clinical Trials and Statistics Unit (ICR-CTSU; Sutton, UK) coordinated the study and carried out central statistical data monitoring and all analyses. The trial management group was overseen by an independent trial steering committee.

ice. All patients provided written informed consent. The Institute of Cancer Research Clinical Trials and Statistics Unit (ICR-CTSU; Sutton, UK) coordinated the study and carried out central statistical data monitoring and all analyses. The trial management group was overseen by an independent trial steering committee. Randomisation and masking Men were registered in the trial before or after starting initial hormone therapy. 4–6 weeks before radiotherapy, participants were randomly assigned (1:1:1) to receive a standard fractionation (control) or one of two hypofractionated schedules. Randomisation was done centrally via telephone calls to the ICR-CTSU. Computer-generated random permuted blocks were used, with block sizes of six and nine. Patients were stratified by centre and National Comprehensive Cancer Network (NCCN) risk group. Neither treatment allocation nor clinical assessment were masked because sham radiotherapy was not given.

elephone calls to the ICR-CTSU. Computer-generated random permuted blocks were used, with block sizes of six and nine. Patients were stratified by centre and National Comprehensive Cancer Network (NCCN) risk group. Neither treatment allocation nor clinical assessment were masked because sham radiotherapy was not given. Procedures Men with NCCN intermediate-risk or high-risk disease received short-course androgen suppression for 3–6 months before and during radiotherapy; this was optional for patients with low-risk disease. Individuals assigned to the control group received standard radiotherapy with 2 Gy daily fractions (Monday to Friday treatment) for 7·4 weeks, to give a total dose of 74 Gy in 37 fractions. Individuals in the experimental groups received hypofractionated treatment with 3 Gy daily fractions to a total dose of either 60 Gy in 20 fractions in 4·0 weeks or 57 Gy in 19 fractions in 3·8 weeks. For the hypofractionated schedules, the protocol stated that the overall duration of treatment should be at least 28 days for the 20-fraction schedule and at least 27 days for the 19-fraction schedule. This was to avoid undue shortening of overall treatment time and, in practice, meant that treatment started on a Wednesday to Friday. Forward or inverse three-dimensional methods were used to plan radiotherapy treatment. Further details of treatment and quality assurance have been reported previously.4

action schedule. This was to avoid undue shortening of overall treatment time and, in practice, meant that treatment started on a Wednesday to Friday. Forward or inverse three-dimensional methods were used to plan radiotherapy treatment. Further details of treatment and quality assurance have been reported previously.4 Men consenting to participate in the QoL substudy were eligible to complete questionnaires at trial entry if they had not already started endocrine treatment, to minimise the effect of toxicity of hormone deprivation on QoL at this timepoint. All men were eligible to complete further questionnaires pre-radiotherapy, and at 10 weeks and 6, 12, 18, and 24 months after the start of radiotherapy. From trial entry to 6 months after radiotherapy, questionnaires were administered in the clinic, and subsequent questionnaires were posted to patients from the ICR-CTSU after local verification of their current health status. All QoL questionnaires were self-administered.

and 24 months after the start of radiotherapy. From trial entry to 6 months after radiotherapy, questionnaires were administered in the clinic, and subsequent questionnaires were posted to patients from the ICR-CTSU after local verification of their current health status. All QoL questionnaires were self-administered. During the planning stages of the CHHiP trial, the University of California, Los Angeles Prostate Cancer Index (UCLA-PCI) was an important QoL instrument available for use in patients with localised prostate cancer.5 Subsequently it became apparent that this instrument needed augmentation to better capture the broad range of urinary, bowel, sexual, and hormonal symptoms in patients receiving external beam radiotherapy or brachytherapy, or undergoing radical prostatectomy. Consequently the Expanded Prostate Cancer Index Composite (EPIC) QoL instrument was developed that had item content that better represented typical symptoms after radiotherapy.6 To maximise the sensitivity of the PROs, the QoL instruments were updated during the trial to include the EPIC instrument. Therefore from trial initiation to early 2009, the UCLA-PCI, including the Short Form 36 (SF-36) and Functional Assessment of Cancer Therapy-Prostate (FACT-P) QoL instruments were used.7 Following a protocol amendment on March 12, 2009, the EPIC and Short Form 12 (SF-12) QoL instruments replaced UCLA-PCI, SF-36, and FACT-P, although some old questionnaires were received back from participants after this date.8 EPIC-50 was used for bowel and urinary domains and EPIC-26 for sexual and hormonal domains.9

Following a protocol amendment on March 12, 2009, the EPIC and Short Form 12 (SF-12) QoL instruments replaced UCLA-PCI, SF-36, and FACT-P, although some old questionnaires were received back from participants after this date.8 EPIC-50 was used for bowel and urinary domains and EPIC-26 for sexual and hormonal domains.9 UCLA-PCI consists of 20 items organised into six domains, including bowel function (four items), bowel bother (one item), urinary function (five items), urinary bother (one item), sexual function (eight items), and sexual bother (one item). EPIC-50 includes a bowel function domain (seven items) and a bowel bother domain (seven items), which together form the bowel summary domain, and a urinary function domain (five items) and a urinary bother domain (seven items), which together form a urinary summary domain. EPIC-26 includes a sexual function domain (five items) and a sexual bother domain (one item), which are combined to form a sexual summary domain; there is also a hormonal domain (five items). Items in the UCLA-PCI and EPIC QoL instruments differed: for example, the EPIC bowel function domain included rectal bleeding, faecal incontinence, and daily bowel movements, which were absent from UCLA-PCI, whereas bowel distress was represented in the UCLA-PCI bowel function domain and absent from EPIC-50. Additionally, haematuria and dysuria were represented in the EPIC urinary function domain, but absent from UCLA-PCI. All QoL instrument scores range from 0 to 100 and a higher score represents better QoL.

rom UCLA-PCI, whereas bowel distress was represented in the UCLA-PCI bowel function domain and absent from EPIC-50. Additionally, haematuria and dysuria were represented in the EPIC urinary function domain, but absent from UCLA-PCI. All QoL instrument scores range from 0 to 100 and a higher score represents better QoL. Health-related QoL was assessed using the FACT-P and SF-36 instruments (with UCLA-PCI) or the SF-12 instrument (with EPIC). FACT-P consists of physical, social, functional, and emotional wellbeing domains; each domain has seven items, and scores per domain range from 0 to 28, except for emotional wellbeing, which ranges from 0 to 24. The SF-36 instrument includes eight domains of physical functioning, social functioning, vitality, role limitations (physical), role limitations (emotional), mental health, general health, and bodily pain, and each domain score ranges from 0 to 100. SF-12 consists of a physical composite score (PCS) and mental composite score (MCS), each of which have six items, and scores range from 0 to 100. For all three instruments, a higher score represents better quality of life. All questionnaires were scored in accordance with the recommended scoring manuals.10, 11, 12, 13, 14 Not all respondents answered all questions; all available data points were used in analyses. Separate QoL analyses were planned after 2 and 5 years of follow-up, and this report describes PROs up to 2 years.

Health-related QoL was assessed using the FACT-P and SF-36 instruments (with UCLA-PCI) or the SF-12 instrument (with EPIC). FACT-P consists of physical, social, functional, and emotional wellbeing domains; each domain has seven items, and scores per domain range from 0 to 28, except for emotional wellbeing, which ranges from 0 to 24. The SF-36 instrument includes eight domains of physical functioning, social functioning, vitality, role limitations (physical), role limitations (emotional), mental health, general health, and bodily pain, and each domain score ranges from 0 to 100. SF-12 consists of a physical composite score (PCS) and mental composite score (MCS), each of which have six items, and scores range from 0 to 100. For all three instruments, a higher score represents better quality of life. All questionnaires were scored in accordance with the recommended scoring manuals.10, 11, 12, 13, 14 Not all respondents answered all questions; all available data points were used in analyses. Separate QoL analyses were planned after 2 and 5 years of follow-up, and this report describes PROs up to 2 years. Outcomes The primary endpoint was the single item “Overall how much of a problem have your bowels been for you during the last 4 weeks” (overall bowel bother). This question was chosen because it gives a good overall measure of bowel-associated morbidity and is common to both the UCLA-PCI and EPIC instruments, so was reported by all patients. The main secondary endpoints were overall urinary bother and overall sexual bother. Additional secondary endpoints were individual bowel, urinary, and sexual items and domain scores assessed within EPIC and UCLA-PCI and the general health-related QoL domain scores in FACT-P, SF-36, and SF-12.

reported by all patients. The main secondary endpoints were overall urinary bother and overall sexual bother. Additional secondary endpoints were individual bowel, urinary, and sexual items and domain scores assessed within EPIC and UCLA-PCI and the general health-related QoL domain scores in FACT-P, SF-36, and SF-12. Statistical analysis For all endpoints, the control group was compared with each of the experimental groups, as per the statistical analysis plan of the main trial. Formal statistical tests were done at 24 months. After this analysis, a post-hoc pragmatic comparison was done between the 60 Gy and 57 Gy experimental schedules to support clinical management choices if both hypofractionated schedules were shown to be non-inferior to standard fractionation for disease control. With 443 patients per experimental group, this substudy would have 80% power and 2·5% two-sided significance to detect changes in the proportion of patients with overall bowel bother scores as follows: from 65% in the standard fractionation group to 60% in an experimental hypofractionation group for scores of 1 (no bother), from 22% to 20% for scores of 2 (very small bother), from 7% to 10% for scores of 3 (small bother), and from 6% to 10% for scores of 4 or 5 (moderate or severe bother). Because the trial was not originally powered for QoL analyses, these calculations were done retrospectively, but before any analysis occurred. Power calculations were based on comparisons of two independent groups of ordered categorical data, with constant odds ratios computed across all categories, and assumed complete data would be available for 70% of patients (1330 individuals) at 2 years. A significance level of 0·001 and 99% confidence intervals were used, guided by a Bonferroni adjustment, to make some allowance for multiple testing.

ategorical data, with constant odds ratios computed across all categories, and assumed complete data would be available for 70% of patients (1330 individuals) at 2 years. A significance level of 0·001 and 99% confidence intervals were used, guided by a Bonferroni adjustment, to make some allowance for multiple testing. We did cross-sectional, time-to-event, and change-from-baseline analyses. Cross-sectional analysis was done at each timepoint, with formal comparisons between treatment groups at 24 months via the χ2 test for trend and the Mann-Whitney U test. We combined moderate and severe events for the formal comparisons because of the small number of severe events. We did time-to-event analysis using Kaplan-Meier methods and the log-rank test to assess time to small or worse, and moderate or worse events for individual items. This analysis aimed to detect differences in late radiation toxic effects between treatment groups and therefore did not include the 10-week post-radiotherapy assessment, which assessed acute symptoms. Time-to-event was therefore measured from the start of radiotherapy to the QoL assessments at 6, 12, 18, or 24 months. Patients who reached the relevant endpoint at trial entry or pre-radiotherapy were excluded from that specific time-to-event analysis.

-week post-radiotherapy assessment, which assessed acute symptoms. Time-to-event was therefore measured from the start of radiotherapy to the QoL assessments at 6, 12, 18, or 24 months. Patients who reached the relevant endpoint at trial entry or pre-radiotherapy were excluded from that specific time-to-event analysis. We assessed change from baseline (post-radiotherapy score minus baseline score) to account for differences in pre-existing comorbidity between groups. For bowel and urinary items and domain scores, we used the pre-radiotherapy score as a surrogate baseline score unless it was missing, in which case the baseline score was used. We used this surrogate to maximise numbers. To assess the sensitivity of the results to this assumption, analyses were repeated using baseline data only. For sexual endpoints, only the baseline score at trial entry was used. A sensitivity analysis was also done to confirm the robustness of including the baseline assessments of patients receiving less than 1 month of endocrine treatment which involved repeating analyses using baseline assessments of patients receiving no endocrine therapy beforehand.

baseline score at trial entry was used. A sensitivity analysis was also done to confirm the robustness of including the baseline assessments of patients receiving less than 1 month of endocrine treatment which involved repeating analyses using baseline assessments of patients receiving no endocrine therapy beforehand. We modelled the odds of any specific change from baseline or pre-radiotherapy to 24 months using ordinal logistic regression after checking the validity of the proportional odds assumption.15 Odds ratios less than one favour the relevant experimental group. For the ordinal logistic regression models, the dependent variable is the post-radiotherapy score minus the baseline or pre-radiotherapy score, taking values of −4, −3, −2, −1, 0, 1, 2, 3, or 4, where negative numbers represent an improvement in QoL and positive numbers represent worsening QoL. We used ANCOVA modelling to assess change from baseline for continuous variables such as domain scores, adjusting for baseline or pre-radiotherapy score as indicated above. We assessed the normality assumption of the ANCOVA model visually via histograms and we did not deem formal tests to be necessary. Patients were excluded from the fixed timepoint analyses if their QoL assessments were dated outside prespecified acceptable time intervals, as outlined in figure 1.

ore as indicated above. We assessed the normality assumption of the ANCOVA model visually via histograms and we did not deem formal tests to be necessary. Patients were excluded from the fixed timepoint analyses if their QoL assessments were dated outside prespecified acceptable time intervals, as outlined in figure 1. No imputation of missing PRO data was done. For missing individual items, domain scores were only calculated if sufficient items were completed in accordance with the relevant scoring manual. For absent whole instruments, the effect of these missing data was assessed by comparison of the baseline characteristics of patients present in the analysis versus, first, those who consented but were missing entirely, and second, those who consented but were missing at 24 months, when formal statistical testing was done. Analysis was on an intention-to-treat basis and all analyses were done with Stata version 13.1. The CHHiP trial is registered as an International Standard Randomised Controlled Trial, number ISRCTN97182923. Role of the funding source The funders provided peer-reviewed approval for the study concept but had no role in study design, data collection, data analysis, data interpretation, or writing of the report. AW, HM, CG, and EH had access to all the raw data. The lead and corresponding authors had full access to all data in the study and had final responsibility for the decision to submit for publication.

concept but had no role in study design, data collection, data analysis, data interpretation, or writing of the report. AW, HM, CG, and EH had access to all the raw data. The lead and corresponding authors had full access to all data in the study and had final responsibility for the decision to submit for publication. Results Between Oct 18, 2002, and Nov 1, 2009, 2100 patients were recruited from 57 centres in the UK (figure 1 and appendix p 14) into the QoL substudy of the CHHiP trial; subsequently, the substudy closed to accrual. 696 patients were assigned to the standard 74 Gy schedule, 698 were assigned to the 60 Gy schedule, and 706 were assigned to the 57 Gy schedule. Median follow-up was 50·0 months (IQR 38·4–64·2) on April 9, 2014, which was the most recent follow-up measurement of all data collected before the QoL data snapshot for this analysis in September, 2014. At trial entry, 700 (86%) of the 812 eligible patients who had not started endocrine treatment plus 252 patients who had started endocrine treatment returned questionnaires. Questionnaires were returned by 1659 (79%) patients pre-radiotherapy, 1470 (70%) patients at 10 weeks, 1597 (76%) patients at 6 months, 1551 (74%) patients at 12 months, 1456 (69%) patients at 18 months, and 1444 (69%) patients at 24 months. Baseline characteristics of patients were balanced between treatment groups except for an imbalance in T stage between the 74 Gy and 60 Gy groups (table 1). 1490 (71%) patients had intermediate NCCN risk disease.16

Median follow-up was 50·0 months (IQR 38·4–64·2) on April 9, 2014, which was the most recent follow-up measurement of all data collected before the QoL data snapshot for this analysis in September, 2014. At trial entry, 700 (86%) of the 812 eligible patients who had not started endocrine treatment plus 252 patients who had started endocrine treatment returned questionnaires. Questionnaires were returned by 1659 (79%) patients pre-radiotherapy, 1470 (70%) patients at 10 weeks, 1597 (76%) patients at 6 months, 1551 (74%) patients at 12 months, 1456 (69%) patients at 18 months, and 1444 (69%) patients at 24 months. Baseline characteristics of patients were balanced between treatment groups except for an imbalance in T stage between the 74 Gy and 60 Gy groups (table 1). 1490 (71%) patients had intermediate NCCN risk disease.16 For 46 patients reported to have consented to enter the substudy, no QoL assessments were received by the ICR-CTSU. Baseline characteristics for these patients were not significantly different from those of patients present in the analysis (appendix p 1). 828 (39%) patients who had consented to participate in the substudy had no QoL assessments available at 24 months. The only significant difference in baseline characteristics between groups with and without 24-month PRO data was that patients with missing questionnaires were more likely to have high NCCN risk disease at trial entry than were patients who provided data (appendix p 2). Of patients with data from at least one QoL assessment, 665 (98%) of 676 in the 74 Gy treatment group, 674 (98%) of 686 in the 60 Gy treatment group, and 677 (98%) of 692 in the 57 Gy treatment group received endocrine therapy.

e high NCCN risk disease at trial entry than were patients who provided data (appendix p 2). Of patients with data from at least one QoL assessment, 665 (98%) of 676 in the 74 Gy treatment group, 674 (98%) of 686 in the 60 Gy treatment group, and 677 (98%) of 692 in the 57 Gy treatment group received endocrine therapy. The incidence of overall bowel bother was low (figure 2, appendix p 3). At 24 months post-radiotherapy, we recorded no overall bowel bother for 269 (66%) of 410 men treated with 74 Gy, 266 (65%) of 411 men treated with 60 Gy, and 282 (65%) of 437 men treated with 57 Gy; very small bother for 92 (22%), 91 (22%), and 93 (21%) men; small bother for 26 (6%), 28 (7%), and 38 (9%) men; moderate bother for 19 (5%), 23 (6%), 21 (5%) men; and severe bother for four (<1%), three (<1%), and three (<1%) men, respectively. Cross-sectional analysis at 24 months showed no significant differences in overall bowel bother between the treatment groups (74 Gy vs 60 Gy, ptrend=0·64; 74 Gy vs 57 Gy, ptrend=0·59; appendix p 3).

ther for 19 (5%), 23 (6%), 21 (5%) men; and severe bother for four (<1%), three (<1%), and three (<1%) men, respectively. Cross-sectional analysis at 24 months showed no significant differences in overall bowel bother between the treatment groups (74 Gy vs 60 Gy, ptrend=0·64; 74 Gy vs 57 Gy, ptrend=0·59; appendix p 3). A temporary increase in any bowel bother was seen at 10 weeks (a change from 413 [27%] of 1509 patients pre-radiotherapy to 745 [57%] of 1309 patients at 10 weeks). At 6 months, any bowel bother had decreased (581 [38%] of 1519 patients), and remained around this level to 24 months, when any overall bowel bother was reported by 441 (35%) of 1258 patients. A sensitivity analysis showed that using pre-radiotherapy scores as a surrogate for baseline scores at trial entry for some patients was valid (appendix p 11–12). Because 252 men completed baseline questionnaires after starting endocrine treatment, a sensitivity analysis was done, which confirmed the robustness of including patients receiving less than 1 month of endocrine treatment at baseline (appendix p 13).

cores at trial entry for some patients was valid (appendix p 11–12). Because 252 men completed baseline questionnaires after starting endocrine treatment, a sensitivity analysis was done, which confirmed the robustness of including patients receiving less than 1 month of endocrine treatment at baseline (appendix p 13). The pattern in overall urinary bother was similar to that for overall bowel bother (figure 2, appendix p 4) and cross-sectional analysis at 24 months showed no significant differences in overall urinary bother between treatment groups (appendix p 4). The baseline incidence of overall sexual bother was higher than that for bowel or urinary bother, with 412 (57%) of 719 patients having any bother at baseline, which increased to 975 (68%) of 1440 patients pre-radiotherapy and improved from 6 months to 24 months (figure 2, appendix p 4). There were no significant differences between treatment groups At 24 months, we noted no significant differences between treatment groups for all other individual bowel, urinary, and sexual items assessed (appendix pp 3–4). Bowel, urinary, and sexual domain scores assessed within UCLA-PCI or EPIC QoL instruments also showed no difference between treatments at 24 months (table 2).

The pattern in overall urinary bother was similar to that for overall bowel bother (figure 2, appendix p 4) and cross-sectional analysis at 24 months showed no significant differences in overall urinary bother between treatment groups (appendix p 4). The baseline incidence of overall sexual bother was higher than that for bowel or urinary bother, with 412 (57%) of 719 patients having any bother at baseline, which increased to 975 (68%) of 1440 patients pre-radiotherapy and improved from 6 months to 24 months (figure 2, appendix p 4). There were no significant differences between treatment groups At 24 months, we noted no significant differences between treatment groups for all other individual bowel, urinary, and sexual items assessed (appendix pp 3–4). Bowel, urinary, and sexual domain scores assessed within UCLA-PCI or EPIC QoL instruments also showed no difference between treatments at 24 months (table 2). Time-to-event analysis of small or worse overall bowel, urinary, and sexual bother showed no significant differences between treatment groups for any endpoints (figure 2). The appendix contains absolute numbers of cumulative small or worse and moderate or worse events, the prevalence of the relevant bowel, urinary, and sexual symptoms before radiotherapy, and the hazard ratios for the time-to-event analysis time from start of radiotherapy to small or worse and moderate or worse events for all individual items in all treatment groups (appendix pp 5 and 6). The number of patients reporting symptoms that were represented only in EPIC (faecal incontinence, rectal bleeding, daily bowel movements, dysuria, and haematuria) was considerably lower than that for other endpoints, so these analyses were underpowered.

l individual items in all treatment groups (appendix pp 5 and 6). The number of patients reporting symptoms that were represented only in EPIC (faecal incontinence, rectal bleeding, daily bowel movements, dysuria, and haematuria) was considerably lower than that for other endpoints, so these analyses were underpowered. Although we saw no significant differences between treatment groups, the cumulative incidence of some symptoms, including faecal incontinence, rectal bleeding, and use of urinary pads, was higher in patients treated with hypofractionated radiation than in those treated with standard fractionation (appendix p 5). However, at 24 months, differences in the prevalence of these symptoms between groups were smaller (appendix pp 3–4). Figure 3 shows change from baseline for UCLA-PCI domain scores and EPIC domain summary scores; additional EPIC domain scores are shown in the appendix (p 10). There were no significant differences between treatment groups. For all urinary and bowel items and domain scores, to maximise numbers, the pre-radiotherapy score was used as a surrogate baseline score unless missing, in which case the baseline score was used; exact numbers are: 749 pre-radiotherapy plus 65 baseline for change in UCLA-PCI bowel function to 24 months; 146 plus 14 for change in EPIC bowel summary to 24 months; 751 plus 64 for change in UCLA-PCI urinary function to 24 months; and 139 plus 16 for change in EPIC urinary summary to 24 months (numbers per treatment group shown in appendix p 14).

otherapy plus 65 baseline for change in UCLA-PCI bowel function to 24 months; 146 plus 14 for change in EPIC bowel summary to 24 months; 751 plus 64 for change in UCLA-PCI urinary function to 24 months; and 139 plus 16 for change in EPIC urinary summary to 24 months (numbers per treatment group shown in appendix p 14). Figure 3 also shows change from baseline in scores for the individual items of overall bowel bother, overall urinary bother, and overall sexual bother; all other endpoints are shown in the appendix (p 8). Most patients had no change in score from baseline and we noted no significant differences between treatment groups in change from baseline to 24 months for any individual items. Compared to the 74 Gy control group, the odds of a one-point increase in overall bowel bother were reduced, although not significantly, in the 60 Gy treatment group (odds ratio [OR] 0·85 [99% CI 0·57–1·26]; p=0·29) and the 57 Gy treatment group (OR 0·84 [0·57–1·24]; p=0·25). For some endpoints, the odds of a patient developing side-effects were slightly increased for the 60 Gy group compared with the 57 Gy group, but none of these increases were significant (appendix p 8). Again for urinary and bowel overall bother items, the pre-radiotherapy score was used as a surrogate for baseline. Exact numbers are 971 pre-radiotherapy plus 85 baseline for overall bowel bother; 969 plus 84 for overall urinary bother (per treatment group shown in appendix p 14).

eases were significant (appendix p 8). Again for urinary and bowel overall bother items, the pre-radiotherapy score was used as a surrogate for baseline. Exact numbers are 971 pre-radiotherapy plus 85 baseline for overall bowel bother; 969 plus 84 for overall urinary bother (per treatment group shown in appendix p 14). We identified no significant differences in health-related QoL domain scores measured by FACT-P, SF-12, and SF-36 between treatment groups at 24 months (appendix p 9). For most domains, there was a consistent pattern of stable scores across all timepoints, including pre-radiotherapy and 10 weeks post-radiotherapy timepoints. However, for the SF-36 domains of vitality, physical role functioning, and social wellbeing, we noted a reduction of more than 10 points between the median domain score at baseline and at 10 weeks after the start of radiotherapy in all treatment groups. By 6 months, median scores had increased to within 10 points of the median scores at baseline for all three measures.

l role functioning, and social wellbeing, we noted a reduction of more than 10 points between the median domain score at baseline and at 10 weeks after the start of radiotherapy in all treatment groups. By 6 months, median scores had increased to within 10 points of the median scores at baseline for all three measures. Discussion In this QoL substudy of the CHHiP trial, PROs were not significantly different between treatment groups for any of the endpoints assessed. Both cross-sectional analysis at 24 months and time-to-event analysis suggest an overall pattern of low incidence of bowel and urinary toxic effects in all treatment groups. The QoL instruments used were sensitive to change because acute toxic effects were clearly distinguished using both individual items and bowel, urinary, and sexual domain scores. These acute toxic effects had a small and short-lived effect on general health-related QoL, especially the SF-36 domains of vitality, physical role functioning, and social wellbeing. Overall, changes from baseline to 24 months for urinary, bowel, and most general health-related QoL domains (except role limitations [physical]), were less than previously reported minimally important differences derived from longitudinal anchor-based methods.17 Although further development of toxic effects is possible after 2 years, recent studies have reported minimal change in late radiotherapy side-effects after 2 years following external beam radiation therapy.18 This suggests that 2 years is an appropriate endpoint for initial PRO reporting.

dinal anchor-based methods.17 Although further development of toxic effects is possible after 2 years, recent studies have reported minimal change in late radiotherapy side-effects after 2 years following external beam radiation therapy.18 This suggests that 2 years is an appropriate endpoint for initial PRO reporting. To our knowledge, this is the first large randomised trial of hypofractionated radiotherapy that used modern radiotherapy techniques to report PROs with follow-up to 24 months. Aluwini and colleagues19 reported preliminary results that included PROs up to 3 months in the HYPRO study,19 which included 820 patients. Radiotherapy doses were higher in HYPRO (standard fractionation of 39 fractions of 2 Gy in 8 weeks vs hypofractionation with 19 fractions of 3·4 Gy in 6·5 weeks) than in CHHiP. Combined clinician-reported outcomes and PROs showed similar acute genitourinary toxic effects between treatments, but increased acute gastrointestinal toxic effects with hypofractionation. A small (124 patients) randomised study of moderate hypofractionation (63 Gy in 20 fractions) versus conventional fractionation (76 Gy in 38 fractions) reported no difference in EPIC scores between treatment groups up to 3 months after radiotherapy.20

creased acute gastrointestinal toxic effects with hypofractionation. A small (124 patients) randomised study of moderate hypofractionation (63 Gy in 20 fractions) versus conventional fractionation (76 Gy in 38 fractions) reported no difference in EPIC scores between treatment groups up to 3 months after radiotherapy.20 The PROs in this study are broadly consistent with preliminary data for clinician-reported outcomes in the CHHiP trial,4 and the clinician-reported outcomes of a small (168 patients) phase 3 trial in Italy.21 Results from another randomised trial that included 203 patients showed a non-significant numerical increase in clinician-reported late gastrointestinal toxic effects with hypofractionation.22 However, the radiotherapy dose schedules used in these studies21, 22 differ substantially from those in CHHiP. So far, randomised trials21, 22, 23, 24, 25 reporting the effects of hypofractionation versus conventional fractionation have reported inconsistent results for side-effects and do not clearly show a difference in the rate of increase of genitourinary or gastrointestinal toxic effects between conventional and hypofractionated radiotherapy treatments.26 These findings emphasise the need for outcome data from large trials of hypofractionation that use modern radiotherapy techniques. Such studies to compare hypofractionated radiotherapy with standard fractionation, together with the clinician-reported outcomes from CHHiP, will help to confirm whether faecal incontinence, rectal bleeding, or use of urinary pads are more common at a dose of 3 Gy per fraction.

that use modern radiotherapy techniques. Such studies to compare hypofractionated radiotherapy with standard fractionation, together with the clinician-reported outcomes from CHHiP, will help to confirm whether faecal incontinence, rectal bleeding, or use of urinary pads are more common at a dose of 3 Gy per fraction. Findings from a trial of hypofractionation25 that included 303 patients raised concerns about increased urinary toxic effects after hypofractionated treatment in patients who had compromised urinary function before enrolment, as assessed by clinician-reported LENT and RTOG scores. A formal comparison restricted to patients with baseline dysfunction has not been done in our study, partly because obstructive and irritative symptoms, and consequently overall urinary dysfunction, are not well represented in the UCLA-PCI instrument. We plan to do a formal comparison between groups using clinician-reported outcomes from RTOG and LENT instruments in a separate study. However urinary co-morbidity was well balanced between treatment groups and change in urinary function from baseline did not differ between fractionation schedules. Furthermore, overall urinary bother seems to decrease during the 2 years after radiotherapy, which is consistent with findings from the RT01 dose escalation trial.27

inary co-morbidity was well balanced between treatment groups and change in urinary function from baseline did not differ between fractionation schedules. Furthermore, overall urinary bother seems to decrease during the 2 years after radiotherapy, which is consistent with findings from the RT01 dose escalation trial.27 Comparison of bowel bother and distress assessed using the UCLA-PCI instrument in both the 74 Gy group of CHHiP and the 74 Gy group of the RT01 trial,27 in which conventional radiotherapy planning techniques were used, suggests that patients benefit substantially from improved treatment methods that use intensity-modulated radiotherapy and the dose constraints used in CHHiP. In RT01, 27 (9%) of 289 patients reported moderate bowel bother and nine (3%) patients reported severe bother in the 74 Gy group at 24 months.28 This compares with 19 (5%) of 410 patients reporting moderate bother and four (<1%) patients reporting severe bother in the 74 Gy group in CHHiP at 24 months. Similarly, at 24 months, 34 (12%) of 288 patients in RT0128 versus 13 (4%) of 312 patients in CHHiP reported moderate bowel distress, and two (<1%) patients in RT0128 versus none in CHHiP reported severe bowel distress.

ate bother and four (<1%) patients reporting severe bother in the 74 Gy group in CHHiP at 24 months. Similarly, at 24 months, 34 (12%) of 288 patients in RT0128 versus 13 (4%) of 312 patients in CHHiP reported moderate bowel distress, and two (<1%) patients in RT0128 versus none in CHHiP reported severe bowel distress. Strengths of our study include the wide age range and large number of patients recruited from different parts of the UK. The use of different QoL instruments was a limitation of the analysis, because it meant that fewer patients reported some important radiotherapy-related toxic effects, including rectal bleeding and faecal incontinence, which were only represented in EPIC. Additionally, domain scores for UCLA-PCI and EPIC are not directly comparable, so separate reporting was necessary with smaller numbers than used for the primary endpoint. EPIC is now regarded as the QoL instrument of choice for localised prostate cancer;29 however, our trial was planned before it was widely available. To our knowledge, minimally clinically important differences have not been published for EPIC-50, but will be a valuable addition when available.

for the primary endpoint. EPIC is now regarded as the QoL instrument of choice for localised prostate cancer;29 however, our trial was planned before it was widely available. To our knowledge, minimally clinically important differences have not been published for EPIC-50, but will be a valuable addition when available. Patients might acclimatise to symptoms over time and therefore the most subjective endpoints, such as overall bowel, urinary, and sexual bother, might under-represent the actual toxicity at later timepoints. However, more objective PROs, including rectal bleeding, dysuria, and quality of erections showed similar patterns of toxic effects over time compared with overall bother items, suggesting that overall bother gives a reliable representation of patient experience. Bother items might incorporate a psychosocial component as well as actual functional change, but we believe that overall perception of toxic effects is a comprehensive and comprehensible endpoint in a randomised comparison of PROs. Patients with missing data at 24 months were more likely to be in a higher NCCN risk group than were those patients with data present. Biochemical recurrence did not exclude patients from the PRO substudy, but patients might have been less willing to complete the questionnaires after relapse. Both NCCN risk group and numbers of patients with missing data did not differ between treatment groups, therefore missing data are unlikely to have substantially biased the randomised comparisons.

ude patients from the PRO substudy, but patients might have been less willing to complete the questionnaires after relapse. Both NCCN risk group and numbers of patients with missing data did not differ between treatment groups, therefore missing data are unlikely to have substantially biased the randomised comparisons. PROs at 5 years will be important to confirm our findings. 5-year outcomes, together with longer-term clinician-reported outcomes, will help to elucidate whether late emergent differences exist between treatment groups. 5-year efficacy data from the CHHiP trial will be available in late 2015. Follow-up from complementary randomised studies is ongoing and together these will clarify the role of moderately hypofractionated radiotherapy treatment for localised prostate cancer. Radiotherapy treatments need to balance the potential increased efficacy of biologically increased doses with the risk of increased side-effects. So far, our results show that the bowel and urinary side-effects of moderate hypofractionation for prostate cancer delivered with modern radiotherapy techniques are low and similar to those of standard fractionation. Supplementary Material Supplementary appendix

PROs at 5 years will be important to confirm our findings. 5-year outcomes, together with longer-term clinician-reported outcomes, will help to elucidate whether late emergent differences exist between treatment groups. 5-year efficacy data from the CHHiP trial will be available in late 2015. Follow-up from complementary randomised studies is ongoing and together these will clarify the role of moderately hypofractionated radiotherapy treatment for localised prostate cancer. Radiotherapy treatments need to balance the potential increased efficacy of biologically increased doses with the risk of increased side-effects. So far, our results show that the bowel and urinary side-effects of moderate hypofractionation for prostate cancer delivered with modern radiotherapy techniques are low and similar to those of standard fractionation. Supplementary Material Supplementary appendix Acknowledgments This work was supported by Cancer Research UK (CRUK/06/16, C8262/A7253, C1491/A9895, C1491/A15955, SP2312/021), the Department of Health, the National Institute for Health Research Cancer Research Network, and NHS funding to the NIHR Biomedical Research Centre at the Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, London. We thank the patients, investigators, physicists, radiographers, and research support staff at the participating centres (appendix). We recognise the role of all the trials unit staff at Bob Champion Unit and at ICR-CTSU who contributed to the central coordination of the study. We would also like to thank the CHHiP Trial Management Group members, past and present, and the Independent Data Monitoring and Trial Steering Committees for overseeing the trial. Helen Patterson died in 2012 shortly after completion of recruitment to the trial and remains greatly missed by her colleagues.

f the study. We would also like to thank the CHHiP Trial Management Group members, past and present, and the Independent Data Monitoring and Trial Steering Committees for overseeing the trial. Helen Patterson died in 2012 shortly after completion of recruitment to the trial and remains greatly missed by her colleagues. Contributors AW did statistical analyses, data interpretation and wrote the report. HM did statistical analyses and data interpretation. DD, EH, IS, JS, VK, CS, HP, JG, CC, and CG are members of the CHHiP Trial Management Group responsible for the design and day-to-day oversight of the study and contributed to data interpretation. DD, IS, JS, VK, CS, HP, JG, DB, CP, JL, AB, ZM, MP, CE, MR, PK, and JMO'S were involved in patient recruitment and data collection. AG was responsible for data collection and study management at the Bob Champion Unit. CC was responsible for central study management and CG was the supervisory statistician at ICR-CTSU. DD is the CHHiP chief investigator. EH is responsible for central management of the trial at ICR-CTSU and for all statistical analyses. DD and EH are joint senior authors and contributed to manuscript writing. All authors reviewed the report.

central study management and CG was the supervisory statistician at ICR-CTSU. DD is the CHHiP chief investigator. EH is responsible for central management of the trial at ICR-CTSU and for all statistical analyses. DD and EH are joint senior authors and contributed to manuscript writing. All authors reviewed the report. Declaration of interests CP reports grants and personal fees from Bayer, personal fees from Janssen, and personal fees from BNIT. JS reports personal fees from Janssen-Cilag Limited and personal fees from Bayer. DD reports grants from Cancer Research UK, during the conduct of the study. EH reports grants from Cancer Research UK, during the conduct of the study; and a grant from Accuracy Inc. to the Institute of Cancer Research to support independent statistical analysis of a phase 3 trial of sterotatic body radiotherapy in prostate cancer, outside the submitted work. VK reports advisory and educational fees and non-financial support from Astellas, educational fees from Bayer, non-financial educational support from Janssen, advisory and educational fees and non-financial support from Ipsen, and educational fees from Tolmar. All other authors declare no competing interests. Figure 1 Quality-of-life study profile

Declaration of interests CP reports grants and personal fees from Bayer, personal fees from Janssen, and personal fees from BNIT. JS reports personal fees from Janssen-Cilag Limited and personal fees from Bayer. DD reports grants from Cancer Research UK, during the conduct of the study. EH reports grants from Cancer Research UK, during the conduct of the study; and a grant from Accuracy Inc. to the Institute of Cancer Research to support independent statistical analysis of a phase 3 trial of sterotatic body radiotherapy in prostate cancer, outside the submitted work. VK reports advisory and educational fees and non-financial support from Astellas, educational fees from Bayer, non-financial educational support from Janssen, advisory and educational fees and non-financial support from Ipsen, and educational fees from Tolmar. All other authors declare no competing interests. Figure 1 Quality-of-life study profile *Trial entry or pre-radiotherapy if patients were receiving endocrine therapy at trial entry. †Patients were excluded from the fixed timepoint analyses if their QoL assessments were dated outside prespecified acceptable time intervals: after 1 month of endocrine treatment or after randomisation for baseline; before 3 months or after 1 week of starting radiotherapy for pre-radiotherapy; outside 2 weeks from the expected date of completion for 10 weeks; and outside of 3 months from the expected date of completion for later timepoints. QoL=quality of life. Figure 2 Overall bowel, urinary, and sexual bother

*Trial entry or pre-radiotherapy if patients were receiving endocrine therapy at trial entry. †Patients were excluded from the fixed timepoint analyses if their QoL assessments were dated outside prespecified acceptable time intervals: after 1 month of endocrine treatment or after randomisation for baseline; before 3 months or after 1 week of starting radiotherapy for pre-radiotherapy; outside 2 weeks from the expected date of completion for 10 weeks; and outside of 3 months from the expected date of completion for later timepoints. QoL=quality of life. Figure 2 Overall bowel, urinary, and sexual bother Data are prevalence of overall bowel bother (A), time to small or worse overall bowel bother (B), prevalence of overall urinary bother (C), time to small or worse overall urinary bother (D), prevalence of overall sexual bother (E), and time to small or worse overall sexual bother (F). Figure 3 Change in domain scores and single item overall bother scores from baseline to 24 months

Data are prevalence of overall bowel bother (A), time to small or worse overall bowel bother (B), prevalence of overall urinary bother (C), time to small or worse overall urinary bother (D), prevalence of overall sexual bother (E), and time to small or worse overall sexual bother (F). Figure 3 Change in domain scores and single item overall bother scores from baseline to 24 months Change in UCLA-PCI bowel function domain score*† (A); change in EPIC bowel summary domain score*† (B); change in overall bowel bother from pre-radiotherapy† to 24 months (C); change in UCLA-PCI urinary function domain score*† (D); change in EPIC urinary summary domain score*† (E); change in overall urinary bother from pre-radiotherapy† to 24 months (F); change in UCLA-PCI sexual function domain score* (G); change in EPIC sexual summary domain score* (H); and change in overall sexual bother from baseline to 24 months (I). EPIC=Expanded Prostate Cancer Index Composite. UCLA-PCI=University of California, Los Angeles Prostate Cancer Index. Error bars are 99% CIs.*Higher domain scores indicate better function. †For all urinary and bowel items and domain scores, to maximise numbers, the pre-radiotherapy score was used as a surrogate baseline score unless missing, in which case the baseline score was used. Table 1 Baseline characteristics and clinical history

Change in UCLA-PCI bowel function domain score*† (A); change in EPIC bowel summary domain score*† (B); change in overall bowel bother from pre-radiotherapy† to 24 months (C); change in UCLA-PCI urinary function domain score*† (D); change in EPIC urinary summary domain score*† (E); change in overall urinary bother from pre-radiotherapy† to 24 months (F); change in UCLA-PCI sexual function domain score* (G); change in EPIC sexual summary domain score* (H); and change in overall sexual bother from baseline to 24 months (I). EPIC=Expanded Prostate Cancer Index Composite. UCLA-PCI=University of California, Los Angeles Prostate Cancer Index. Error bars are 99% CIs.*Higher domain scores indicate better function. †For all urinary and bowel items and domain scores, to maximise numbers, the pre-radiotherapy score was used as a surrogate baseline score unless missing, in which case the baseline score was used. Table 1 Baseline characteristics and clinical history 74 Gy in 37 fractions (n=676) 60 Gy in 20 fractions (n=686) 57 Gy in 19 fractions (n=692) Age (years) 69 (65–73) 69 (64–73) 68 (64–73) T stage T1a/1b/1c/1x 224 (33%) 273 (40%) 252 (36%) T2a/b/c/x 393 (58%) 355 (52%) 368 (53%) T3a/x 59 (9%) 57 (8%) 71 (10%) Unknown 0 (0%) 1 (<1%) 1 (<1%) Gleason score ≤6 248 (37%) 252 (37%) 233 (34%) 7 406 (60%) 413 (60%) 432 (62%) 8 22 (3%) 21 (3%) 27 (4%) Prostate-specific antigen (ng/mL) Median (IQR) 10·4 (7·3–14·6) 11·0 (7·8–15·5) 10·4 (7·2–14·5) Mean (SD) 11·3 (5·3) 11·9 (5·8) 11·3 (5·4) 0·00–4·99 44 (7%) 51 (7%) 46 (7%) 5·00–9·99 267 (39%) 249 (36%) 274 (40%) 10·00–19·90 319 (47%) 327 (48%) 321 (46%) 20·00–40·00 46 (7%) 59 (9%) 49 (7%) Unknown 0 (0%) 0 (0%) 2 (<1%) NCCN risk group Low 104 (15%) 113 (16%) 109 (16%) Intermediate 496 (73%) 498 (73%) 496 (72%) High 76 (11%) 75 (11%) 87 (13%) Diabetes Yes 74 (11%) 65 (9%) 77 (11%) No 599 (89%) 619 (90%) 606 (88%) Unknown 3 (<1%) 2 (<1%) 9 (1%) Hypertension Yes 248 (37%) 281 (41%) 273 (39%) No 423 (63%) 403 (59%) 414 (60%) Unknown 5 (<1%) 2 (<1%) 5 (<1%) Inflammatory bowel or diverticular disease Yes 25 (4%) 21 (3%) 27 (4%) No 646 (96%) 663 (97%) 659 (95%) Unknown 5 (<1%) 2 (<1%) 6 (1%) Previous pelvic surgery Yes 55 (8%) 51 (7%) 53 (8%) No 616 (91%) 633 (92%) 632 (91%) Unknown 5 (<1%) 2 (<1%) 7 (1%) Symptomatic haemorrhoids in past 12 months Yes 39 (6%) 50 (7%) 52 (8%) No 620 (92%) 617 (90%) 623 (90%) Unknown 17 (3%) 19 (3%) 17 (2%) Previous transurethral resection of the prostate Yes 55 (8%) 61 (9%) 62 (9%) No 606 (90%) 615 (90%) 617 (89%) Unknown 15 (2%) 10 (1%) 13 (2%) Hormone treatment duration (days)* 140 (113–169) 132 (102–165) 127 (102–157) Time from androgen suppression to start of radiotherapy (days) 116 (103–138) 118 (103–138) 115 (103–139) Time from radiotherapy start to end of androgen suppression (days)† 16 (−3 to 42) 6 (−6 to 24) 6 (−7 to 23) Data are n (%) or median (IQR) unless otherwise stated. NCCN=National Comprehensive Cancer Network.

02–157) Time from androgen suppression to start of radiotherapy (days) 116 (103–138) 118 (103–138) 115 (103–139) Time from radiotherapy start to end of androgen suppression (days)† 16 (−3 to 42) 6 (−6 to 24) 6 (−7 to 23) Data are n (%) or median (IQR) unless otherwise stated. NCCN=National Comprehensive Cancer Network. * Hormone treatment duration was the total duration, including before study entry. † Negative values are from patients who stopped androgen suppression before starting radiotherapy. Table 2 Bowel, urinary, and sexual domain scores at 24 months for UCLA-PCI and EPIC QoL instruments

02–157) Time from androgen suppression to start of radiotherapy (days) 116 (103–138) 118 (103–138) 115 (103–139) Time from radiotherapy start to end of androgen suppression (days)† 16 (−3 to 42) 6 (−6 to 24) 6 (−7 to 23) Data are n (%) or median (IQR) unless otherwise stated. NCCN=National Comprehensive Cancer Network. * Hormone treatment duration was the total duration, including before study entry. † Negative values are from patients who stopped androgen suppression before starting radiotherapy. Table 2 Bowel, urinary, and sexual domain scores at 24 months for UCLA-PCI and EPIC QoL instruments 74 Gy in 37 fractions (n=676) 60 Gy in 20 fractions (n=686) 57 Gy in 19 fractions (n=692) 74 Gy vs 60 Gy p value* 74 Gy vs 57 Gy p value* 60 Gy vs 57 Gy p value* Number of patients with data Median (IQR) Number of patients with data Median (IQR) Number of patients with data Median (IQR) Bowel function (UCLA) 312 (46%) 93·8 (82·5–100·0) 310 (45%) 91·8 (79·3–100·0) 331 (48%) 93·8 (81·3–100·0) 0·064 0·77 0·12 Urinary function (UCLA) 311 (46%) 100·0 (81·8–100·0) 313 (46%) 100·0 (81·8–100·0) 334 (48%) 100·0 (83·5–100·0) 0·69 0·47 0·74 Sexual function (UCLA) 307 (45%) 27·1 (4·1–53·1) 300 (44%) 23·4 (7·3–57·3) 321 (46%) 26·0 (7·3–56·3) 0·39 0·33 0·92 Bowel function (EPIC) 95 (14%) 96·4 (89·3–96·4) 94 (14%) 96·4 (89·3–100·0) 100 (14%) 92·9 (85·7–98·2) 0·15 0·51 0·059 Bowel bother (EPIC) 95 (14%) 95·8 (87·5–100·0) 96 (14%) 95·8 (83·3–100·0) 101 (15%) 95·8 (79·2–100·0) 0·54 0·15 0·38 Bowel summary (EPIC) 93 (14%) 94·2 (88·5–98·1) 94 (14%) 94·2 (87·5–100·0) 99 (14%) 94·2 (84·6–98·1) 0·41 0·41 0·15 Urinary function (EPIC) 97 (14%) 100·0 (93·4–100·0) 99 (14%) 100·0 (88·4–100·0) 104 (15%) 100·0 (90·9–100·0) 0·18 0·84 0·11 Urinary bother (EPIC) 89 (13%) 89·3 (79·2–96·4) 95 (14%) 89·3 (78·6–96·4) 102 (15%) 90·5 (75·0–96·4) 0·64 0·72 0·38 Urinary summary (EPIC) 89 (13%) 91·0 (85·4–97·9) 95 (14%) 93·1 (82·7–97·9) 101 (15%) 93·8 (82·7–97·9) 1·00 0·91 0·93 Sexual function (EPIC) 89 (13%) 21·6 (0·0–60·0) 88 (13%) 21·6 (0·0–53·4) 98 (14%) 26·6 (0·0–58·4) 0·78 0·74 0·53 Sexual summary (EPIC) 92 (14%) 28·4 (15·2–62·5) 93 (14%) 23·7 (16·7–58·3) 99 (14%) 27·8 (13·8–61·2) 0·60 0·84 0·74 EPIC-50 was used for bowel and urinary domains and EPIC-26 for sexual domains. UCLA-PCI=University of California, Los Angeles Prostate Cancer Index. EPIC=Expanded Prostate Cancer Index Composite.

0·74 0·53 Sexual summary (EPIC) 92 (14%) 28·4 (15·2–62·5) 93 (14%) 23·7 (16·7–58·3) 99 (14%) 27·8 (13·8–61·2) 0·60 0·84 0·74 EPIC-50 was used for bowel and urinary domains and EPIC-26 for sexual domains. UCLA-PCI=University of California, Los Angeles Prostate Cancer Index. EPIC=Expanded Prostate Cancer Index Composite. * Mann-Whitney U test.

Introduction Prostate cancer is the most common cancer in men in the UK, with 41 736 new cases in 2011.1 Since the introduction of prostate-specific antigen (PSA) testing, most men diagnosed have localised disease. Management options include external-beam radiotherapy, brachytherapy, radical prostatectomy, active surveillance (for men with low-risk disease), and watchful waiting (for those unsuitable for radical curative treatment), with management choices often affected by potential treatment-related toxic effects. Prostate cancer and its treatment are the leading cause of cancer years lived with disability.2 Research in context Evidence before this study

Introduction Prostate cancer is the most common cancer in men in the UK, with 41 736 new cases in 2011.1 Since the introduction of prostate-specific antigen (PSA) testing, most men diagnosed have localised disease. Management options include external-beam radiotherapy, brachytherapy, radical prostatectomy, active surveillance (for men with low-risk disease), and watchful waiting (for those unsuitable for radical curative treatment), with management choices often affected by potential treatment-related toxic effects. Prostate cancer and its treatment are the leading cause of cancer years lived with disability.2 Research in context Evidence before this study We searched PubMed for articles published between Jan 1, 1990, and Oct 18, 2002, before trial commencement using the terms “radiotherapy AND prostate cancer AND (hypofractionation OR alpha/beta ratio)” and then updated results to Sept 8, 2015. Before the CHHiP trial began, reports based on retrospective series of patients suggested that the α/β ratio for prostate cancer might be low, but only two small randomised trials had tested hypofractionation compared with conventional fractionation, both using relatively low doses of radiotherapy, and neither trial was large enough to confirm or refute a benefit. Since CHHiP started, more recent results from a meta-analysis of five small trials testing hypofractionation and retrospective reviews of large patient databases have been done, suggesting that the best estimates for the α/β ratio are between 1·4 Gy and 1·9 Gy, although estimates up to 8·3 Gy have been calculated. However, these retrospective analyses and reviews have not changed clinical practice; hence the need for a large randomised controlled trial. Meta-analyses of studies of dose-escalated radiotherapy and neoadjuvant androgen deprivation show improved disease control compared with standard radiotherapy doses, but dose escalation increases bowel side-effects. However, conformal and intensity-modulated radiotherapy improves dose distributions of radiotherapy and conformal radiotherapy reduces side-effects.

adiotherapy and neoadjuvant androgen deprivation show improved disease control compared with standard radiotherapy doses, but dose escalation increases bowel side-effects. However, conformal and intensity-modulated radiotherapy improves dose distributions of radiotherapy and conformal radiotherapy reduces side-effects. Added value of this study The CHHiP trial is, to our knowledge, the largest randomised treatment study undertaken in localised prostate cancer. We tested two experimental hypofractionated radiotherapy schedules using 3 Gy per fraction to total doses of 60 Gy and 57 Gy compared with standard fractionation using 2 Gy per fraction to a total dose of 74 Gy. We have shown that the hypofractionated schedule of 60 Gy in 20 fractions is non-inferior to a standard schedule of 74 Gy in 37 fractions for the endpoint of biochemical and clinical disease control. Overall treatment time was reduced from 7·4 weeks to 4 weeks. 57 Gy in 19 fractions could not be claimed to be non-inferior to the control 74 Gy group. The results give an estimate of 1·8 Gy for the α/β ratio for prostate cancer. Quality controlled IMRT techniques were used and the side-effect profiles were favourable and low in all three randomised groups. Interpretation

The CHHiP trial is, to our knowledge, the largest randomised treatment study undertaken in localised prostate cancer. We tested two experimental hypofractionated radiotherapy schedules using 3 Gy per fraction to total doses of 60 Gy and 57 Gy compared with standard fractionation using 2 Gy per fraction to a total dose of 74 Gy. We have shown that the hypofractionated schedule of 60 Gy in 20 fractions is non-inferior to a standard schedule of 74 Gy in 37 fractions for the endpoint of biochemical and clinical disease control. Overall treatment time was reduced from 7·4 weeks to 4 weeks. 57 Gy in 19 fractions could not be claimed to be non-inferior to the control 74 Gy group. The results give an estimate of 1·8 Gy for the α/β ratio for prostate cancer. Quality controlled IMRT techniques were used and the side-effect profiles were favourable and low in all three randomised groups. Interpretation The findings from this pre-planned analysis of the CHHiP trial show that the hypofractionated IMRT schedule giving 60 Gy in 3 Gy fractions in 4 weeks is both effective and safe and can be recommended as a new standard of care for patients with localised prostate cancer using the high-quality radiotherapy techniques described. The results are most robust for patients with intermediate-risk disease who received short-course androgen deprivation therapy.

ons in 4 weeks is both effective and safe and can be recommended as a new standard of care for patients with localised prostate cancer using the high-quality radiotherapy techniques described. The results are most robust for patients with intermediate-risk disease who received short-course androgen deprivation therapy. External-beam radiotherapy is most appropriate for men with intermediate-risk or high-risk disease,3 and is associated with long-term disease control in most patients.4 About 15 800 men receive radical prostate radiotherapy in the UK every year (Ball C, National Clinical Analysis and Specialised Applications Team, The Clatterbridge Cancer Centre NHS Foundation Trust, personal communication). Several phase 3 randomised controlled trials have shown the benefit of dose escalation5, 6 and high-dose conformal radiotherapy with conventional 2 Gy daily fractions to a total dose of 74 Gy is the standard of care in the UK.7 However, a meta-analysis showed that high-dose radiotherapy (74–80 Gy) is associated with an increased risk (odds ratio 1·58) of late gastrointestinal toxicity of grade 2 or more compared with lower doses of radiotherapy (64–70·2 Gy).8 Therefore, it is important to use advanced radiotherapy techniques that are able to sculpt dose distributions to the prostate target and avoid the organs at risk.

ciated with an increased risk (odds ratio 1·58) of late gastrointestinal toxicity of grade 2 or more compared with lower doses of radiotherapy (64–70·2 Gy).8 Therefore, it is important to use advanced radiotherapy techniques that are able to sculpt dose distributions to the prostate target and avoid the organs at risk. Additionally, there has been interest in the fraction sensitivity of prostate cancer.9, 10, 11 The association between total isoeffective radiation dose and fraction size is described by a linear quadratic model which uses two constants: α and β. The ratio α/β is inversely related to the effect of changes in fraction size on normal and malignant tissues. The α/β ratio for most cancers and acute normal tissue reactions is believed to be high and about 10 Gy. However, for prostate cancer, a value as low as 1·5 Gy has been suggested, which is lower than the 3 Gy reported for the late reactions of most normal tissues (including rectum).12 These findings have potentially important therapeutic implications. Hypofractionated radiotherapy, giving fewer fractions each with a higher dose, might improve the therapeutic ratio, resource use, and patient convenience. The main aims of the CHHiP trial (CRUK/06/016) were to compare the efficacy and toxicity of conventional and hypofractionated radiotherapy using high-quality radiation techniques.

Additionally, there has been interest in the fraction sensitivity of prostate cancer.9, 10, 11 The association between total isoeffective radiation dose and fraction size is described by a linear quadratic model which uses two constants: α and β. The ratio α/β is inversely related to the effect of changes in fraction size on normal and malignant tissues. The α/β ratio for most cancers and acute normal tissue reactions is believed to be high and about 10 Gy. However, for prostate cancer, a value as low as 1·5 Gy has been suggested, which is lower than the 3 Gy reported for the late reactions of most normal tissues (including rectum).12 These findings have potentially important therapeutic implications. Hypofractionated radiotherapy, giving fewer fractions each with a higher dose, might improve the therapeutic ratio, resource use, and patient convenience. The main aims of the CHHiP trial (CRUK/06/016) were to compare the efficacy and toxicity of conventional and hypofractionated radiotherapy using high-quality radiation techniques. Methods Study design and participants CHHiP is an international, multicentre, randomised, phase 3, non-inferiority trial comparing the conventionally fractionated schedule of 74 Gy in 37 fractions with two experimental hypofractionated schedules of 60 Gy in 20 fractions and 57 Gy in 19 fractions in men with localised prostate cancer. Safety of the 3 Gy (fraction) schedules was reported after a pre-planned analysis of the first 457 men recruited.13 Here, we report primary efficacy results and further comparative safety data.

erimental hypofractionated schedules of 60 Gy in 20 fractions and 57 Gy in 19 fractions in men with localised prostate cancer. Safety of the 3 Gy (fraction) schedules was reported after a pre-planned analysis of the first 457 men recruited.13 Here, we report primary efficacy results and further comparative safety data. Men older than 16 years who had histologically confirmed T1b–T3aN0M0 prostate cancer and a WHO performance status of 0 or 1 were eligible. Initially, men with a PSA concentration of less than 40 ng/mL and risk of lymph node involvement14 less than 30% were eligible. On Aug 1, 2006, after 454 patients had been recruited, these criteria were revised to reflect the developing consensus on use of long-term androgen deprivation in locally advanced disease. Thereafter, a PSA concentration less than 30 ng/mL and a risk of seminal vesicle involvement15 less than 30% were needed. Patients were ineligible if they had both T3 tumours and a Gleason score of 8 or higher, or a life expectancy of less than 10 years. Other exclusion criteria included previous pelvic radiotherapy or radical prostatectomy, previous androgen suppression, another active malignancy in the past 5 years (other than cutaneous basal-cell carcinoma), comorbid conditions precluding radical radiotherapy, hip prosthesis (criterion amended to bilateral hip prosthesis Jan 30, 2009), and full anticoagulation treatment (criterion removed July 1, 2009). Full details of trial design, eligibility, and treatment have been reported previously.13 The protocol is available in the appendix (pp 27–90).

ecluding radical radiotherapy, hip prosthesis (criterion amended to bilateral hip prosthesis Jan 30, 2009), and full anticoagulation treatment (criterion removed July 1, 2009). Full details of trial design, eligibility, and treatment have been reported previously.13 The protocol is available in the appendix (pp 27–90). The study was approved in the UK by the London Multi-centre Research Ethics Committee (04/MRE02/10) and by the institutional research board of each participating international site. The trial was sponsored by the Institute of Cancer Research and was done in accordance with the principles of good clinical practice. All patients provided written informed consent. The Institute of Cancer Research Clinical Trials and Statistics Unit (ICR-CTSU; London, UK) coordinated the study and carried out central statistical data monitoring and all analyses. The trial management group was overseen by an independent trial steering committee.

ice. All patients provided written informed consent. The Institute of Cancer Research Clinical Trials and Statistics Unit (ICR-CTSU; London, UK) coordinated the study and carried out central statistical data monitoring and all analyses. The trial management group was overseen by an independent trial steering committee. Randomisation and masking Men were registered into the trial before or after commencement of androgen deprivation therapy. Following registration, and within 4–6 weeks before radiotherapy, patients were randomly assigned (1:1:1) to receive conventional fractionation (control) or one of two hypofractionated schedules. Randomisation was via telephone to the ICR-CTSU. Computer-generated random permuted blocks of sizes six and nine were used, stratified by National Comprehensive Cancer Network (NCCN) risk-classification (low vs intermediate vs high)3 and radiotherapy treatment centre. It was not possible to mask patients or clinicians to treatment allocation.

lephone to the ICR-CTSU. Computer-generated random permuted blocks of sizes six and nine were used, stratified by National Comprehensive Cancer Network (NCCN) risk-classification (low vs intermediate vs high)3 and radiotherapy treatment centre. It was not possible to mask patients or clinicians to treatment allocation. Procedures Short-course androgen deprivation treatment was given for 3–6 months before and during radiotherapy; this was optional for patients with low-risk disease. Injections of a luteinising-hormone-releasing hormone (LHRH) analogue every month, combined with initial anti-androgen to reduce testosterone flare, or an anti-androgen alone, were allowed. Individuals assigned to the 74 Gy in 37 fractions control group received 2 Gy daily fractions (Monday to Friday treatment) for 7·4 weeks. Individuals in the experimental groups received hypofractionated treatment with 3 Gy daily fractions to a total dose of either 60 Gy in 20 fractions in 4·0 weeks (≥28 days) or 57 Gy in 19 fractions in 3·8 weeks (≥27 days). Biological doses in the hypofractionated schedules were calculated to be equivalent to those in the conventional schedule assuming α/β ratios of 2·4 Gy for the 60 Gy group and 1·4 Gy for the 57 Gy group. All treatment groups received intensity-modulated radiation techniques (IMRT). Treatment delays for toxic effects, and for technical reasons of up to 5 days, were permitted.

alculated to be equivalent to those in the conventional schedule assuming α/β ratios of 2·4 Gy for the 60 Gy group and 1·4 Gy for the 57 Gy group. All treatment groups received intensity-modulated radiation techniques (IMRT). Treatment delays for toxic effects, and for technical reasons of up to 5 days, were permitted. Planning of radiotherapy treatment for all three groups was done with forward or inverse three-dimensional methods about 12 weeks after the start of hormonal treatment. The complex forward-planned multisegment technique using an integrated simultaneous boost has been previously described16 using three treatment fields with a total of eight segments. Pelvic lymph nodes were not included in the target volumes. Mandatory dose constraints were defined for target coverage and avoidance of normal tissues including rectum, bowel, bladder, and femoral heads. Treatment plans were reviewed and dose reductions permitted to meet dose constraints. Treatment was delivered with 6–15 MV photons with multileaf collimators to shape beams. Portal imaging was used to verify treatment accuracy, which was to be within 3 mm and was taken at least three times during week 1 and at least weekly intervals thereafter. Use of image-guided techniques (IGRT) was permitted but not required. Details of target volumes, dose parameters, and constraints are given in the appendix (pp 2, 3, 15). The integral quality-assurance programme has previously been described.13

t least three times during week 1 and at least weekly intervals thereafter. Use of image-guided techniques (IGRT) was permitted but not required. Details of target volumes, dose parameters, and constraints are given in the appendix (pp 2, 3, 15). The integral quality-assurance programme has previously been described.13 Staging investigations included PSA measurement, standard haematology and biochemistry, lymph node assessment by pelvic MRI or CT, and bone scans for patients at intermediate or high risk. Histology was locally assessed with diagnostic transrectal ultrasound-guided biopsies (or specimens from transurethral resection of the prostate) and reported with the Gleason system. PSA concentrations were recorded before commencement of androgen deprivation therapy and radiotherapy and subsequently at weeks 10, 18, and 26 after radiotherapy and then at 6-month intervals for 5 years and subsequently annually.

imens from transurethral resection of the prostate) and reported with the Gleason system. PSA concentrations were recorded before commencement of androgen deprivation therapy and radiotherapy and subsequently at weeks 10, 18, and 26 after radiotherapy and then at 6-month intervals for 5 years and subsequently annually. Baseline, pre-radiotherapy treatment, acute, and late toxicity data were collected using physician-completed and patient-reported outcome questionnaires. Instruments chosen reflected practice at the time of trial commencement, the desirability of assessing symptoms before radiotherapy to allow consideration of emergent events, and to facilitate comparison with other studies. Baseline bowel, bladder, and sexual function assessments were made before androgen deprivation therapy and radiotherapy and were graded using the Late Effects on Normal Tissues: Subjective/Objective/Management (LENT/SOM)17 and Royal Marsden Hospital (RMH)18 scoring systems and patient-reported outcome questionnaires. Acute toxicity data were collected for the first 2163 randomly assigned patients. When the sample size was increased (see Statistical analysis) it was felt that sufficient data had been collected on acute toxicity to allow robust conclusions to be drawn about comparisons between the three randomised groups. Reactions were graded every week during radiotherapy and at weeks 10, 12, and 18 from radiotherapy start date using the Radiation Therapy Oncology Group (RTOG) scoring system for acute toxicity.19 Late side-effects were then assessed beginning 26 weeks after the start of radiotherapy and every 6 months for 2 years and then yearly to 5 years, as previously described,13 using the RTOG grades for late side-effects,19 RMH, and LENT/SOM scoring systems. A quality-of-life substudy using patient-reported outcomes was included as previously described.20 From trial initiation to early 2009, the UCLA-PCI, including the Short Form 36 (SF-36), and the Functional Assessment of Cancer Therapy-Prostate (FACT-P) quality-of-life instruments were used. After a protocol amendment on March 12, 2009, the Expanded Prostate Cancer Index Composite (EPIC) and Short Form 12 (SF-12) instruments replaced UCLA-PCI, SF-36, and FACT-P due to EPIC becoming the patient-reported outcome measure of choice. EPIC-50 was used for bowel and urinary domains and EPIC-26 for sexual and hormonal domains. Patient-reported outcomes to 2 years after treatment have been reported.20

te (EPIC) and Short Form 12 (SF-12) instruments replaced UCLA-PCI, SF-36, and FACT-P due to EPIC becoming the patient-reported outcome measure of choice. EPIC-50 was used for bowel and urinary domains and EPIC-26 for sexual and hormonal domains. Patient-reported outcomes to 2 years after treatment have been reported.20 Outcomes The primary outcome measure was time to biochemical or clinical failure, defined as the time from randomisation to biochemical failure or prostate cancer recurrence. The initial definition of biochemical failure (PSA >2 ng/mL 6 months or more after the commencement of radiotherapy and a PSA rising by 50% or more from the nadir) was updated in 2007, and applied retrospectively, to reflect the Phoenix consensus guidelines as a PSA concentration greater than nadir plus 2 ng/mL.21 The nadir PSA was the lowest concentration recorded at any time after commencement of androgen deprivation therapy or radiotherapy. A consecutive confirmatory PSA concentration was required. Biochemical failure events were determined centrally from PSA concentrations and confirmed by the local investigator. Prostate cancer recurrence events were as reported by the investigator and included recommencement of androgen deprivation therapy, local recurrence, lymph node or pelvic recurrence, and distant metastases.

hemical failure events were determined centrally from PSA concentrations and confirmed by the local investigator. Prostate cancer recurrence events were as reported by the investigator and included recommencement of androgen deprivation therapy, local recurrence, lymph node or pelvic recurrence, and distant metastases. Secondary efficacy outcome measures were disease-free survival (time from randomisation to any prostate cancer-related event or death from any cause); overall survival (time from randomisation to death from any cause); development of metastases; and recommencement of hormonal treatment for disease recurrence. Cause of death was centrally reviewed by a panel of three trial investigators (DD, JG, IS), masked to treatment allocation. Additional secondary endpoints were acute and late side-effects. Acute toxicity outcomes were summarised by reporting the peak and week 18 bowel and bladder side-effects. Clinician-reported late toxicity outcomes were the proportion of patients with a grade 2 or worse toxic effect at 2 and 5 years, and time to development of grade 1, grade 2, and grade 3 toxicity (assessed using each scoring method). Patient-reported outcomes included overall bowel, bladder, and sexual dysfunction bother reported as single items on the UCLA-PCI and EPIC-50 instruments.

tients with a grade 2 or worse toxic effect at 2 and 5 years, and time to development of grade 1, grade 2, and grade 3 toxicity (assessed using each scoring method). Patient-reported outcomes included overall bowel, bladder, and sexual dysfunction bother reported as single items on the UCLA-PCI and EPIC-50 instruments. Statistical analysis The trial was powered to assess non-inferiority in biochemical or clinical failure-free rate between the hypofractionated and conventional radiotherapy schedules. A three-arm design allowed estimation of isoeffective doses for both efficacy and complications. We assumed a 70% failure-free rate at 5 years in the control group and, with 2163 patients, initially wished to exclude a decrease of 6% in a hypofractionated group. Due to accrual exceeding expectations, a protocol amendment on Nov 23, 2009, increased the sample size to allow a smaller non-inferiority margin of 5%, corresponding to a critical hazard ratio (HR) of 1·208, to be used. This critical HR was used for all non-inferiority analyses. To conclude non-inferiority with 80% power (one-sided α=0·05), 3163 men (1054 per treatment group) were required. This analysis would require 349 events in the control group but, as agreed with the Independent Data Monitoring Committee, data could also be considered sufficiently mature for analysis after a median follow-up of 5 years. A small allowance (1·5%) for dropout or loss to follow-up was incorporated.

group) were required. This analysis would require 349 events in the control group but, as agreed with the Independent Data Monitoring Committee, data could also be considered sufficiently mature for analysis after a median follow-up of 5 years. A small allowance (1·5%) for dropout or loss to follow-up was incorporated. Analyses for all time-to-event endpoints were on an intention-to-treat basis. The primary outcome was also analysed in the per-protocol population, including all patients receiving at least one fraction of their allocated radiotherapy schedule. For time to biochemical or clinical failure, patients event free at the time of analysis were censored at their last known PSA assessment. For disease-free and overall survival, patients were censored at the date they were last known to be alive. For development of metastases and recommencement of hormonal treatment patients were censored at the date they were last seen or date of death.

e time of analysis were censored at their last known PSA assessment. For disease-free and overall survival, patients were censored at the date they were last known to be alive. For development of metastases and recommencement of hormonal treatment patients were censored at the date they were last seen or date of death. Kaplan-Meier methods were used to estimate event rates. Estimates of treatment effect were made using unadjusted and also adjusted Cox regression models, with an HR less than 1 indicating a decreased risk of the event in the hypofractionated treatment group compared with control. Covariates included in adjusted Cox regression models were age (≤69 years vs >69 years), NCCN risk group (low vs intermediate vs high), Gleason score (≤6 vs >6), clinical stage, and pre-androgen deprivation therapy PSA (<10 ng/mL vs 10–20 ng/mL vs >20 ng/mL). Although the trial was not designed to directly compare the hypofractionated schedules, hypothesis generating comparisons have been made, with an HR less than 1 indicating a decreased risk of the event in the 60 Gy group compared with the 57 Gy group. For time to biochemical or clinical failure, HRs are provided with two-sided 90% CIs (equivalent to one-sided 95% CIs) in accordance with the one-sided non-inferiority design. p values to reject the null hypothesis of HR of 1·208 or greater (pNI) are reported. In all other instances, 95% CIs are reported. Comparisons were made between the control group and each hypofractionated group using the log-rank test, with a p value less than 0·05 indicating statistical significance.

feriority design. p values to reject the null hypothesis of HR of 1·208 or greater (pNI) are reported. In all other instances, 95% CIs are reported. Comparisons were made between the control group and each hypofractionated group using the log-rank test, with a p value less than 0·05 indicating statistical significance. Absolute treatment differences (δ) in time to biochemical or clinical failure have been calculated based on the Kaplan-Meier estimate of the failure-free rate in the control group and the HR. A competing risks analysis was done using the methods of Fine and Gray for the primary outcome measure, with death due to any cause as the competing event with consistent results (data not shown). Pre-planned subgroup analyses of the primary outcome by NCCN risk group were done in addition to multivariable analyses adjusting for risk group and prespecified clinically prognostic factors. Heterogeneity of the treatment effect was tested using χ2 tests for interaction. The α/β ratio for prostate cancer was estimated10 assuming a linear fit for data from the hypofractionated groups.

oup were done in addition to multivariable analyses adjusting for risk group and prespecified clinically prognostic factors. Heterogeneity of the treatment effect was tested using χ2 tests for interaction. The α/β ratio for prostate cancer was estimated10 assuming a linear fit for data from the hypofractionated groups. Acute toxicity analyses were done in the safety population, including all patients who received at least one fraction of radiotherapy. All available data were used irrespective of timing of assessment, with the exception of comparisons at 18 weeks, where a 2 week window either side of the expected date was used. Pairwise comparisons of the distribution of acute toxicity scores were compared using Mann-Whitney tests. For each late toxicity scale, the proportion of late grade 2 or worse toxicity at 2 years and 5 years is reported with exact binomial 95% CIs. Fisher's exact tests were used to compare each hypofractionated group with control. Time to first late adverse event was compared with the use of the Kaplan-Meier method. Patients not experiencing an event were censored at last known toxicity assessment. The proportion of small or worse patient-reported bother and time to first very small, small, moderate, or worse late bother score were analysed as for late toxicity endpoints. To make some allowance for multiple testing of toxicity and patient-reported endpoints, a p value of less than 0·01 was considered statistically significant.

of small or worse patient-reported bother and time to first very small, small, moderate, or worse late bother score were analysed as for late toxicity endpoints. To make some allowance for multiple testing of toxicity and patient-reported endpoints, a p value of less than 0·01 was considered statistically significant. For all time-to-event analyses the proportional hazards assumption of the Cox model was tested using Schoenfeld residuals and found to hold (appendix pp 25–26). Analyses were based on a database snapshot taken on Sept 8, 2015, and were done using Stata version 13. The CHHiP trial is registered as an International Standard Randomised Controlled Trial, number ISRCTN97182923. Role of the funding source The funding source provided peer-reviewed approval for the trial, but had no other role in study design, collection, analysis, interpretation of data, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. HMo, CG, and EH also had full access to the data.

l for the trial, but had no other role in study design, collection, analysis, interpretation of data, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. HMo, CG, and EH also had full access to the data. Results Between Oct 18, 2002, and June 17, 2011, 3216 men were recruited from 71 centres in the UK, Republic of Ireland, Switzerland, and New Zealand (appendix pp 1, 4–5). 1065 patients were assigned to the conventional 74 Gy schedule, 1074 to the 60 Gy schedule, and 1077 to the 57 Gy schedule; 64 patients received no radiotherapy (figure 1). Demographic and clinical characteristics were balanced across treatment groups (table). 3213 (97%) patients received concurrent androgen deprivation therapy, with 2700 (84%) patients receiving LHRH analogues and short-term anti-androgens. Most patients who did not receive androgen deprivation therapy had low-risk disease (73 [78%] of 93 patients). For patients who received androgen deprivation treatment, the median duration before commencement of radiotherapy was 16 weeks (IQR 15–20). Overall, 3152 (98%) patients received radiotherapy, and 3117 (97%) received the allocated dose and fractionation schedule (figure 1). Treatment delays of 1 week or more occurred in only 23 (1%) patients. At the time of analysis, median follow-up was 62·6 months (54·1–77·2) in the 74 Gy group, 62·2 months (53·9–77·2) in the 60 Gy group, and 62·4 months (53·7–76·6) in the 57 Gy group.

) received the allocated dose and fractionation schedule (figure 1). Treatment delays of 1 week or more occurred in only 23 (1%) patients. At the time of analysis, median follow-up was 62·6 months (54·1–77·2) in the 74 Gy group, 62·2 months (53·9–77·2) in the 60 Gy group, and 62·4 months (53·7–76·6) in the 57 Gy group. By 5 years, the number of patients with biochemical or clinical events were 111 of 1065 in the 74 Gy group, 88 of 1074 in the 60 Gy group, and 132 of 1077 in the 57 Gy group, respectively. 5-year biochemical or clinical failure-free rates were 88·3% (95% CI 86·0–90·2) in the 74 Gy group, 90·6% (88·5–92·3) in the 60 Gy group, and 85·9% (83·4–88·0) in the 57 Gy group (figure 2A, appendix p 6). With reference to the critical HR for non-inferiority, 60 Gy was non-inferior to 74 Gy with HR 0·84 (90% CI 0·68–1·03), pNI=0·0018. Since the upper limit of the 90% CI for the HR comparing 57 Gy with 74 Gy (HR 1·20 [0·99–1·46]) exceeds 1·208 (pNI=0·48), non-inferiority of the 57 Gy schedule relative to 74 Gy cannot be claimed. To facilitate comparison with other studies, 95% CIs were estimated as 0·65–1·07 for the HR in the 60 Gy group and 0·96–1·51 for the HR in the 57 Gy group. The estimated absolute difference in the proportion of patients in the hypofractionated groups free from biochemical or clinical failure compared with that in the control group at 5 years is δ=1·80% (90% CI −0·34 to 3·58) for 60 Gy versus 74 Gy and δ=–2·20% (–4·88 to 0·08) for 57 Gy versus 74 Gy. Analyses in the per-protocol population confirmed these results (60 Gy, HR 0·83 [90% CI 0·68–1·02], pNI=0·0015, δ=1·88% [90% CI −0·27 to 3·67]; 57 Gy, HR 1·17 [0·97 to 1·42], pNI=0·40, δ=–1·92% [–4·59 to 0·34]).

s δ=1·80% (90% CI −0·34 to 3·58) for 60 Gy versus 74 Gy and δ=–2·20% (–4·88 to 0·08) for 57 Gy versus 74 Gy. Analyses in the per-protocol population confirmed these results (60 Gy, HR 0·83 [90% CI 0·68–1·02], pNI=0·0015, δ=1·88% [90% CI −0·27 to 3·67]; 57 Gy, HR 1·17 [0·97 to 1·42], pNI=0·40, δ=–1·92% [–4·59 to 0·34]). Estimates of the HR adjusted for age (≤69 years vs >69 years), NCCN risk group, Gleason score (≤6 vs ≥7), clinical stage, and pre-androgen deprivation therapy PSA (<10 ng/mL vs 10–20 ng/mL vs >20 ng/mL) also confirmed these results (60 Gy vs 74 Gy, HR 0·86 [90% CI 0·70–1·06], p=0·25; 57 Gy vs 74 Gy, HR 1·21 [90% CI 0·99–1·46], p=0·11; appendix p 7). Prespecified subgroup analyses of time to biochemical or clinical failure showed no significant interactions with treatment group, except for age, where older men (age >69 years) had a reduced biochemical or clinical failure rate with 60 Gy compared with 74 Gy, but younger men (age ≤69 years) showed no difference in biochemical or clinical failure rate between treatment groups; however, this difference was not seen for the 57 Gy group (figure 3). In an exploratory secondary analysis to compare 60 Gy in 20 fractions with 57 Gy in 19 fractions the HR was 0·70 (95% CI 0·55–0·88, log-rank p=0·0026; δ=4·07% [95% CI 1·56–6·10]; appendix p 6).

mical or clinical failure rate between treatment groups; however, this difference was not seen for the 57 Gy group (figure 3). In an exploratory secondary analysis to compare 60 Gy in 20 fractions with 57 Gy in 19 fractions the HR was 0·70 (95% CI 0·55–0·88, log-rank p=0·0026; δ=4·07% [95% CI 1·56–6·10]; appendix p 6). At 5 years, biochemical and clinical failure-free rates for the NCCN low-risk, intermediate-risk, and high-risk groups were: 96·7% (95% CI 92·3–98·6), 86·8% (84·0–89·1), and 86·5% (78·4–91·7) for the 74 Gy group; 96·6% (92·1–98·6), 90·2% (87·7–92·3), and 84·2% (75·7–90·0) for the 60 Gy group; and 90·9% (85·1–94·5), 86·0% (83·1–88·5), and 78·3% (69·2–85·0) for the 57 Gy group, respectively (appendix p 16). 92 (9%) deaths were reported in the 74 Gy group, 73 (7%) in the 60 Gy group, and 87 (8%) in the 57 Gy group. Of 252 deaths reported, 40 (16%) were prostate cancer related, 88 (35%) were due to a second malignancy, 111 (44%) were non-cancer causes, and 13 (5%) were of unknown cause. No significant differences in overall survival were observed between the control group and either of the hypofractionated groups (figure 2B; appendix p 6).

aths reported, 40 (16%) were prostate cancer related, 88 (35%) were due to a second malignancy, 111 (44%) were non-cancer causes, and 13 (5%) were of unknown cause. No significant differences in overall survival were observed between the control group and either of the hypofractionated groups (figure 2B; appendix p 6). Progression events or death occurred in 209 (20%) of 1065 patients in the 74 Gy group, 179 (17%) of 1074 in the 60 Gy group, and 227 (21%) of 1077 in the 57 Gy group. Disease-free survival at 5 years was 82·3% (95% CI 79·6–84·6) in the 74 Gy group, 85·3% (82·8–87·5) in the 60 Gy group, and 80·1% (77·3–82·6) in the 57 Gy group. Compared with 74 Gy, the HR for disease-free survival was 0·83 (95% CI 0·68–1·01) in the 60 Gy group and 1·08 (0·90–1·31) in the 57 Gy group. 103 (3%) patients developed distant metastases: 32 (3%) in the 74 Gy group, 29 (3%) in the 60 Gy group, and 42 (4%) in the 57 Gy group. Androgen deprivation therapy was recommenced in 80 (8%) patients in the 74 Gy group, 70 (7%) in the 60 Gy group, and 89 (8%) in the 57 Gy group. Time to recommencement of androgen deprivation therapy and development of distant metastases were not significantly different between either of the hypofractionated schedules and the 74 Gy schedule (appendix pp 6, 17).

d in 80 (8%) patients in the 74 Gy group, 70 (7%) in the 60 Gy group, and 89 (8%) in the 57 Gy group. Time to recommencement of androgen deprivation therapy and development of distant metastases were not significantly different between either of the hypofractionated schedules and the 74 Gy schedule (appendix pp 6, 17). Acute RTOG bowel and bladder symptoms peaked sooner in the hypofractionated schedules than in the control, at 4–5 weeks compared with 7–8 weeks (figure 4, appendix p 18). There was significantly more acute bowel toxicity at the peak in the hypofractionated schedules compared with the control; the proportion of patients reporting RTOG grade 2 or worse bowel toxicity was 176 (25%) of 715 patients with available assessments in the 74 Gy group, 277 (38%) of 720 in the 60 Gy group (vs 74 Gy, p<0·0001), and 270 (38%) of 713 in the 57 Gy group (vs 74 Gy, p<0·0001; figure 4A). However, the distribution of bladder toxicity by grade was similar across all groups; the proportion of patients with available assessments reporting RTOG grade 2 or worse bladder toxicity was 331 (46%) of 715 in the 74 Gy group compared with 356 (49%) of 720 in the 60 Gy group (p=0·34), and 327 (46%) of 713 in the 57 Gy group (p=0·90; figure 4B). By 18 weeks, both bowel and bladder toxicity by RTOG assessment were similar between treatment groups (figure 4). Of 592 patients treated with 74 Gy with available assessments, 15 (3%) and 34 (6%) reported RTOG grade 2 or worse bowel and bladder toxicity, respectively. Corresponding proportions in the 60 Gy group (607 patients treated with available assessments) were 20 (3%) bowel (p=0·38) and 30 (5%) bladder (p=1·00), and in the 57 Gy group (508 patients treated with available assessments) were 15 (3%) bowel (p=0·60) and 30 (5%) bladder (p=0·10).

and bladder toxicity, respectively. Corresponding proportions in the 60 Gy group (607 patients treated with available assessments) were 20 (3%) bowel (p=0·38) and 30 (5%) bladder (p=1·00), and in the 57 Gy group (508 patients treated with available assessments) were 15 (3%) bowel (p=0·60) and 30 (5%) bladder (p=0·10). All radiotherapy schedules showed a low frequency of late bowel and bladder side-effects (figure 5, appendix pp 19–22). 2 years from the start of treatment fewer assessable patients treated with 57 Gy reported RTOG grade 2 or worse bowel symptoms as compared with 74 Gy control (74 Gy, 35 [4%] of 922 vs 57 Gy, 17 [2%] of 962; p=0·0075); however, no difference was observed between the 60 Gy group (28 [3%] of 959) and the 74 Gy group (p=0·31); this was evident across all clinician-reported toxicity scales (RTOG, RMH, and LENT-SOM; figure 5A, appendix pp 8, 19–20). The proportion of RTOG grade 2 or worse bladder symptoms at 2 years was similar across all treatment groups (74 Gy, 13 [1%] of 922; 60 Gy, 16 [2%] of 959 [vs 74 Gy, p=0·71]; 57 Gy, 11 [1%] of 962 [vs 74 Gy, p=0·68]), with no significant difference observed between control and either hypofractionated group on any scale (figure 5B, appendix pp 9, 21–22). Sexual dysfunction was common at baseline and increased with androgen deprivation therapy; although this partially recovered after radiotherapy it remained higher than baseline in all groups (appendix pp 10, 13, 23–24). The proportion of LENT-SOM grade 2 or worse sexual symptoms at 2 years was similar in each treatment group (74 Gy, 550 [67%] of 826 assessable patients; 60 Gy, 562 [65%] of 864 [vs 74 Gy, p=0·54]; 57 Gy, 552 [64%] of 859, [vs 74 Gy, p=0·33).

py it remained higher than baseline in all groups (appendix pp 10, 13, 23–24). The proportion of LENT-SOM grade 2 or worse sexual symptoms at 2 years was similar in each treatment group (74 Gy, 550 [67%] of 826 assessable patients; 60 Gy, 562 [65%] of 864 [vs 74 Gy, p=0·54]; 57 Gy, 552 [64%] of 859, [vs 74 Gy, p=0·33). At 5 years post-radiotherapy, the frequency of grade 2 or worse bowel, bladder, and sexual toxicity across clinician-reported toxicity scales was similar across fractionation schedules (appendix pp 8–10). We identified no significant differences in the incidence of late grade 1 or worse, grade 2 or worse, or grade 3 or worse bowel, bladder, or sexual symptoms in either hypofractionated group compared with control at any timepoint using any clinician-reported toxicity scale (appendix pp 8–13). Estimated cumulative incidences of grade 2 or worse bowel toxicity at 5 years measured with the RTOG scale were 13·7% (111 events) for the 74 Gy group, 11·9% (105 events) for the 60 Gy group (HR compared with 74 Gy 0·94 [95% CI 0·72–1·23], p=0·65) and 11·3% (95 events) for the 57 Gy group (HR compared with 74 Gy 0·84 [0·64–1·11], p=0·22). Estimated cumulative incidences of grade 2 or worse bladder toxicity at 5 years measured with the RTOG scale were 9·1% (66 events) for the 74 Gy group, 11·7% (88 events) for the 60 Gy group (HR compared with 74 Gy 1·34 [95% CI 0·98–1·85], p=0·07) and 6·6% (57 events) for the 57 Gy group (HR 0·85 [0·60–1·21], p=0·37; figure 5, appendix p 9). There was a slightly higher frequency of grade 2 or worse bowel and bladder side-effects in the 60 Gy group compared with 57 Gy at 2 and 5 years on most clinician-reported scales (appendix pp 8–9). Cumulative incidence of LENT-SOM grade 2 or worse bowel side-effects was higher in the 60 Gy group compared with the 57 Gy group (HR 1·39 [95% CI 1·14–1·70], p=0·0010; appendix p 8), but this difference was not seen with other toxicity scales. Cumulative incidence of grade 2 or worse RTOG bladder toxicity was also greater in the 60 Gy group versus the 57 Gy group (HR 1·58 [1·13–2·20], p=0·0073; figure 5, appendix p 9), although this was not seen with other toxicity scales. More serious grade 3 or worse toxicities were rare in all groups (figure 5).

y scales. Cumulative incidence of grade 2 or worse RTOG bladder toxicity was also greater in the 60 Gy group versus the 57 Gy group (HR 1·58 [1·13–2·20], p=0·0073; figure 5, appendix p 9), although this was not seen with other toxicity scales. More serious grade 3 or worse toxicities were rare in all groups (figure 5). At 5 years, the proportions of patients with RTOG grade 3 or worse bowel events were none of 534 patients in the 74 Gy group, two (<1%) of 569 patients in the 60 Gy group, and three of 549 (<1%) in the 57 Gy group; RTOG grade 3 or worse bladder events occurred in two (<1%) of 534 patients in the 74 Gy group, four (<1%) of 569 patients in the 60 Gy group, and five (1%) of 549 patients in the 57 Gy group. The estimated cumulative incidence of grade 3 or worse bowel and bladder adverse events at 5 years was 2% (17 events) and 3% (27 events) in the 74 Gy group, 3% (20 events) and 6% (38 events) in the 60 Gy group, and 4% (23 events) and 3% (21 events) in the 57 Gy group, respectively (figure 5C, D). No treatment-related deaths were reported.

tive incidence of grade 3 or worse bowel and bladder adverse events at 5 years was 2% (17 events) and 3% (27 events) in the 74 Gy group, 3% (20 events) and 6% (38 events) in the 60 Gy group, and 4% (23 events) and 3% (21 events) in the 57 Gy group, respectively (figure 5C, D). No treatment-related deaths were reported. Patient-reported outcomes showed no significant differences in the proportion of small or worse bowel, bladder, or sexual bother at 2 or 5 years (figure 5, appendix pp 8–13). At 5 years, the proportion of assessable patients reporting small or worse bowel bother was 49 (14%) of 341 in the 74 Gy group, 57 (15%) of 375 in the 60 Gy group, and 59 (15%) of 387 in the 57 Gy group. The corresponding figures for bladder bother were 56 (17%) of 333, 63 (17%) of 371, and 60 (16%) of 376, respectively, and for sexual bother were 187 (52%) of 357, 184 (52%) of 357 and 195 (53%) of 370, respectively. The cumulative incidence of late small or worse bother in each hypofractionated schedule was not significantly different from control for bowel, bladder, and sexual symptoms, and there were no significant differences between the hypofractionated groups (figure 5E, F, appendix pp 8–10).

d 195 (53%) of 370, respectively. The cumulative incidence of late small or worse bother in each hypofractionated schedule was not significantly different from control for bowel, bladder, and sexual symptoms, and there were no significant differences between the hypofractionated groups (figure 5E, F, appendix pp 8–10). Discussion In this pre-planned analysis with a median follow-up of over 5 years, we found that the 60 Gy hypofractionated schedule is non-inferior to the 74 Gy conventionally fractionated schedule in terms of time to biochemical or clinical failure for patients with localised prostate cancer. Evaluation of the lower 57 Gy hypofractionated schedule was inconclusive: it cannot be stated to be non-inferior to the 74 Gy control group but it was inferior to the 60 Gy group. Notably, the proportion of patients free from biochemical or clinical failure at 5 years in all treatment groups (88·3% for the 74 Gy group, 90·6% for the 60 Gy group, and 85·9% for the 57 Gy group) were considerably higher than the 71% reported for the 74 Gy group in the national MRC RT01 trial.6 Recent meta-analyses5, 22 identified five relatively small phase 3 trials comparing modest hypofractionation with conventional 1·8–2·0 Gy fractions. A total of 1825 patients were included, although 1153 received relatively low doses of radiotherapy (≤67 Gy in the control groups). There were no consistent differences between randomised groups, and the investigators concluded that larger trials were required to establish the non-inferiority of hypofractionation for clinical effectiveness. In the CHHiP trial, we estimate the α/β ratio to be 1·8 Gy, which is in keeping with previous reports23 and recent large series and meta-analyses which have suggested the α/β ratio to be between 1·4 Gy and 1·93 Gy.24, 25, 26, 27 Our estimate does not account for any time factor related to overall treatment duration and the potential effect of accelerated repopulation.26 The improvement in 5-year disease control for a 3 Gy dose difference between the 57 Gy and 60 Gy groups is in keeping with the review of the six randomised controlled dose-escalation trials previously reported5, 6 and a recent meta-analysis of biologically equivalent dose escalation.28

ct of accelerated repopulation.26 The improvement in 5-year disease control for a 3 Gy dose difference between the 57 Gy and 60 Gy groups is in keeping with the review of the six randomised controlled dose-escalation trials previously reported5, 6 and a recent meta-analysis of biologically equivalent dose escalation.28 Five other contemporary phase 3 studies have reported side-effects related to hypofractionated radiotherapy29, 30, 31, 32, 33 (appendix p 14). We initially reported no differences between the three randomised groups in the frequency of side-effects in the first 457 patients recruited to the CHHiP trial with 2 years follow-up.13, 20 Analysis of acute side-effects in the full CHHiP cohort has confirmed no difference in bladder side-effects except that the wave of toxicity occurs earlier in the hypofractionated groups. However, we have now documented that the peak acute bowel toxicity is greater in patients receiving a hypofractionated schedule, although it is noteworthy that only 2% of patients had grade 3 or worse toxic effects and only 1% had a prolongation of treatment time of more than 1 week, indicating the tolerability of the hypofractionated schedules. An increased acute gastrointestinal reaction in patients treated with hypofractionated radiotherapy has been noted by other investigators29, 32 but not by Norkus and colleagues,33 who treated patients using a 4 day per week schedule. The different treatment schedules and total dose given might account for these differences (appendix p 14). There were no significant differences in late bowel toxicity between the 74 Gy and 60 Gy groups in any of the domains at any timepoint. The 57 Gy group had less grade 1 or worse and grade 2 or worse LENT-SOM bowel and grade 2 or worse RTOG bladder side-effects than the 60 Gy group with outcomes on other toxicity scales supporting this effect, although they were not significant; there were no significant differences between the 57 Gy and 60 Gy groups in sexual function domains. The results are in accord with a recent meta-analysis indicating the general tolerability of hypofractionated radiotherapy,5, 22 although two studies have reported approximate doublings of grade 2 gastrointestinal31 or genitourinary30 side-effects compared with conventional 1·8–2·0 Gy fractions (appendix p 14).

omains. The results are in accord with a recent meta-analysis indicating the general tolerability of hypofractionated radiotherapy,5, 22 although two studies have reported approximate doublings of grade 2 gastrointestinal31 or genitourinary30 side-effects compared with conventional 1·8–2·0 Gy fractions (appendix p 14). Post-radiotherapy symptoms relate to treatment methods, total dose, patient factors, and fractionation schedules. The favourable results in the CHHiP study reflect the mandated radiotherapy technique using an integrated simultaneous boost with relatively narrow planning target volume margins16 and normal tissue dose constraints. The cumulative rate of RTOG grade 2 or worse gastrointestinal side-effects observed by 5 years was 14% compared with 33% for the 74 Gy group in the RT01 trial,34 which used conformal radiotherapy without specified dose constraints. We previously reported complementary findings from the CHHiP trial showing a greater than 50% reduction in bowel bother or distress compared with the RT01 trial using the UCLA-PCI instrument.20 Additionally, the apparently more favourable bladder toxicity results reported in the CHHiP trial compared with other studies might relate to lower total delivered dose29, 30 or amount of bladder and trigone included in the high-dose volume,31 which would be in keeping with our observation of lower side-effects in the 57 Gy compared with 60 Gy hypofractioned groups. This finding suggests that a steep dose–volume association might exist for late bladder complications using hypofractionated schedules.

responsible for central study management and CG was the supervisory statistician at ICR-CTSU. SH was the trial manager. JP was the data manager. EH is responsible for central management of the trial at ICR-CTSU and for all statistical analyses and contributed to manuscript writing. All authors reviewed the manuscript. Declaration of interests We declare no competing interests. Figure 1 Trial profile Figure 2 Biochemical or clinical failure-free survival (A) and overall survival (B) *Number of events reported after 7 years. Figure 3 Univariable subgroup analyses of biochemical or clinical failure comparing 60 Gy (A) and 57 Gy (B) with conventional radiotherapy *Stratified by risk group; all other analyses are unstratified. Figure 4 Acute RTOG toxicity by timepoint and randomised treatment group (A) Prevalence of bowel toxicity and (B) prevalence of bladder toxicity. RTOG=Radiation Therapy Oncology Group. Grade 1+=grade 1 or worse adverse event. Grade 2+=grade 2 or worse adverse event. Grade 3+=grade 3 or worse adverse event. Figure 5 Late bowel and bladder toxicity by timepoint, assessment, and randomised treatment group

of bladder and trigone included in the high-dose volume,31 which would be in keeping with our observation of lower side-effects in the 57 Gy compared with 60 Gy hypofractioned groups. This finding suggests that a steep dose–volume association might exist for late bladder complications using hypofractionated schedules. Strengths of the CHHiP trial include its size and multicentre recruitment. Consistent and quality assured radiotherapy delivery in 40 centres demonstrates the generalisability of the radiotherapy techniques. The study design, using two experimental hypofractionated groups, enables clarity of clinical interpretation. Limitations are that results are primarily applicable to patients receiving short-course androgen deprivation therapy (97% of the men treated) and might not be generalisable for populations who do not receive androgen deprivation therapy, and are most robust for patients with intermediate-risk disease (73% of the men treated), although there was no heterogeneity of effect across all risk groups. We have found a low level of side-effects in all groups. However, further follow-up is required to assess 10-year and 15-year outcomes, and it is possible that differences might yet emerge. Because the hypofractionated and conventional radiotherapy were given over different lengths of time, the study does not address the issue of treatment duration and accelerated repopulation in prostate cancer and our estimate of the α/β ratio of 1·8 Gy does not include a time factor.

that differences might yet emerge. Because the hypofractionated and conventional radiotherapy were given over different lengths of time, the study does not address the issue of treatment duration and accelerated repopulation in prostate cancer and our estimate of the α/β ratio of 1·8 Gy does not include a time factor. Other complementary phase 3 studies treating patients with prostate cancer with radiotherapy alone without androgen deprivation therapy and using hypofractionated radiotherapy with different treatment durations will clarify these issues. In the Netherlands, the HYPRO study (ISRCTN85133859)32 randomised 820 patients with intermediate-risk or high-risk disease to high-dose conventional radiotherapy (78 Gy in 39 fractions) or dose-escalated hypofractionated radiotherapy of 64·6 Gy in 19 fractions given over a period of 6–7 weeks with or without androgen deprivation therapy. In Canada, the PROFIT trial (ISRCTN43853433) has randomised 1204 men with intermediate-risk disease to receive either 60 Gy in 20 fractions over 4 weeks or 78 Gy in 39 fractions. For the low-risk patient group, the RTOG 0415 study recruited 1115 patients who were randomly assigned to receive 70 Gy in 28 fractions over 5·5 weeks or 73·8 Gy in 41 fractions over 8·1 weeks. More extreme forms of hypofractionated radiotherapy are now being studied. The HYPO trial (ISRCTN45905321) will shortly complete recruitment of 1200 men comparing 43·7 Gy in seven fractions over 15–19 days with 78 Gy in 39 fractions over 7·8 weeks and the PACE trial (ISRCTN17627211) compares 36·25 Gy in five fractions over 1–2 weeks with 78 Gy in 39 fractions over 7·8 weeks. Both trials use IGRT, which permits reduced target margins around the prostate, and might reduce treatment side-effects. Outcome results will not be available for several years.

er 7·8 weeks and the PACE trial (ISRCTN17627211) compares 36·25 Gy in five fractions over 1–2 weeks with 78 Gy in 39 fractions over 7·8 weeks. Both trials use IGRT, which permits reduced target margins around the prostate, and might reduce treatment side-effects. Outcome results will not be available for several years. Prostate cancer radiotherapy accounts for 27% of the workload of radiotherapy departments in the UK.35 Radical external-beam radiotherapy was given to 14 364 patients in 2014–15 in England and Wales involving 455 638 attendances. Before commencement of the CHHiP trial, 3 Gy fraction schedules were rarely used, but by 2014–15, after publication of initial safety results,13 19% of patients received 3 Gy hypofractionated schedules (Ball C, personal communication). A uniform change to a 20-fraction schedule could reduce the number of treatment fractions and attendances by over 200 000 annually in the UK. The CHHiP trial results show that the combination of hypofractionation and high-quality radiotherapy techniques gives excellent tumour control, a low level of side-effects, and increased convenience for patients compared with a conventional fractionation schedule. High-dose modest hypofractionation using high-quality treatment methods such as those used in this trial should become a new standard of care for external-beam radiotherapy. The trial results might act as a benchmark for comparison with other treatment approaches, including radical prostatectomy, brachytherapy, and external beam radiation therapy without androgen deprivation therapy or using extreme hypofractionation schedules.

become a new standard of care for external-beam radiotherapy. The trial results might act as a benchmark for comparison with other treatment approaches, including radical prostatectomy, brachytherapy, and external beam radiation therapy without androgen deprivation therapy or using extreme hypofractionation schedules. This online publication has been corrected. The corrected version first appeared at thelancet.com/oncology on June 24, 2016 Supplementary Material Supplementary appendix

become a new standard of care for external-beam radiotherapy. The trial results might act as a benchmark for comparison with other treatment approaches, including radical prostatectomy, brachytherapy, and external beam radiation therapy without androgen deprivation therapy or using extreme hypofractionation schedules. This online publication has been corrected. The corrected version first appeared at thelancet.com/oncology on June 24, 2016 Supplementary Material Supplementary appendix Acknowledgments We thank the patients and all investigators and research support staff at the participating centres. Recognition goes to all the trials unit staff at the Bob Champion Unit and at ICR-CTSU who contributed to the central coordination of the study. We would also like to thank the CHHiP Trial Management Group members past and present and the Independent Data Monitoring Committee (Matthew Sydes [chair], Christopher Tyrell, Peter Barrett-Lee and, previously, Peter Hoskin, Christopher Nutting) and Trial Steering Committee (Anthony Zietman [chair], Soren Bentzen, Vivian Cosgrove, Heather Payne) for overseeing the trial. We acknowledge support of Cancer Research UK (C8262/A7253, C1491/A9895, C1491/A15955, SP2312/021), the Department of Health, the National Institute for Health Research (NIHR) Cancer Research Network, and NHS funding to the NIHR Biomedical Research Centre at the Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, London. Helen Patterson died in 2012 shortly after completion of recruitment to the trial and remains greatly missed by her colleagues. We acknowledge the data supplied by Chris Ball (National Clinical Analysis and Specialised Applications Team, Clatterbridge Cancer Centre NHS Foundation Trust) from the National Radiotherapy Dataset (RTDS).

n died in 2012 shortly after completion of recruitment to the trial and remains greatly missed by her colleagues. We acknowledge the data supplied by Chris Ball (National Clinical Analysis and Specialised Applications Team, Clatterbridge Cancer Centre NHS Foundation Trust) from the National Radiotherapy Dataset (RTDS). Contributors DD is the Chief Investigator and was involved with study design, recruiting patients, data interpretation, and manuscript writing. HMo did the statistical analyses, and contributed to data interpretation and manuscript writing. DD, IS, HMo, CC SH, JP, JG, VK, HMa, ON, HP, CSc, CSo, JS, JT, CG, and EH are members of the CHHiP Trial Management Group responsible for the design and day-to-day oversight of the study and contributed to data interpretation. JL shared the Christie Hospital pilot experience concerning safety of the hypofractionated schedules. ON leads the Physics Quality Assurance Group. HMa, CSo, and MB contributed to radiotherapy planning and quality assurance. DD, IS, AB, DB, JG, VK, PK, JL, ZM, JM-K, JMO'S, MP, CP, HP, CSc, JS, AS, and JT were involved in patient recruitment and data collection. AG was responsible for data collection and study management at the Bob Champion Unit. CC was responsible for central study management and CG was the supervisory statistician at ICR-CTSU. SH was the trial manager. JP was the data manager. EH is responsible for central management of the trial at ICR-CTSU and for all statistical analyses and contributed to manuscript writing. All authors reviewed the manuscript.

(A) Prevalence of bowel toxicity and (B) prevalence of bladder toxicity. RTOG=Radiation Therapy Oncology Group. Grade 1+=grade 1 or worse adverse event. Grade 2+=grade 2 or worse adverse event. Grade 3+=grade 3 or worse adverse event. Figure 5 Late bowel and bladder toxicity by timepoint, assessment, and randomised treatment group Grade distribution of (A) bowel adverse events and (B) bladder adverse events measured with RTOG. Cumulative incidence of (C) bowel adverse events measured with RTOG and (E) bowel symptom scores measured with UCLA PCI/EPIC. Cumulative incidence of (D) bladder adverse events measured with RTOG and (F) bladder symptom scores measured with UCLA PCI/EPIC. Late toxicity data have been included in analyses if they were reported within 6 weeks of the 6 month visit, within 3 months of the 12–24 month visit, and within 6 months of the 36–60 month visit. For UCLA/EPIC, before androgen deprivation therapy data were included if they were reported within 3 months before starting androgen deprivation therapy and within 1 month after starting androgen deprivation therapy. Before radiotherapy data are included if they were reported within 3 months before radiotherapy and no more than 7 days after starting radiotherapy. Time-to-event analyses use all data reported from 6 weeks before the 6 month visit onwards. RTOG=Radiation Therapy Oncology Group scale. UCLA PCI=UCLA Prostate Cancer Index. EPIC=Expanded Prostate Cancer Index Composite. Grade 1+=grade 1 or worse adverse event. Grade 2+=grade 2 or worse adverse event. Grade 3+=grade 3 or worse adverse event. Very small+=score of very small, small, moderate, or big bother. Small+=score of small, moderate, or big bother. Moderate+=score of moderate or worse bother.

panded Prostate Cancer Index Composite. Grade 1+=grade 1 or worse adverse event. Grade 2+=grade 2 or worse adverse event. Grade 3+=grade 3 or worse adverse event. Very small+=score of very small, small, moderate, or big bother. Small+=score of small, moderate, or big bother. Moderate+=score of moderate or worse bother. Table Baseline demographics, clinical characteristics, and treatment details by randomised group

panded Prostate Cancer Index Composite. Grade 1+=grade 1 or worse adverse event. Grade 2+=grade 2 or worse adverse event. Grade 3+=grade 3 or worse adverse event. Very small+=score of very small, small, moderate, or big bother. Small+=score of small, moderate, or big bother. Moderate+=score of moderate or worse bother. Table Baseline demographics, clinical characteristics, and treatment details by randomised group 74 Gy in 37 fractions (N=1065) 60 Gy in 20 fractions (N=1074) 57 Gy in 19 fractions (N=1077) Age (years; range) 69 (48–85) 69 (48–84) 69 (44–83) NCCN risk group Low risk 157 (15%) 164 (15%) 163 (15%) Intermediate risk 779 (73%) 784 (73%) 784 (73%) High risk 129 (12%) 126 (12%) 130 (12%) Gleason score ≤6 371 (35%) 387 (36%) 364 (34%) 7 656 (62%) 658 (61%) 681 (63%) 8 38 (4%) 29 (3%) 32 (3%) Clinical T stage T1a–T1b–T1c–T1x 356 (33%) 422 (39%) 392 (36%) T2a–T2b–T2c–T2x 623 (58%) 561 (52%) 582 (54%) T3a–T3x 85 (8%) 90 (8%) 102 (9%) Missing, unknown, or not done 1 (<1%) 1 (<1%) 1 (<1%) Pre-androgen deprivation therapy PSA (ng/mL) Median 10 (7–14) 10 (7–15) 10 (9–14) Mean 11 (5) 11 (6) 11 (5) <10 510 (48%) 518 (48%) 539 (50%) 10–20 477 (45%) 476 (44%) 462 (43%) ≥20 67 (6%) 75 (7%) 66 (6%) Comorbidity Diabetes 107 (10%) 115 (11%) 120 (11%) Hypertension 400 (38%) 441 (41%) 435 (40%) Inflammatory bowel disease 41 (4%) 39 (4%) 44 (4%) Previous pelvic surgery 86 (8%) 88 (8%) 78 (7%) Symptomatic haemorrhoids 68 (6%) 78 (7%) 63 (6%) Previous TURP 82 (8%) 88 (8%) 89 (8%) Intended androgen deprivation therapy LHRH plus short-term AA 881 (83%) 910 (85%) 909 (84%) 150 mg bicalutamide 144 (14%) 133 (12%) 126 (12%) Other 6 (1%) 1 (<1%) 2 (<1%) None 29 (3%) 27 (3%) 34 (3%) Median duration of androgen deprivation therapy (weeks)* 25 (21–28) 24 (19–27) 23 (20–27) Median time from start of androgen deprivation therapy to radiotherapy (weeks)† 16 (14–20) 16 (14–19) 16 (15–20) Median time from randomisation to start of radiotherapy (weeks)‡ 8 (5–11) 8 (5–11) 7 (5–11) Median duration of radiotherapy (days) 53 (51–55) 29 (28–29) 28 (27–28) Radiotherapy planning Forward planned 626 (59%) 624 (58%) 626 (58%) Inverse planned 304 (29%) 337 (31%) 322 (30%) Unavailable 135 (13%) 113 (11%) 129 (12%) Radiotherapy delivery No image guidance 563 (53%) 568 (53%) 563 (52%) Image-guided 312 (29%) 322 (30%) 326 (30%) Unavailable 190 (18%) 184 (17%) 188 (17%) Data are n (%), mean (SD), or median (IQR), unless otherwise stated. NCCN=National Comprehensive Cancer Network. PSA=prostate-specific antigen. TURP=transurethral resection of the prostate. LHRH=luteinising-hormone-releasing hormone.

) Image-guided 312 (29%) 322 (30%) 326 (30%) Unavailable 190 (18%) 184 (17%) 188 (17%) Data are n (%), mean (SD), or median (IQR), unless otherwise stated. NCCN=National Comprehensive Cancer Network. PSA=prostate-specific antigen. TURP=transurethral resection of the prostate. LHRH=luteinising-hormone-releasing hormone. AA=anti-androgen. * Data presented for patients who received androgen deprivation therapy and a start and end date of treatment is known (n=950 in the 74 Gy group, n=966 in the 60 Gy group, and n=970 in the 57 Gy group). † Data presented for patients who received androgen deprivation therapy and started radiotherapy (n=1008 in the 74 Gy group, n=1022 in the 60 Gy group, and n=1020 in the 57 Gy group). ‡ Data presented for patients who started radiotherapy (n=1043 in the 74 Gy group, n=1051 in the 60 Gy group, and n=1056 in the 57 Gy group).

Research in context Evidence before this study Hyperbaric oxygen is widely used to treat chronic adverse effects of curative radiotherapy in long-term survivors of pelvic malignancy, especially those with rectal bleeding. We searched PubMed from Jan 1, 1970, to Dec 31, 2008, with the terms “clinical trials” AND “hyperbaric oxygen” AND “pelvic” OR “pelvis” OR “bowel” AND radiotherapy”. We identified 11 relevant publications, including reviews, relatively small case studies, and case reports. The results of a single randomised, sham-controlled trial from 2008 (HORTIS) reported significant clinical benefits for patients treated with hyperbaric oxygen 2 weeks post-treatment. A Cochrane intervention review from 2012 confirmed the retrospective nature of much research and detected no other level 1 evidence on which to base an assessment of this treatment modality for patients with chronic radiation-induced bowel dysfunction. Added value of this study The results of this double-blind, sham-controlled clinical trial fail to confirm earlier positive results of hyperbaric therapy for cancer survivors with chronic bowel dysfunction, including a subset of patients with rectal bleeding, after curative radiotherapy for pelvic malignancy, with a similarly sized minority of volunteers in each randomised group reporting some improvement in symptoms. This trial is only the second randomised study in this important patient population. Implications of all the available evidence

The results of this double-blind, sham-controlled clinical trial fail to confirm earlier positive results of hyperbaric therapy for cancer survivors with chronic bowel dysfunction, including a subset of patients with rectal bleeding, after curative radiotherapy for pelvic malignancy, with a similarly sized minority of volunteers in each randomised group reporting some improvement in symptoms. This trial is only the second randomised study in this important patient population. Implications of all the available evidence The contribution of hyperbaric oxygen to the management of a growing population of long-term cancer survivors with severe restrictions on daily activities and impaired quality of life as a consequence of bowel injuries after curative radiotherapy for pelvic malignancy remains unclear and requires more evidence from well designed clinical trials.

ric oxygen to the management of a growing population of long-term cancer survivors with severe restrictions on daily activities and impaired quality of life as a consequence of bowel injuries after curative radiotherapy for pelvic malignancy remains unclear and requires more evidence from well designed clinical trials. Introduction More than 1 million patients worldwide are estimated to need curative radiotherapy for pelvic cancer annually, with up to a third of these patients subsequently developing chronic moderate or severe gastrointestinal symptoms.1 Hyperbaric oxygen has been used as a therapy for symptomatic patients for decades, yet the evidence to support the use of this therapy is based almost exclusively on non-randomised studies.2 The authors of a 2012 Cochrane intervention review3 identified a single well designed, controlled, randomised trial (HORTIS)4 that showed clinical benefit of hyperbaric oxygen therapy in patients with gastrointestinal symptoms after radiotherapy for cancers of the colon, endometrium, uterine corpus, uterine cervix, prostate, or rectum. We conducted a double-blind, randomised controlled trial (HOT2) to test long-term benefits of hyperbaric oxygen therapy in patients with chronic adverse effects of curative pelvic radiotherapy after failure of optimum medical therapy for symptoms of pelvic radiation disease. Methods Study design and participants The HOT2 trial was a randomised, double-blind, sham-controlled phase 3 study involving ten UK hyperbaric medicine facilities registered with the British Hyperbaric Association (appendix p 2).

Introduction More than 1 million patients worldwide are estimated to need curative radiotherapy for pelvic cancer annually, with up to a third of these patients subsequently developing chronic moderate or severe gastrointestinal symptoms.1 Hyperbaric oxygen has been used as a therapy for symptomatic patients for decades, yet the evidence to support the use of this therapy is based almost exclusively on non-randomised studies.2 The authors of a 2012 Cochrane intervention review3 identified a single well designed, controlled, randomised trial (HORTIS)4 that showed clinical benefit of hyperbaric oxygen therapy in patients with gastrointestinal symptoms after radiotherapy for cancers of the colon, endometrium, uterine corpus, uterine cervix, prostate, or rectum. We conducted a double-blind, randomised controlled trial (HOT2) to test long-term benefits of hyperbaric oxygen therapy in patients with chronic adverse effects of curative pelvic radiotherapy after failure of optimum medical therapy for symptoms of pelvic radiation disease. Methods Study design and participants The HOT2 trial was a randomised, double-blind, sham-controlled phase 3 study involving ten UK hyperbaric medicine facilities registered with the British Hyperbaric Association (appendix p 2). Eligible participants were men and women aged 18 years or older with at least grade 2 gastrointestinal symptoms in any category of the Late Effects Normal Tissue scoring system (LENT SOMA) for radiation injury or grade 1 gastrointestinal symptoms with intermittent symptoms attributed to radiotherapy for carcinoma of the rectum, prostate, testis, bladder, uterine cervix, uterine corpus, vagina, vulva, or ovary for at least 12 months before enrolment. Grade 2 symptoms defined by LENT SOMA are moderate, requiring only conservative treatment, whereas grade 3 symptoms are severe, having a substantial negative effect on daily activities, and necessitating more aggressive treatment.5 Participants were screened for eligibility if they presented with gastrointestinal symptoms such as onset or worsening of anal, rectal, atypical abdominal, or back pain; endoscopic evidence of anal, rectal, or sigmoid stricture; worsening of intestinal symptoms after months or years of stable symptoms; worsening of urinary symptoms; or new vaginal bleeding. Potentially eligible participants were assessed using a clinical algorithm6 to identify individuals with symptoms attributable to radiotherapy. Eligible patients could show no evidence of cancer recurrence, as assessed by magnetic resonance imaging of the pelvis, abdomen, and spine. Additional exclusion criteria included medical history of cancer recurrence, rectal surgery, previous hyperbaric oxygen therapy (except for treatment of decompression illness), exposure to bleomycin, claustrophobia, epilepsy, uncontrolled asthma, bullous lung disease, some types of ear surgery, and inability to equalise the middle ear. Individuals with a past history of prostate cancer had to have three serial measurements of serum prostate-specific antigen within the normal concentration range (less than 3 ng/mL for men aged 50–59 years, 4 ng/mL for men aged 60–69 years, 5 ng/mL for men 70 years or older).

, and inability to equalise the middle ear. Individuals with a past history of prostate cancer had to have three serial measurements of serum prostate-specific antigen within the normal concentration range (less than 3 ng/mL for men aged 50–59 years, 4 ng/mL for men aged 60–69 years, 5 ng/mL for men 70 years or older). Patients with symptoms attributed to radiotherapy entered a minimum 3-month period of optimum standard treatment, including antibiotic treatment for small bowel bacterial overgrowth, treatment of bile acid malabsorption,7 lifestyle advice, or several of these interventions, and were supervised by a gastroenterologist. Individuals were considered eligible for the study only if the 3-month period of optimal standard treatment was unsuccessful. All patients provided written informed consent. We listed no criteria for removing a patient from the trial once written informed consent was gained. The study was approved by the MHRA (2008-002152-26) and the NRES Committee North East-York (08/H0903/40). The full case study report and trial protocol are available online.

All patients provided written informed consent. We listed no criteria for removing a patient from the trial once written informed consent was gained. The study was approved by the MHRA (2008-002152-26) and the NRES Committee North East-York (08/H0903/40). The full case study report and trial protocol are available online. Randomisation and masking Eligible participants were randomly assigned (2:1) to receive hyperbaric oxygen treatment or sham. Randomisation was arranged by a telephone call from the treating hyperbaric medicine facility to the Institute of Cancer Research Clinical Trials and Statistics Unit (ICR-CTSU). Randomisation was by computer-generated random permuted blocks (block size of nine and 12), and participants were stratified by centre. Computer-generated lists were used to allocate patients within a block. To deliver the correct treatment, only engineers and technicians operating the hyperbaric chamber were informed of the allocated treatment by the trials office, and care was taken to ensure that patients, clinicians, nurse practitioners, and other health-care professionals associated with patients' care remained masked to treatment allocation. The most important precaution was to disallow any non-trial patient sharing the chamber with a trial patient.

ent by the trials office, and care was taken to ensure that patients, clinicians, nurse practitioners, and other health-care professionals associated with patients' care remained masked to treatment allocation. The most important precaution was to disallow any non-trial patient sharing the chamber with a trial patient. Procedures Participants attended the participating hyperbaric oxygen medicine facility most convenient for them, where they were assessed for suitability for hyperbaric oxygen therapy. Patients in the hyperbaric oxygen therapy group received 40 pressure exposures at 2·4 atmospheres of absolute pressure (ATA; 243 kPa) breathing 100% oxygen for 90 min (including 5-min air breaks at 30-min intervals), whereas patients in the control group received 40 pressure exposures at 1·3 ATA (131 kPa) breathing 21% oxygen (ie, air) for 90 min with two simulated 5-min air breaks. We aimed to deliver the pressure exposures once a day for 5 days per week for 8 weeks for a total of 40 pressure exposures. Additional treatments were delivered beyond the 8-week timeframe if any scheduled sessions were missed. Dose reductions were not permitted.

ie, air) for 90 min with two simulated 5-min air breaks. We aimed to deliver the pressure exposures once a day for 5 days per week for 8 weeks for a total of 40 pressure exposures. Additional treatments were delivered beyond the 8-week timeframe if any scheduled sessions were missed. Dose reductions were not permitted. Participants were asked to complete the modified Inflammatory Bowel Disease Questionnaire (IBDQ)8 and the European Organisation for Research and Treatment of Cancer (EORTC) C30 core quality of life questionnaire (QLQ-C30) and CR38 colorectal module (QLQ-CR38)9, 10 at baseline, 2 weeks after end of treatment, and again at 3 months, 6 months, 9 months, and 12 months after start of treatment. At each timepoint, patients were asked to base their responses on symptoms experienced within the previous 2 weeks. The IBDQ bowel function component (panel) was adopted on the basis of previous application for the characterisation of chronic gastrointestinal morbidity after pelvic radiotherapy in a comparable population of former patients.8, 12 Late radiation-induced adverse effects were clinically assessed within 2 weeks of treatment completion and again at 12 months after start of treatment and were based on the LENT SOMA intestinal and rectal scales of radiation injury (version 2) and 11 questions selected from the Common Terminology Criteria for Adverse Events (CTCAE) gastrointestinal scale (version 4), which were considered most relevant to the study population.5, 13 Telephone interviews were substituted for the minority of patients unable to attend appointments at the Royal Marsden as per protocol.

questions selected from the Common Terminology Criteria for Adverse Events (CTCAE) gastrointestinal scale (version 4), which were considered most relevant to the study population.5, 13 Telephone interviews were substituted for the minority of patients unable to attend appointments at the Royal Marsden as per protocol. Outcomes The two primary clinical endpoints of the study were the change in gastrointestinal symptoms score using the IBDQ and the change in rectal bleeding score (Question 22) in the IBDQ between baseline and 12 months (panel). Secondary clinical endpoints were adverse effects (bowel dysfunction) assessed according to LENT SOMA scales of radiation injury, clinical assessments of gastrointestinal symptoms according to the 11 questions selected from the CTCAE gastrointestinal scale (version 4), and patient self-assessments of quality of life recorded by the EORTC QLQ-C30 core questionnaire and QLQ-CR38 colorectal module between baseline and 12 months.

of radiation injury, clinical assessments of gastrointestinal symptoms according to the 11 questions selected from the CTCAE gastrointestinal scale (version 4), and patient self-assessments of quality of life recorded by the EORTC QLQ-C30 core questionnaire and QLQ-CR38 colorectal module between baseline and 12 months. Statistical analysis The sample size was calculated on the basis of the bowel component of the modified IBDQ primary endpoint. On the basis of results from a previous study,14 we considered a reduction in IBDQ bowel component score of 7 (SD 10) from baseline to 12 months to be clinically relevant. To detect this minimum change at a two-sided significance level of 5% and an estimated power of 80%, we planned to enrol 75 evaluable patients. During the recruitment phase of the trial (February, 2012), the independent data monitoring committee agreed that the significance level of 5% could be split to allow additional analyses in patients reporting rectal bleeding in the IBDQ at baseline. We estimated that 75 evaluable patients would allow us to detect a difference in IBDQ bowel symptom score of 7·5 with 80% power at a two-sided significance level of 3%. On the basis of the assumption that 30 of 75 patients would report grade 2–4 rectal bleeding at baseline on the LENT SOMA Management Scale, corresponding to a score of 1–5 on the IBDQ rectal bleeding scale, this subgroup would allow us to detect a difference of 70% of patients showing any improvement in rectal bleeding (10% in the control group, 80% in the hyperbaric oxygen therapy group) with 80% power at a two-sided significance level of 2%.

ement Scale, corresponding to a score of 1–5 on the IBDQ rectal bleeding scale, this subgroup would allow us to detect a difference of 70% of patients showing any improvement in rectal bleeding (10% in the control group, 80% in the hyperbaric oxygen therapy group) with 80% power at a two-sided significance level of 2%. Analysis of primary endpoints was by modified intention-to-treat, which included analysis of data from forms returned by patients within timeframes agreed to by the independent data monitoring committee. All patients who received any treatment (hyperbaric oxygen therapy or sham) were included in the safety population. Forms were processed as follow-up assessments according to the period that had elapsed between start of treatment and time of completion (table 1). IBDQ questions are scored from 1 to 7 with a low score indicating poorer function or worse symptoms. The bowel component is made up of ten questions, and we used all ten items in the bowel component of the modified IBDQ to analyse overall bowel function and analysed rectal bleeding using the single rectal bleeding question in the modified IBDQ. The difference in change from baseline to 12 months between the two study groups was analysed using the Mann-Whitney U test due to the non-normality of the data. We planned sensitivity analyses of the primary endpoints and the LENT SOMA secondary endpoint in the population of patients who were registered into the study and returned IBDQ forms, irrespective of timelines (intention-to-treat), and in the per-protocol population, which included all patients registered into the study who received at least 32 pressure exposures within a 10-week period. These sensitivity analyses excluded individuals who received less than three treatments per week for at least 2 weeks or who missed five consecutive treatments.

and in the per-protocol population, which included all patients registered into the study who received at least 32 pressure exposures within a 10-week period. These sensitivity analyses excluded individuals who received less than three treatments per week for at least 2 weeks or who missed five consecutive treatments. For the comparison of change in LENT SOMA scores from baseline to 12 months for rectum and intestine (secondary endpoints; table 2) between the active treatment and control groups, we scored individual symptoms within each of three LENT SOMA descriptors (subjective, objective, management) using a four-point scale (with high scores denoting worse symptoms) and summed these scores to develop overall subjective, objective, and management scores for each anatomical site (rectum and intestine). We did no formal statistical analyses of other secondary endpoints (CTCAE scales, EORTC QLQ-C30, and QLQ-CR3810), although the descriptive results were used to strengthen interpretation of changes in the primary endpoints. In an exploratory analysis we tested for a difference in the proportion of patients reporting an improvement in rectal bleeding at 12 months between the two study groups using all available questionnaires (ie, the CTCAE rectal bleeding questions, rectal LENT SOMA objective and management scores, intestinal LENT SOMA management score, and EORTC QLQ-CR38 question 59 “Have you had blood with your stools?”). Patients reporting no rectal bleeding at baseline on an individual scale were excluded from the analysis of that scale. We did two exploratory subgroup analyses of the primary endpoints; one analysis considered the group of patients who received radiotherapy 1–5 years before randomisation, and the other considered the group of patients whose trial treatment was delivered by hood (or monochamber).

xcluded from the analysis of that scale. We did two exploratory subgroup analyses of the primary endpoints; one analysis considered the group of patients who received radiotherapy 1–5 years before randomisation, and the other considered the group of patients whose trial treatment was delivered by hood (or monochamber). We used Stata version 13 for all statistical analyses. The trial is registered with the ISRCTN registry, number ISRCTN86894066. Role of funding source The funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data and had final responsibility for the decision to submit for publication.

We used Stata version 13 for all statistical analyses. The trial is registered with the ISRCTN registry, number ISRCTN86894066. Role of funding source The funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data and had final responsibility for the decision to submit for publication. Results Between Aug 14, 2009, and Oct 23, 2012, 241 patients were given a rigorous initial assessment followed by a 3-month period of optimised medication. 84 participants were considered eligible for trial entry and were randomly assigned to treatment with hyperbaric oxygen (active treatment group; n=55) or with sham control (control group; n=29; figure). The trial ended when all patients had been followed up for 12 months from start of treatment; the final data were collected on Oct 28, 2013. Median follow-up was 13·2 months (IQR 12·4–14·2). Baseline characteristics of the study population are summarised in table 3. There was a small imbalance in the proportion of patients reporting a medical history of rectal bleeding at trial entry, but this was not reflected in the baseline IBDQ or LENT SOMA scales (online case study report). Two-thirds of participants had faecal frequency, incontinence, or both, symptoms that suggest injury to the colon as well as rectum, and a similar proportion reported rectal bleeding. 75 (89%) participants received all 40 planned pressure exposures, and nine (11%) patients received 38 exposures or less (one patient received 38 exposures, one patient received 31 exposures, one patient received 18 exposures, one patient received 11 exposures, one patient received four exposures, one patient received two exposures, and three patients received no exposures). Table 4 details the number of IBDQ and LENT SOMA assessment forms returned within prespecified timeframes.

ient received 31 exposures, one patient received 18 exposures, one patient received 11 exposures, one patient received four exposures, one patient received two exposures, and three patients received no exposures). Table 4 details the number of IBDQ and LENT SOMA assessment forms returned within prespecified timeframes. We found no significant differences in the improvement of overall bowel function (Mann-Whitney U score 0·67; p=0·50) or rectal bleeding (U score 1·69; p=0·092) after 12 months between randomised groups (table 5). Of the patients in the modified intention-to-treat population who reported slight increase in frequency or worse rectal bleeding on IBDQ at baseline, ten (67%) of 15 patients in the control group and 26 (74%) of 35 patients in the treatment group reported an improvement of at least 1 point in the IBDQ rectal bleeding score at 12 months (absolute difference 7·6% [95% CI −20·3 to 35·5]; p=0·58; appendix p 2). Analysis of the IBDQ baseline data did not show imbalances in the pattern or severity of symptoms between treatment groups (figure and appendix p 1).

group reported an improvement of at least 1 point in the IBDQ rectal bleeding score at 12 months (absolute difference 7·6% [95% CI −20·3 to 35·5]; p=0·58; appendix p 2). Analysis of the IBDQ baseline data did not show imbalances in the pattern or severity of symptoms between treatment groups (figure and appendix p 1). Sensitivity analyses of both primary endpoints, including all data returned for the 12-month timepoint irrespective of time of return, showed that the difference in change from baseline to 12 months between the two study groups was consistent with the modified intention-to-treat analysis (U score 0·71 [p=0·48] for overall bowel function; U score 2·06 [p=0·040] for rectal bleeding). Per-protocol analyses of the primary endpoints were also consistent with the modified intention-to-treat analysis (U score 0·94 [p=0·35] for overall bowel function; U score 1·44 [p=0·15] for rectal bleeding; appendix p 3).

s (U score 0·71 [p=0·48] for overall bowel function; U score 2·06 [p=0·040] for rectal bleeding). Per-protocol analyses of the primary endpoints were also consistent with the modified intention-to-treat analysis (U score 0·94 [p=0·35] for overall bowel function; U score 1·44 [p=0·15] for rectal bleeding; appendix p 3). Both treatment groups had a non-significant decrease in subjective LENT SOMA scores for rectum and intestine indicative of an improvement in symptoms (table 6). Sensitivity analyses including all data irrespective of specified timelines gave similar results (U score 1·62 [p=0·11] for rectal LENT SOMA scores; U score −1·41 [p=0·16] for intestinal LENT SOMA scores). Planned descriptive analysis of changes in CTCAE grades at baseline, 2 weeks post-treatment, and at 12 months also did not show differences between the treatment groups (appendix p 4). In view of these negative results, we did not report the planned descriptive analyses of the EORTC QLQ-C30 and QLQ-CR38 since they could not affect the interpretation or conclusions of the trial.

aseline, 2 weeks post-treatment, and at 12 months also did not show differences between the treatment groups (appendix p 4). In view of these negative results, we did not report the planned descriptive analyses of the EORTC QLQ-C30 and QLQ-CR38 since they could not affect the interpretation or conclusions of the trial. Exploratory analysis comparing patient-reported rectal bleeding obtained from the IBDQ questionnaire with scores from other scales including the CTCAE rectal bleeding, rectal LENT SOMA objective and management, intestinal LENT SOMA management, and EORTC QLQ-CR38 questionnaires were in line with those obtained using IBDQ with the exception of the rectal LENT SOMA management score (appendix p 2). Five (100%) of five patients in the control group reported an improvement in rectal bleeding LENT SOMA Management score compared with four (31%) of 13 patients in the hyperbaric oxygen therapy group. Exploratory subgroup analyses of patients who completed radiotherapy 1–5 years before entering the study did not show any difference in IBDQ scores between the two groups (U score 0·59 [p=0·56] for overall bowel function; U score 1·57 [p=0·12] for rectal bleeding). Exploratory subgroup analysis in patients receiving treatment using a hood or monochamber showed no difference in overall bowel function but did suggest a difference in rectal bleeding (U score −0·31 [p=0·76] for overall bowel function; U score 2·9 [p=0·004] for rectal bleeding).

ore 1·57 [p=0·12] for rectal bleeding). Exploratory subgroup analysis in patients receiving treatment using a hood or monochamber showed no difference in overall bowel function but did suggest a difference in rectal bleeding (U score −0·31 [p=0·76] for overall bowel function; U score 2·9 [p=0·004] for rectal bleeding). We analysed toxic effects in the safety population, which included the 81 patients who received at least one treatment (53 in the hyperbaric oxygen therapy group and 28 in the sham control group). Treatment-emergent toxic effects were reported for 41 (51%) of 81 patients receiving at least one treatment in either treatment group. The most commonly reported adverse events were eye refractive change, including myopia (three [11%] of 28 patients in the control group vs 16 [30%] of 53 patients in the treatment group), increased fatigue or tiredness (three [11%] vs two [4%]), and ear pain or barotrauma (six [21%] vs 15 [28%]). Eight serious adverse events were reported in eight patients: two were reported in two patients in the control group (tonsillitis requiring surgery [grade 3]; recurrent cancer of the vulva [grade 4]) and six serious adverse events were reported in six patients in the treatment group (malignant spinal cord compression requiring surgery [grade3]; malignant paraortic lymph node involvement requiring surgery [grade 3]; recurrence of vomiting and dehydration [grade 3]; diarrhoea and fever associated with Campylobacter infection [grade 3]; recurrence of abdominal pain, bloating, diarrhoea, and urinary tract infection [grade 3]; aneurysm [grade 4]). No reported adverse event was considered related to treatment. One patient had an improvement in eyesight during treatment. Only two of the patients who stopped treatment early did so for reasons related to treatment (anxiety). No treatment-related deaths were noted.

nary tract infection [grade 3]; aneurysm [grade 4]). No reported adverse event was considered related to treatment. One patient had an improvement in eyesight during treatment. Only two of the patients who stopped treatment early did so for reasons related to treatment (anxiety). No treatment-related deaths were noted. Discussion Despite some clinical evidence and plausible pathophysiological mechanisms justifying an expectation of therapeutic effect of hyperbaric oxygen therapy, the HOT2 trial results detected no clinically relevant benefit of hyperbaric oxygen therapy in individuals with a wide range of chronic gastrointestinal dysfunction, including rectal bleeding, after curative radiotherapy for pelvic malignancy. The modified IBDQ was adopted to assess the primary outcome in HOT2, given its successful application in characterising patients with gastrointestinal dysfunction after pelvic radiotherapy.6, 8, 11, 15, 16, 17 None of the exploratory analyses using other instruments to measure rectal bleeding, including LENT SOMA, CTCAE, and EORTC, suggested any clinical benefit of hyperbaric oxygen.

2, given its successful application in characterising patients with gastrointestinal dysfunction after pelvic radiotherapy.6, 8, 11, 15, 16, 17 None of the exploratory analyses using other instruments to measure rectal bleeding, including LENT SOMA, CTCAE, and EORTC, suggested any clinical benefit of hyperbaric oxygen. Pelvic radiation syndrome describes a range of physiological disorders that often take a remittent course and are best characterised by investigation according to structured algorithms before treatment.6 The symptoms include pain, bloating, flatulence, diarrhoea, urgency, faecal incontinence, and rectal bleeding. Histologically, progressive obliterative endarteritis is a classic feature and ischaemic atrophy is an important element of the pathophysiology, but direct radiation effects on other tissue elements, including epithelia, also contribute to symptoms.18 The tissues rendered ischaemic by vascular atrophy do not share the steep oxygen gradients that stimulate angiogenesis in acute surgical wounds unless these gradients are artificially introduced.19 In studies of animal and human skin,20, 21 hyperbaric oxygen therapy has been shown to restore virtually normal small vessel density and transcutaneous oxygen tension after high-dose radiotherapy, an effect that peaked after 20–30 treatments in human beings. The proposed therapeutic mechanisms include marrow stem-cell mobilisation and consequent vasculogenesis, although our results do not suggest that these processes, if activated by hyperbaric oxygen, were of therapeutic value.22

n after high-dose radiotherapy, an effect that peaked after 20–30 treatments in human beings. The proposed therapeutic mechanisms include marrow stem-cell mobilisation and consequent vasculogenesis, although our results do not suggest that these processes, if activated by hyperbaric oxygen, were of therapeutic value.22 Our trial results are inconsistent with a long history of striking anecdotes and reviews of non-randomised studies.2, 23, 24, 25 A Cochrane intervention review3 identified two randomised trials testing hyperbaric oxygen in patients with chronic gastrointestinal symptoms after pelvic radiotherapy, but the analysis was restricted to the HORTIS trial4 because of a high risk of bias identified in the other study. The HORTIS trial randomly assigned 150 patients from Mexico, Turkey, South Africa, and Australia with a 3-month or longer medical history of radiation proctitis to breathe air at 1·1 ATA (sham group) or 100% oxygen at 2·0 ATA (active treatment group) for 90 min for 30 sessions within 6–8 weeks, with an additional ten sessions depending on individual responses. Improvements in the LENT SOMA score (primary endpoint) were found in 120 evaluable patients with radiation proctitis; patients in the active treatment group recording significantly lower average scores than patients in the sham group (p=0·015), with an estimated difference of 1·93 points (95% CI 0·38–3·48). The HORTIS investigators also reported a significant benefit of hyperbaric oxygen in patients with bowel bother, a group of symptoms that include faecal incontinence, faecal urgency, and pain. The authors of the Cochrane review interpreted these results as non-significant and sensitive to randomised patients excluded from primary analysis but concluded that HORTIS supported the continued use of hyperbaric oxygen for patients with radiation proctitis.

hat include faecal incontinence, faecal urgency, and pain. The authors of the Cochrane review interpreted these results as non-significant and sensitive to randomised patients excluded from primary analysis but concluded that HORTIS supported the continued use of hyperbaric oxygen for patients with radiation proctitis. It is unclear why the results of HOT2 fail to reproduce the HORTIS4 findings. Although a single-centre study in terms of patient referral and selection, HOT2 trial participants were treated at one of ten UK-registered hyperbaric facilities. Patient selection was unusually rigorous, including assessment by a gastroenterologist specialised in radiation enteropathy and a 3-month run-in period of optimised oral drug treatment to ensure that eligible patients had radiation-induced symptoms that could not be controlled by standard measures. The trial population is considered representative of patients with radiation enteropathy in terms of their symptoms, although patients with severe faecal incontinence or transfusion-dependent rectal bleeding are likely to be under-represented, the former being too restricted to leave their homes and the latter considered too seriously at risk to be considered by their primary physicians for entry into a trial with a sham treatment option. We assessed 20 characteristics relating to bowel dysfunction at baseline, and despite small imbalances in the proportion of patients with a medical history of rectal bleeding (23 [79%] of 29 patients in the control group vs 34 [62%] of 55 patients in the treatment group), this imbalance did not apply to baseline IBDQ rectal bleeding scores analysed as the primary endpoint. In other respects, symptom duration in HOT2, at a median 3·7 years (IQR 2·4–6·8) post-radiotherapy, is consistent with that of the HORTIS population, and HOT2 patient characteristics are well balanced between randomised groups.

ance did not apply to baseline IBDQ rectal bleeding scores analysed as the primary endpoint. In other respects, symptom duration in HOT2, at a median 3·7 years (IQR 2·4–6·8) post-radiotherapy, is consistent with that of the HORTIS population, and HOT2 patient characteristics are well balanced between randomised groups. In a literature review26 of ten retrospective studies published between 1960 and 2004 reporting favourable results of hyperbaric oxygen therapy for patients with radiation proctitis, patients had an average of 24 treatments each. The randomised, sham-controlled HORTIS trial4 reported beneficial clinical effects of hyperbaric oxygen after 30–40 treatments. Hence, the 40 treatments used in HOT2 can be considered an appropriate test of hyperbaric oxygen. Compliance with treatment in HOT2 was reasonably high, with 75 (88%) of 84 patients eligible for inclusion in the intention-to-treat population, as required by the analysis plan. Post-hoc analyses in the per-protocol population failed to detect any treatment effect. Pre-trial investigations designed to exclude patients with residual malignant disease ensured that only three patients developed cancer recurrence while participating in the study.

treat population, as required by the analysis plan. Post-hoc analyses in the per-protocol population failed to detect any treatment effect. Pre-trial investigations designed to exclude patients with residual malignant disease ensured that only three patients developed cancer recurrence while participating in the study. Other relevant points of difference between the HORTIS4 and HOT2 trials include the immediate post-treatment timepoint for the primary analysis in HORTIS, compared with the primary analysis at 12 months post-treatment in HOT2. Exploratory analyses of the 2-week post-treatment effects in HOT2 showed no difference between randomised groups for any of the primary or secondary endpoints. This included change in total LENT SOMA score, which was the primary endpoint analysed by the HORTIS investigators. Unlike HOT2, in which we analysed the primary endpoint in a modified intention-to-treat population, the primary analysis in HORTIS excluded 30 of 150 randomised patients who did not complete the treatment protocol (plus one patient lost to follow-up), although unplanned analyses of the outcomes of clinical assessments by intention-to-treat were also consistent with a beneficial effect of hyperbaric oxygen in HORTIS. In other exploratory analyses of HOT2 endpoints (including subgroup analysis of patients completing radiotherapy 1–5 years before randomisation and subgroup analysis of patients receiving treatment by hood or hyperbaric chamber rather than mask), we could not identify variables that might explain differences in reported outcomes between these two trials. Randomisation appears to have resulted in a reasonably even distribution of patient characteristics between the treatment and control groups in our study, to the extent that the total effect of these covariates would not be expected to mask any effect of hyperbaric oxygen. The very small number of patients with transfusion-dependent rectal bleeding in this study prevents us from commenting on the use of hyperbaric oxygen therapy for patients referred for this potentially life-threatening complication.

fect of these covariates would not be expected to mask any effect of hyperbaric oxygen. The very small number of patients with transfusion-dependent rectal bleeding in this study prevents us from commenting on the use of hyperbaric oxygen therapy for patients referred for this potentially life-threatening complication. Our trial was designed to have a power of 80%; with 69 evaluable patients, we had a power of about 75% to detect a difference of the magnitude expected in the first of the primary endpoints. We did not detect a clinically or statistically significant clinical benefit of hyperbaric oxygen therapy for patients with chronic gastrointestinal dysfunction, including rectal bleeding, after pelvic radiotherapy. The findings contrast with previous reports, highlighting an urgent need for more level 1 evidence to determine with confidence whether hyperbaric oxygen therapy can be recommended as a standard of care for this group of patients. Supplementary Material Supplementary appendix Acknowledgments We thank the volunteers and site personnel participating in this study for their support and commitment. Cancer Research UK, the Department of Health, and the Welsh Assembly Government jointly funded the work (C181/A9694). Statistical input was undertaken at the ICR Clinical Trials and Statistics Unit with funding from Cancer Research UK. The National Institute for Health Research Biomedical Research Centre at the Royal Marsden and the ICR received funding from the National Health Service.

Government jointly funded the work (C181/A9694). Statistical input was undertaken at the ICR Clinical Trials and Statistics Unit with funding from Cancer Research UK. The National Institute for Health Research Biomedical Research Centre at the Royal Marsden and the ICR received funding from the National Health Service. Contributors JY was the chief investigator of the trial, was involved in study design, was responsible for the clinical supervision of patients and performance of the study, and contributed to the preparation and writing of the report. MG, GRS, HJA, BEB, PB, OF, LG, JH, MI, GL, SM, DM, CELP, SP, and GS were were study investigators responsible for patient recruitment, clinical supervision, and treatment of patients, were involved in the acquisition, analysis, and interpretation of the data, and contributed to the writing of the report. LM was responsible for the acquisition and analysis of the data, and contributed to the writing of the report. Declaration of interests We declare no competing interests. Figure Trial profile IBDQ=modified Inflammatory Bowel Disease Questionnaire. *Includes one patient in the control group and two patients in the hyperbaric oxygen therapy group who received no treatment. Table 1 Classification of patient case report forms according to time of return after start of treatment

Declaration of interests We declare no competing interests. Figure Trial profile IBDQ=modified Inflammatory Bowel Disease Questionnaire. *Includes one patient in the control group and two patients in the hyperbaric oxygen therapy group who received no treatment. Table 1 Classification of patient case report forms according to time of return after start of treatment Form to be processed as Completed after 2-week assessment and less than 4·5 months after start of treatment 3-month assessment Completed 4·5–7·5 months after start of treatment 6-month assessment Completed 7·5–10·5 months after start of treatment 9-month assessment Completed 10·5–14 months after start of treatment 12-month assessment Completed more than 14 months after start of treatment Exclude forms Table 2 Subjective parameter score of the 4-point rectal and intestinal LENT SOMA scoring scales5

th assessment Completed 7·5–10·5 months after start of treatment 9-month assessment Completed 10·5–14 months after start of treatment 12-month assessment Completed more than 14 months after start of treatment Exclude forms Table 2 Subjective parameter score of the 4-point rectal and intestinal LENT SOMA scoring scales5 Score of 1 Score of 2 Score of 3 Score of 4 Rectal* Stool frequency 2–4 per day 5–8 per day >8 per day Uncontrolled diarrhoea Sphincter control Occasional Intermittent Persistent Refractory Pain Occasional and minimal Intermittent and tolerable Persistent and intense Refractory and excruciating Tenesmus Occasional urgency Intermittent urgency Persistent urgency Refractory Mucosal loss Occasional Intermittent Persistent Refractory Intestinal† Stool frequency 2–4 per day 5–8 per day >8 per day Refractory diarrhoea Stool consistency Bulky Loose Mucous, dark, watery ·· Pain Occasional and minimal Intermittent and tolerable Persistent and intense Refractory/rebound Constipation 3–4 per week Only twice per week Only once per week No stool in 10 days LENT SOMA=Late Effects in Normal Tissues Subjective, Objective, Management and Analytic scales. * The possible range of summed results for the five questions is 0–20, where 0 indicates that no symptoms are present and 20 represents the worst possible symptomatology. † The possible range of summed results for the four questions is 0–15, where 0 indicates that no symptoms are present and 15 represents the worst possible symptomatology (there is no grade 4 stool consistency).

* The possible range of summed results for the five questions is 0–20, where 0 indicates that no symptoms are present and 20 represents the worst possible symptomatology. † The possible range of summed results for the four questions is 0–15, where 0 indicates that no symptoms are present and 15 represents the worst possible symptomatology (there is no grade 4 stool consistency). Table 3 Patient characteristics at pretrial eligibility assessments

* The possible range of summed results for the five questions is 0–20, where 0 indicates that no symptoms are present and 20 represents the worst possible symptomatology. † The possible range of summed results for the four questions is 0–15, where 0 indicates that no symptoms are present and 15 represents the worst possible symptomatology (there is no grade 4 stool consistency). Table 3 Patient characteristics at pretrial eligibility assessments Sham control (n=29) Hyperbaric oxygen (n=55) Age Mean 62·0 (11) 62·3 (11) Median 63·7 (53·6–69·9) 63·7 (53·9–71·2) Range 37·3–79·3 34·5–80·9 Sex Male 14 (48%) 23 (42%) Female 15 (52%) 32 (58%) Origin of cancer Prostate 12 (41%) 21 (38%) Anus 4 (14%) 4 (7%) Vagina 3 (10%) 1 (2%) Cervix 5 (17%) 17 (31%) Uterus 3 (10%) 8 (15%) Other* 2 (7%) 4 (7%) Medical history Back pain 3 (10%) 7 (13%) Bloating 18 (62%) 30 (55%) Constipation 5 (17%) 5 (9%) Cramps or abdominal pain 14 (48%) 38 (69%) Diarrhoea 14 (48%) 30 (55%) Faecal incontinence 19 (66%) 35 (64%) Frequency 18 (62%) 38 (69%) Mucus discharge 10 (34%) 21 (38%) Nausea 4 (14%) 13 (24%) Other 6 (21%) 8 (15%) Rectal bleeding 23 (79%) 34 (62%) Rectal or perineal pain 8 (28%) 10 (18%) Steatorrhoea 1 (3%) 10 (18%) Subacute obstructive symptoms 3 (10%) 14 (25%) Tenesmus 18 (62%) 35 (64%) Unable to differentiate need to defecate or pass urine 1 (3%) 2 (4%) Unable to differentiate solid or liquid stool 6 (21%) 11 (20%) Urgency 20 (69%) 48 (87%) Weight loss 2 (7%) 10 (18%) Wind 17 (59%) 39 (71%) Time since pelvic radiotherapy (years) Median 3·9 (2·5–5·7) 3·5 (2·3–9·7) Range 1·5–21·2 1·2–34·0 Data are n (%) or median (IQR).

tiate need to defecate or pass urine 1 (3%) 2 (4%) Unable to differentiate solid or liquid stool 6 (21%) 11 (20%) Urgency 20 (69%) 48 (87%) Weight loss 2 (7%) 10 (18%) Wind 17 (59%) 39 (71%) Time since pelvic radiotherapy (years) Median 3·9 (2·5–5·7) 3·5 (2·3–9·7) Range 1·5–21·2 1·2–34·0 Data are n (%) or median (IQR). * Others were anal canal (n=1) and vulva (n=1) in the control group and retroperitoneum (n=1), pelvis (n=1), rectum (n=1), and bladder (n=1) in the hyperbaric oxygen therapy group. Table 4 Overall returns of IBDQ and LENT SOMA assessment forms and those returned within prespecified timeframes Baseline 2 weeks 3 months 6 months 9 months 12 months IBDQ forms returned 84 75 79 78 78 79 IBDQ forms returned* 84 75 68 76 74 74 Rectal LENT SOMA 84 78 ·· ·· ·· 79 Rectal LENT SOMA* 84 78 ·· ·· ·· 72 Intestinal LENT SOMA 84 78 ·· ·· ·· 79 Intestinal LENT SOMA* 84 78 ·· ·· ·· 72 CTCAE 83 78 ·· ·· ·· 79 CTCAE* 83 78 ·· ·· ·· 72 QLQ-C30 84 ·· 77 78 78 79 QLQ-C30* 84 ·· 65 76 75 74 QLQ-CR38 84 ·· 77 78 78 79 QLQ-CR38* 84 ·· 65 76 74 74 IBDQ=Inflammatory Bowel Disease Questionnaire. LENT SOMA=Late Effects in Normal Tissues Subjective, Objective, Management and Analytic scales. CTCAE=Common Terminology Criteria for Adverse Events. * Includes only forms returned within the prespecified permissible timeframes detailed in table 1. Table 5 Median changes in the IBDQ bowel function component and IBDQ rectal bleeding scores from baseline to 12 months in patients assessed within 10–14 months

Baseline 2 weeks 3 months 6 months 9 months 12 months IBDQ forms returned 84 75 79 78 78 79 IBDQ forms returned* 84 75 68 76 74 74 Rectal LENT SOMA 84 78 ·· ·· ·· 79 Rectal LENT SOMA* 84 78 ·· ·· ·· 72 Intestinal LENT SOMA 84 78 ·· ·· ·· 79 Intestinal LENT SOMA* 84 78 ·· ·· ·· 72 CTCAE 83 78 ·· ·· ·· 79 CTCAE* 83 78 ·· ·· ·· 72 QLQ-C30 84 ·· 77 78 78 79 QLQ-C30* 84 ·· 65 76 75 74 QLQ-CR38 84 ·· 77 78 78 79 QLQ-CR38* 84 ·· 65 76 74 74 IBDQ=Inflammatory Bowel Disease Questionnaire. LENT SOMA=Late Effects in Normal Tissues Subjective, Objective, Management and Analytic scales. CTCAE=Common Terminology Criteria for Adverse Events. * Includes only forms returned within the prespecified permissible timeframes detailed in table 1. Table 5 Median changes in the IBDQ bowel function component and IBDQ rectal bleeding scores from baseline to 12 months in patients assessed within 10–14 months Median score at baseline (IQR) Median score at 12 months (IQR) Median change from baseline to 12 months (IQR) Mann-Whitney testUscore p value Sham Hyperbaric oxygen Sham Hyperbaric oxygen Sham Hyperbaric oxygen Bowel function* 51 (44 to 59) 48 (42 to 52) 53 (40 to 59) 51 (36 to 62) 4 (−6 to 9) 3·5 (−3 to 11) 0·67 0·50 Rectal bleeding† 3 (2 to 4) 3 (2 to 4) 4 (2 to 6) 6 (3 to 7) 1 (1 to 2) 3 (1 to 3) 1·69 0·092 Positive changes indicate higher median Inflammatory Bowel Disease Questionnaire (IBDQ) scores, which signify improvement in symptoms. * Analysis included 23 patients in the sham control group and 46 patients in the hyperbaric oxygen treatment group.

Median score at baseline (IQR) Median score at 12 months (IQR) Median change from baseline to 12 months (IQR) Mann-Whitney testUscore p value Sham Hyperbaric oxygen Sham Hyperbaric oxygen Sham Hyperbaric oxygen Bowel function* 51 (44 to 59) 48 (42 to 52) 53 (40 to 59) 51 (36 to 62) 4 (−6 to 9) 3·5 (−3 to 11) 0·67 0·50 Rectal bleeding† 3 (2 to 4) 3 (2 to 4) 4 (2 to 6) 6 (3 to 7) 1 (1 to 2) 3 (1 to 3) 1·69 0·092 Positive changes indicate higher median Inflammatory Bowel Disease Questionnaire (IBDQ) scores, which signify improvement in symptoms. * Analysis included 23 patients in the sham control group and 46 patients in the hyperbaric oxygen treatment group. † Analysis included 11 patients in the sham control group and 29 patients in the hyperbaric oxygen treatment group. 40 (47%) of 84 patients scored grade 1–5 (clinically significant) rectal bleeding on the IBDQ bowel function component, compared with 57 (68%) patients reporting any rectal bleeding in their medical history (panel). 46 (55%) patients had grade 1–3 rectal bleeding at baseline according to Common Terminology Criteria for Adverse Events scoring system, and 47 (59%) patients had grade 1–3 rectal bleeding in response to EORTC QLQ-CR38 Question 59. Table 6 Median changes in LENT SOMA aggregate parameter scores for rectum and intestine from baseline to 12 months in patients assessed within 10–14 months

† Analysis included 11 patients in the sham control group and 29 patients in the hyperbaric oxygen treatment group. 40 (47%) of 84 patients scored grade 1–5 (clinically significant) rectal bleeding on the IBDQ bowel function component, compared with 57 (68%) patients reporting any rectal bleeding in their medical history (panel). 46 (55%) patients had grade 1–3 rectal bleeding at baseline according to Common Terminology Criteria for Adverse Events scoring system, and 47 (59%) patients had grade 1–3 rectal bleeding in response to EORTC QLQ-CR38 Question 59. Table 6 Median changes in LENT SOMA aggregate parameter scores for rectum and intestine from baseline to 12 months in patients assessed within 10–14 months Median score at baseline (IQR) Median score at 12 months (IQR) Median change from baseline to 12 months (IQR) Mann-Whitney testUscore p value Sham (n=26) Hyperbaric oxygen (n=46) Sham (n=26) Hyperbaric oxygen (n=46) Sham (n=26) Hyperbaric oxygen (n=46) Rectum 6 (5 to 8) 6 (4 to 8) 4·5 (2 to 8) 5 (3 to 8) −1·5 (−4 to 0) −1 (−2 to 1) 1·56 0·12 Intestine 2·5 (1 to 4) 4 (2 to 5) 1 (1 to 4) 2·5 (1 to 4) 0 (−1 to 1) 0 (−2 to 0) −1·30 0·20 High scores indicate worse symptoms; a negative change indicates a lower score at 12 months, signifying improvement in function. LENT SOMA=Late Effects in Normal Tissues Subjective, Objective, Management and Analytic scale. 46 (55%) patients had grade 1–4 rectal bleeding at baseline as assessed by the LENT SOMA scale.

High scores indicate worse symptoms; a negative change indicates a lower score at 12 months, signifying improvement in function. LENT SOMA=Late Effects in Normal Tissues Subjective, Objective, Management and Analytic scale. 46 (55%) patients had grade 1–4 rectal bleeding at baseline as assessed by the LENT SOMA scale. Panel Bowel function component of the modified Inflammatory Bowel Disease Questionnaire (IBDQ)11 Question 1: Have you had your bowel open? Question 5: Have you had loose bowel movements? Question 9: Have you been troubled by pain in your bottom? Question 13: Have you had cramp in tummy or bottom? Question 17: Have you passed a large amount of gas? Question 20: Have you been troubled by bloating? Question 22: Have you had a problem with bleeding from your bottom? Question 24: Have you felt like you need to have your bowel open but nothing happens? Question 26: Have you been troubled by accidental soiling? Question 29: Have you felt disgusted about your bowel problems? Each question is linked to the following response options on a 7-point graded scale: 1=more than ever before; 2=extremely frequently; 3=very frequently; 4=moderate increase in frequency; 5=some increase in frequency; 6=slight increase in frequency; 7=normal/not at all. The possible range of summed results for the 10 questions is 10–70, where 10 represents the most severe, and 70 the least severe, levels of effect, this metric represents a co-primary endpoint. Question 22, analysed separately, is the second co-primary endpoint.

Research in context Evidence before this study Standard treatment for anal cancer is chemoradiotherapy. Guidelines previously recommended assessment of tumour response and biopsy at 6–12 weeks after starting treatment on the basis of several randomised trials and a population study. On the basis of this evidence salvage surgery was recommended to be done on patients with residual tumour shortly after completing chemoradiotherapy. However, present guidelines offer discordant advice on how often and when biopsy should be done and offer uncertainty over the optimum timing of response. Added value of this study Our post-hoc analysis of our trial data shows that tumour assessment at 26 weeks from the start of chemoradiotherapy is most strongly associated with progression and mortality compared with any earlier assessment. Many patients who do not have a complete clinical response at 11 weeks from the start of chemoradiotherapy do respond by 26 weeks and are therefore considered slow to respond to treatment. Implications of all the available evidence Present guidelines on the best timing of tumour response for anal cancer should be strengthened and an assessment of response at 26 weeks should be used in future treatment trials, and should be explored as a surrogate endpoint for survival and progression.

Our post-hoc analysis of our trial data shows that tumour assessment at 26 weeks from the start of chemoradiotherapy is most strongly associated with progression and mortality compared with any earlier assessment. Many patients who do not have a complete clinical response at 11 weeks from the start of chemoradiotherapy do respond by 26 weeks and are therefore considered slow to respond to treatment. Implications of all the available evidence Present guidelines on the best timing of tumour response for anal cancer should be strengthened and an assessment of response at 26 weeks should be used in future treatment trials, and should be explored as a surrogate endpoint for survival and progression. Introduction Standard treatment for anal cancer is chemoradiotherapy with concurrent fluorouracil and mitomycin.1, 2, 3, 4 Randomised phase 3 trials by the Radiotherapy Therapy Oncology Group (RTOG 98–11),5 the Action Clinique Coordonnées en Cancérologie Digestive (ACCORD 03) trial6 and the ACT II trial7 did not show benefit in terms of progression-free survival by increasing the radiotherapy boost dose,6 replacing mitomycin with cisplatin during chemoradiotherapy,5, 6 or by giving maintenance chemoradiotherapy after chemoradiotherapy.7

ique Coordonnées en Cancérologie Digestive (ACCORD 03) trial6 and the ACT II trial7 did not show benefit in terms of progression-free survival by increasing the radiotherapy boost dose,6 replacing mitomycin with cisplatin during chemoradiotherapy,5, 6 or by giving maintenance chemoradiotherapy after chemoradiotherapy.7 Guidelines for anal cancer recommend assessment of response at 6–12 weeks after starting treatment but discordance exists regarding consideration of early biopsy.8, 9, 10 Several randomised trials and one population study used a single response assessment as an early endpoint (4–8 weeks after the completion of trial treatments) and showed that 10–60% of patients did not respond to chemoradiotherapy.1, 2, 3, 11 On the basis of this evidence, salvage surgery can be done on patients who have residual tumour after completion of chemoradiotherapy.3, 11

le response assessment as an early endpoint (4–8 weeks after the completion of trial treatments) and showed that 10–60% of patients did not respond to chemoradiotherapy.1, 2, 3, 11 On the basis of this evidence, salvage surgery can be done on patients who have residual tumour after completion of chemoradiotherapy.3, 11 One of the primary endpoints of the ACT II trial was to assess whether cisplatin given concurrently with fluorouracil and radiotherapy produces a higher complete clinical response than mitomycin alone. A complete clinical response is defined as no evidence of residual tumour or nodal disease. Response was assessed at three timepoints up to 26 weeks from the start of chemoradiotherapy (figure 1). This post-hoc analysis of ACT II aimed to assess the difference in response at each of these timepoints and the association between having complete clinical response and progression-free or overall survival at each of these timepoints. This evidence can then be used indirectly to consider the best time for surgery. To the best of our knowledge, this assessment has not been prospectively investigated before.

se timepoints and the association between having complete clinical response and progression-free or overall survival at each of these timepoints. This evidence can then be used indirectly to consider the best time for surgery. To the best of our knowledge, this assessment has not been prospectively investigated before. Methods Study design and participants ACT II was a randomised phase 3 trial done in 59 centres in the UK and designed to investigate whether replacing mitomycin with cisplatin in the chemoradiotherapy schedule improves the complete response rate, and whether maintenance chemotherapy (fluorouracil and cisplatin) after chemoradiotherapy increases progression-free survival. Patients were randomised to one of four groups to receive mitomycin (12 mg/m2 on day 1) or cisplatin (60 mg/m2 on days 1 and 29) with fluorouracil (1000 mg/m2 per day on days 1–4 and 29–32) and radiotherapy (50·4 Gy in 28 day fractions); with or without two courses of maintenance chemotherapy (fluorouracil and cisplatin at weeks 11 and 14). The full trial methods and results have been reported previously.7

r cisplatin (60 mg/m2 on days 1 and 29) with fluorouracil (1000 mg/m2 per day on days 1–4 and 29–32) and radiotherapy (50·4 Gy in 28 day fractions); with or without two courses of maintenance chemotherapy (fluorouracil and cisplatin at weeks 11 and 14). The full trial methods and results have been reported previously.7 Participants Patients were eligible if they had newly diagnosed, histologically confirmed squamous cell carcinoma basaloid or cloacogenic carcinoma of the anal canal or margin, without metastatic disease, and considered fit for trial treatment; a glomerular filtration rate of 50 mL/min or more; acceptable haematological parameters (haemoglobin >100 g per L, platelets >100 × 109 per L, white blood cells >3 × 109 per L); liver function tests within twice normal range; and adequate cardiac function. There were no age limits. Exclusion criteria were other major malignancies likely to compromise life expectancy or completion of trial therapy, comorbidity including HIV-positive status and cardiac diseases, previous complete local excision, and previous radiotherapy to the pelvis. All patients provided written informed consent and the trial was approved by UK research ethics committees.

cies likely to compromise life expectancy or completion of trial therapy, comorbidity including HIV-positive status and cardiac diseases, previous complete local excision, and previous radiotherapy to the pelvis. All patients provided written informed consent and the trial was approved by UK research ethics committees. Randomisation Randomisation was done by minimisation and stratified by site, T and N stage, sex, age, and renal function. Allocation was concealed by use of a computer program (in the trial co-ordinating centre) to generate the treatment allocation. Site staff would telephone the trial co-ordingating centre, and the assigned treatment was provided for the next patient. Patients, clinicians (including those assessing patients) and investigators analysing data were not masked to treatment allocation. Procedures Briefly, all patients received 50·4 Gy delivered in 28 daily fractions over 38 days with fluorouracil on days 1–4 and 29–32 by continuous intravenous infusion and either mitomycin as bolus on day 1 only or cisplatin by infusion on days 1 and 29.7 Patients randomly allocated to receive maintenance were given two additional courses of fluorouracil and cisplatin on days 71–74 and 92–95 after the start of chemoradiotherapy—ie, weeks 11 and 14. Before treatment, patients were staged according to the UICC 1990 staging system.12 Abdominopelvic CT scans and chest radiographs or thoracic CT scans were mandated, but not MRI or PET.

two additional courses of fluorouracil and cisplatin on days 71–74 and 92–95 after the start of chemoradiotherapy—ie, weeks 11 and 14. Before treatment, patients were staged according to the UICC 1990 staging system.12 Abdominopelvic CT scans and chest radiographs or thoracic CT scans were mandated, but not MRI or PET. There were three primary tumour assessments (figure 1), made by the patient's clinician (single clinical oncologist review). Assessment 1 (11 weeks from the start of chemoradiotherapy) was timed to allow any adverse events from radiotherapy to resolve and before patients randomly assigned to maintenance treatment started therapy. Assessment 2 was at 18 weeks from the start of chemoradiotherapy (4 weeks after completion of maintenance therapy for those receiving it) to assess the effect of maintenance therapy. Assessment 3 was at 26 weeks from the start of chemoradiotherapy in case of treatment delay and this timepoint has been used in other squamous cell cancers to allow for tumours that relapse or progress early.13 Information about examination under anaesthetic was not collected during the trial and biopsies were not routinely done unless there was a high suspicion of residual disease because of anxieties from the radiation oncologist regarding healing in an irradiated area (according to UK practice). Patients who did not have a complete clinical response (those with partial response, stable disease, or progressive disease who did not have salvage surgery before week 26) at either assessment 1 or 2 could have subsequent assessments and delay interventions at the time (to determine slow responders). Patients diagnosed with progressive disease at assessment 1 or any other time before 26 weeks could still have assessment 2 or 3, but as their complete clinical response status could be influenced by any salvage treatment received they were excluded from all analyses.

entions at the time (to determine slow responders). Patients diagnosed with progressive disease at assessment 1 or any other time before 26 weeks could still have assessment 2 or 3, but as their complete clinical response status could be influenced by any salvage treatment received they were excluded from all analyses. Response was assessed according to Response Evaluation Criteria in Solid Tumors (version 1.0).14 Digital rectal examination was done at all three assessments, with mandatory abdominopelvic CT scan and chest radiograph or whole body CT scan at assessment 3 (figure 1). Residual or recurrent disease was confirmed by biopsy before further therapy if results from other evaluations were ambiguous.

umors (version 1.0).14 Digital rectal examination was done at all three assessments, with mandatory abdominopelvic CT scan and chest radiograph or whole body CT scan at assessment 3 (figure 1). Residual or recurrent disease was confirmed by biopsy before further therapy if results from other evaluations were ambiguous. Patients were classified into two groups at each assessment: patients with a complete clinical response or patient without a complete clinical response (ie, patients with a partial response, stable disease, or disease progression). Patients who attended the assessments with insufficient response data were classified as “unknown” whereas those who either did not attend assessments or whose data were not reported were classified as “missing”. Where possible, the missing nodal status was extrapolated from the most recent previous and subsequent assessments or follow-up information. In contrast to our previous publication of ACT II trial data,7 which defined complete response according to primary disease status only, in this report we included absence of nodal disease in the definition of complete clinical response to more accurately describe a group of patients with complete disappearance of tumour.

contrast to our previous publication of ACT II trial data,7 which defined complete response according to primary disease status only, in this report we included absence of nodal disease in the definition of complete clinical response to more accurately describe a group of patients with complete disappearance of tumour. Outcomes In the main trial evaluating chemoradiotherapy the primary endpoints were complete clinical response at 26 weeks from the start of chemoradiotherapy, acute toxic effects for patients that received chemoradiotherapy and progression-free survival for patients that received maintenance chemotherapy. This post-hoc analysis investigated complete clinical response at all three assessments (11 weeks, 18 weeks, and 26 weeks from the start of chemoradiotherapy) as well as progression-free survival and overall survival measured from the time of randomisation.

survival for patients that received maintenance chemotherapy. This post-hoc analysis investigated complete clinical response at all three assessments (11 weeks, 18 weeks, and 26 weeks from the start of chemoradiotherapy) as well as progression-free survival and overall survival measured from the time of randomisation. Statistical analysis The association between tumour response and progression-free survival or overall survival at each timepoint was examined by Kaplan-Meier curves and Cox regression models. Crude and baseline adjusted hazard ratios (HRs) for achieving versus not achieving complete clinical response were calculated and adjusted for prognostic baseline factors and trial treatment. Progression-free survival events were defined as progressive disease, local recurrence (with or without metastases), metastases, or death from any cause. New tumours were not defined as progression-free survival events. Overall survival events included deaths from any cause. Time-to-event endpoints were measured from the date of randomisation, and patients without the event of interest were censored at date of last follow-up. Sensitivity analyses were done to check the effect of extrapolating nodal status when missing on the proportion of patients with a complete response.

from any cause. Time-to-event endpoints were measured from the date of randomisation, and patients without the event of interest were censored at date of last follow-up. Sensitivity analyses were done to check the effect of extrapolating nodal status when missing on the proportion of patients with a complete response. To ensure that the analyses were not biased by the inclusion of patients who died before assessment 3, progression-free survival and overall survival were analysed both in all randomised patients using complete clinical response status where known, and those patients who attended clinic at all three timepoints and did not have salvage treatment before assessment 3.

the inclusion of patients who died before assessment 3, progression-free survival and overall survival were analysed both in all randomised patients using complete clinical response status where known, and those patients who attended clinic at all three timepoints and did not have salvage treatment before assessment 3. Two sensitivity analyses of progression-free survival and overall survival was done using two extreme assumptions. The first was done on all randomised patients, and where response status was unknown it was assumed to be complete clinical response, and the second was done on all randomised patients, and where response status was unknown it was assumed to be not- complete clinical response. Analyses other than overall survival and progression-free survival were based on those patients who had tumour assessment data at all timepoints excluding those patients who had salvage treatment, to have a uniform dataset for these analyses, unaffected by missing tumour response data. All reported p values are two-sided, and analyses were done using Stata (version 12). We also examined the prognostic performance of complete clinical response status at each timepoint, using sensitivity and false-positive rates. This study is registered as at ISRCTN, number 26715889. Role of the funding source The funder had no role in study design, report writing, or collecting, analysing, or interpretation of the data. RG-J, DS-M, HMM, RB, SB, AH, and LK had full access to the data. All authors made the decision to submit the report for publication and gave final approval for submission.

Two sensitivity analyses of progression-free survival and overall survival was done using two extreme assumptions. The first was done on all randomised patients, and where response status was unknown it was assumed to be complete clinical response, and the second was done on all randomised patients, and where response status was unknown it was assumed to be not- complete clinical response. Analyses other than overall survival and progression-free survival were based on those patients who had tumour assessment data at all timepoints excluding those patients who had salvage treatment, to have a uniform dataset for these analyses, unaffected by missing tumour response data. All reported p values are two-sided, and analyses were done using Stata (version 12). We also examined the prognostic performance of complete clinical response status at each timepoint, using sensitivity and false-positive rates. This study is registered as at ISRCTN, number 26715889. Role of the funding source The funder had no role in study design, report writing, or collecting, analysing, or interpretation of the data. RG-J, DS-M, HMM, RB, SB, AH, and LK had full access to the data. All authors made the decision to submit the report for publication and gave final approval for submission. Results 940 patients were enrolled from June 4, 2001, until Dec 16, 2008. The baseline characteristics of all individuals enrolled in the trial have been reported previously.7 Overall, the median age was 58 years, 486 (52%) of 940 had a primary tumour of 5 cm or smaller (T1 or T2), 430 (46%) of 940 had a primary tumour larger than 5 cm or invasion to neighbouring organs (T3 or T4), 305 (32%) of 940 had positive lymph nodes, 787 (84%) of 940 had a tumour in the anal canal, and 132 (14%) of 940 had a tumour in the anal margin. Of the 940 patients, 249 were excluded for subgroup analysis: 241 patients did not attend all tumour assessments, or had results where it was not possible to determine response status, and eight patients had salvage treatment before the third assessment. The baseline characteristics of these 691 remaining patients who were analysed in this subgroup analysis were similar to both the entire trial population 940 patients (appendix p 1), and the 249 patients who were excluded (appendix p 2). The proportions of patients with primary tumour response data at all three assessments were similar in the two treatment groups (345 [73%] of 472 patients who received mitomycin vs 346 (74%) of 468 patients who received cisplatin; appendix p 3). Eight of 19 patients with confirmed early progressive disease at assessments 1 or 2 had a potentially curative resection before week 26 and therefore had a complete clinical response at the third assessment (week 26). Excluding these patients in the analysis had a negligible effect on the results.

latin; appendix p 3). Eight of 19 patients with confirmed early progressive disease at assessments 1 or 2 had a potentially curative resection before week 26 and therefore had a complete clinical response at the third assessment (week 26). Excluding these patients in the analysis had a negligible effect on the results. The median follow-up, censoring deaths, was 5·1 years (IQR 3·9–6·9) for all 940 patients and 5·2 years (4·0–6·8) for the 691 patients who attended all three assessments. There were 211 deaths and 292 progression-free survival events in the whole trial population. 23 of these deaths occurred before assessment 3. 12 patients died before assessment 1 (six from chemotherapy-related adverse events, two from anal cancer, three from reasons unrelated to cancer, and one from an unknown cause). Three died between assessments 1 and 2 (two by reasons unrelated to cancer and one from an unknown cause). Eight patients died between assessments 2 and 3 (six from anal cancer, one from radiotherapy, and one by suicide). There were 127 deaths and 182 progression-free survival events in the subgroup of 691 patients who had attended all three assessments. Compliance to chemoradiotherapy in the subgroup pf 691 patients who were assessable at all three timepoints was high and similar between the mitomycin and cisplatin groups (appendix p 4) and similar to that for all 940 trial patients.7 The median overall treatment time was 38 days (IQR 38–39 days) in both the mitomycin and cisplatin groups.

iotherapy in the subgroup pf 691 patients who were assessable at all three timepoints was high and similar between the mitomycin and cisplatin groups (appendix p 4) and similar to that for all 940 trial patients.7 The median overall treatment time was 38 days (IQR 38–39 days) in both the mitomycin and cisplatin groups. In our subgroup analysis of the 691 patients with response data at all three timepoints, the proportion of patients with complete clinical response increased over time: 441 (64%, 95% CI 61–67) of 691 patients had a complete clinical response at assessment 1, 556 (81%, 78–88) at assessment 2, and 590 (85%, 83–88) at assessment 3 (Table 1, Table 2). 421 (95%) of 441 patients who had a complete clinical response at assessment 1 maintained this complete clinical response at assessment 2 and 411 (93%) of 441 still had a complete clinical response at assessment 3 (table 3). The remaining 30 (7%) of 441 patients either had a suspected relapse at assessment 2 or 3 (n=25) or they were missing data for nodal status (n=5). Complete clinical response was achieved in 492 (52%) of 940 patients at assessment 1 (11 weeks), 665 (71%) of patients at assessment 2 (18 weeks), and 730 (78%) of patients at assessment 3 (26 weeks). However, 151 (72%) of 209 patients who were not in complete clinical response at assessment 1 achieved complete clinical response by assessment 3, and 115 (76%) of 151 were alive and disease-free on last follow-up after treatment. Therefore, 115 (55%) of 209 patients who did not have a complete clinical response at assessment 1 could be considered slow responders.

t in complete clinical response at assessment 1 achieved complete clinical response by assessment 3, and 115 (76%) of 151 were alive and disease-free on last follow-up after treatment. Therefore, 115 (55%) of 209 patients who did not have a complete clinical response at assessment 1 could be considered slow responders. Of the overall trial population, 119 (13%) of 940 patients did not have a complete clinical response at assessment 3 (table 3). Of these 119, two (2%) patients had defunctioning stomas for side-effects of radiotherapy, 27 (23%) had salvage surgery (abdominoperineal excision of rectum or anorectal excision), and three had other types of surgery (all done after 26 weeks). Disease was pathologically confirmed before radical surgery.

t 3 (table 3). Of these 119, two (2%) patients had defunctioning stomas for side-effects of radiotherapy, 27 (23%) had salvage surgery (abdominoperineal excision of rectum or anorectal excision), and three had other types of surgery (all done after 26 weeks). Disease was pathologically confirmed before radical surgery. The difference in the population of patients achieving a complete clinical response between patients in the cisplatin and mitomycin groups or between patients in the groups that received or did not receive maintenance was not significant at any assessment (table 1). There was no interaction between maintenance treatment and mitomycin or cisplatin (p=0·88 for both). The proportion of patients with a complete clinical response at assessment 3 (disregarding nodal status), was 311 (90%) of 345 for mitomycin and 313 (91%) of 346 for cisplatin (appendix p 3), similar to those among all 940 patients as previously reported (391 [91%] of 432 patients treated with mitomycin and 386 [90%] of 431 patients treated with cisplatin).7 When nodal status was included at assessment 3 (26 weeks), the results were similar, with 290 (84%) of 345 patients in the mitomycin group achieving a complete clinical response compared with 294 (85%) of 346 patients in the cisplatin group (appendix p 3).

tomycin and 386 [90%] of 431 patients treated with cisplatin).7 When nodal status was included at assessment 3 (26 weeks), the results were similar, with 290 (84%) of 345 patients in the mitomycin group achieving a complete clinical response compared with 294 (85%) of 346 patients in the cisplatin group (appendix p 3). Regardless of when patients were assessed, clinical complete response was affected by patient's tumour size and nodal stage (appendix p 5). Although the clinical complete response was significantly affected by tumour metastasis to neighbouring organs (compared with no tumour metastasis) at assessment 1 (p=0·0009), there was no longer a significant difference between patients with and without metastasis to neighbouring organs at assessment 3 (p=0·08; appendix p 5). Clinical complete response was unaffected by age (patients aged 65 years and older compared with patients younger than 65 years). However, clinical complete response was affected by a patient's sex at assessment 3, but not at assessment 1 (appendix p 5). Overall survival of the overall trial population was analysed (with tumour response data, where available at any of the three assessments; figure 2). Table 4 shows the crude and adjusted HRs for overall survival. The 5 year overall survival for patients with complete clinical response and patients without a complete clinical response groups were: 83% (79–86) and 72% (66–78) at assessment 1; 84% (81–87) and 59% (49–67) at assessment 2; and 87% (84–89) and 46% (37–55) at assessment 3. The overall survival for the subgroup of 691 patients who had tumour assessments at all three timepoints was also assessed (appendix p 8) and was similar to that found in the overall trial population (table 4). The 5 year overall survival in this subgroup for patients with a complete clinical response and patients without a complete clinical response group was 85% (95% CI 81–88) and 75% (68–80) at assessment 1, 86% (82–88) and 61% (50–70) at assessment 2, and 87% (84–90) and 48% (36–58) at assessment 3.

population (table 4). The 5 year overall survival in this subgroup for patients with a complete clinical response and patients without a complete clinical response group was 85% (95% CI 81–88) and 75% (68–80) at assessment 1, 86% (82–88) and 61% (50–70) at assessment 2, and 87% (84–90) and 48% (36–58) at assessment 3. Progression-free survival for the overall trial population and the subgroup of patients with a response at all timepoints was also analysed (table 4, figure 3, appendix p 9). The 5 year progression-free survival in patients who had a complete clinical response compared with those who did not have a complete clinical response was 75% versus 63% at assessments 1, 75% versus 53% at assessment 2, and 80% versus 33% at assessment 3. Analysis of the subgroup of 691 patients who had data for all three timepoints yielded similar results (appendix p 9). There was no difference in progression-free survival events in patients who had a complete clinical response at assessment 1 (105 events in 441 patients) and those who had a complete clinical response only at assessment 3 (nine events in 43 patients) (absolute difference 2·9 [95% CI −9·9 to 15·7], p=0·67).

pendix p 9). There was no difference in progression-free survival events in patients who had a complete clinical response at assessment 1 (105 events in 441 patients) and those who had a complete clinical response only at assessment 3 (nine events in 43 patients) (absolute difference 2·9 [95% CI −9·9 to 15·7], p=0·67). We did several analyses for overall survival and progression-free survival to check for consistency in the results because 241 of 940 patients had missing tumour assessments at one or more timepoints (where the complete clinical response status was unavailable: unknown or missing) and eight patients who had available data were excluded because they had salvage treatment. Sensitivity analyses on the basis of imputing data for missing tumour response (with assumptions) and those who did not attend assessments (which includes those for which response data were not reported) provided similar results for both overall survival and progression-free survival as the main analysis (table 4, appendix pp 10–13); as well as for imputation for missing nodal status, assumed to be either positive or negative (appendix p 6). Furthermore, after excluding 23 deaths occurring before assessment 3 (which would otherwise bias the group of patients with missing data), there was no difference in overall survival between patients who attended, patients who did not attend this assessment, or those with data not reported (table 3). At 1 year the overall survival was 96% (95% 95–98) for 809 patients who attended assessment 1 versus 91% (84–95) for patients who either did not attend or had data unreported at assessment 1. 1 year overall survival at assessment 2 was 97% (95–98) for the 852 patients who attended compared with 90% (80–95) for 73 patients who did not attend or had data not reported, and 1 year overall survival at assessment 3 was 98% (96–99) for the 871 patients who attended versus 84% (70–92) for the 46 patients who did not attend or data not reported at assessment 3 (data not shown).

852 patients who attended compared with 90% (80–95) for 73 patients who did not attend or had data not reported, and 1 year overall survival at assessment 3 was 98% (96–99) for the 871 patients who attended versus 84% (70–92) for the 46 patients who did not attend or data not reported at assessment 3 (data not shown). The sensitivity (ie the number of patients with a complete response who are alive relative to the total number of patient alive) of complete clinical response to predict overall survival was analysed at all three timepoints (appendix p 7). Sensitivity of complete clinical response to predict overall survival at 1 year was the highest at assessment 3 (88%) compared with at assessment 2 (85%) or assessment 1 (68%). The probability of a false-positive at 1 year was also the lowest at assessment 3 (20%) compared with assessment 2 (32%) and assessment 1 (50%). Discussion Our results show that the proportion of those with a complete clinical response at 26 weeks (assessment 3) is more informative than either of the two earlier assessments, and that it is acceptable to monitor partial responders carefully up to the 26 week assessment. This extended period with careful monitoring has not been international practice to date, and there are patients who have salvage surgery because of residual tumour found shortly after completing chemoradiotherapy. Therefore it is important for patients and clinicians to establish the best time to assess tumour response, and we have done so with data from our clinical trial.

international practice to date, and there are patients who have salvage surgery because of residual tumour found shortly after completing chemoradiotherapy. Therefore it is important for patients and clinicians to establish the best time to assess tumour response, and we have done so with data from our clinical trial. There was no evidence that maintenance therapy acted as a confounding factor for the association between complete clinical response status and overall survival and progression-free survival, as maintenance therapy did not influence whether a patient had complete clinical response. Moreover, maintenance treatment started only after assessment 1, there was no difference in the complete clinical response rates between patients with and without maintenance therapy (table 1), and there was no effect of maintenance therapy on either progression-free survival or overall survival, as detailed in our original report on this trial.7 Since the median overall treatment time was the same in both the mitomycin and cisplatin groups (38 days [IQR 38–39 days]) the timing of assessment of complete clinical response was probably not confounded by variations in treatment duration. There was a substantial increase in the proportion of patients with a complete clinical response at 26 weeks (assessment 3) from the start of chemoradiotherapy compared with earlier assessments (table 1). These data are compatible with the RTOG-8704 trial results,1 which mandated a biopsy after chemoradiotherapy at 4–6 weeks after completion of chemotherapy as part of the assessment and showed that the combination of mitomycin and fluorouracil with radiation produced a pathological complete response in 92% of patients at 6 weeks after completing chemoradiotherapy.1 Patients with residual cancer shown in the biopsy after chemoradiotherapy were treated with a salvage regimen of additional pelvic radiotherapy (9 Gy), fluorouracil, and cisplatin (100 mg/m2). Of the 24 assessable patients who had salvage chemoradiotherapy, 12 (50%) were rendered disease-free. Our results imply that simply waiting longer might have achieved similar results.

chemoradiotherapy were treated with a salvage regimen of additional pelvic radiotherapy (9 Gy), fluorouracil, and cisplatin (100 mg/m2). Of the 24 assessable patients who had salvage chemoradiotherapy, 12 (50%) were rendered disease-free. Our results imply that simply waiting longer might have achieved similar results. Our findings are also consistent with those previously described in a series of sequential chemoradiotherapy studies with mitomycin,15 which showed that some patients with squamous cell carcinoma of the anus required 9–12 months to achieve a complete clinical response (defined as complete resolution of all clinical signs of the primary cancer for a duration of 2 months; if not, the tumour was scored as residual). A limitation of our analysis is that the group which achieved a complete clinical response at assessment 3 (week 26) does not truly represent all responders to chemoradiotherapy because it also includes patients who have had an early and sustained response and those who have a late response, but not patients who had a complete clinical response at assessment 1 but subsequently relapsed or did not have a complete clinical response by assessment 3.

y represent all responders to chemoradiotherapy because it also includes patients who have had an early and sustained response and those who have a late response, but not patients who had a complete clinical response at assessment 1 but subsequently relapsed or did not have a complete clinical response by assessment 3. Eventual outcomes (overall survival and progression-free survival) were independent of the timing that complete clinical response was achieved. These outcomes seem to be independent of whether complete clinical response was achieved at assessment 3 or at assessment 1. The HRs from both the overall survival and progression-free survival curves suggest that assessment 3 (26 weeks from the start of treatment) gives the most widely discriminating effect on survival outcomes, and suggests this is the optimum timepoint for assessment. Tumour assessment (including nodal status) at week 26 should therefore be explored as a surrogate endpoint for progression-free survival and overall survival in future trials.

start of treatment) gives the most widely discriminating effect on survival outcomes, and suggests this is the optimum timepoint for assessment. Tumour assessment (including nodal status) at week 26 should therefore be explored as a surrogate endpoint for progression-free survival and overall survival in future trials. These findings challenge guidelines8, 9 for the post-treatment follow-up of patients with anal cancer, which include serial digital rectal examination, with biopsy of clinically suspicious lesions recommended at weeks 4–6, 6–8, and 8–12 after chemoradiotherapy completion, respectively. In our trial, not all 940 randomised patients had a known complete clinical response status at all three assessment timepoints. To address whether substantial bias could have arisen we made extreme assumptions about unknown status in two sensitivity analyses, and both produced the same conclusions irrespective of whether we considered the whole trial population or only those who attended all three assessments—ie, that the strongest association between complete clinical response status and either overall survival or progression-free survival is at assessment 3.

ity analyses, and both produced the same conclusions irrespective of whether we considered the whole trial population or only those who attended all three assessments—ie, that the strongest association between complete clinical response status and either overall survival or progression-free survival is at assessment 3. Anal cancers can regress slowly after completion of chemoradiotherapy treatment. Accurate identification of response is a crucial component to optimise patient management. Clinical groups in the USA have recommended salvage surgery should only be considered after at least 12 weeks because complete resolution might take 3–6 months.1 The ideal method and timing of response evaluation and when maximum response occurs after chemoradiotherapy is unclear. Retrospective reports suggested initial early clinical response is an independent prognostic factor for survival.16, 17, 18 Randomised studies have done clinical assessments of the tumour between 6 weeks and 12 weeks after completion of chemoradiotherapy.1, 2, 3, 4, 5, 6

ponse occurs after chemoradiotherapy is unclear. Retrospective reports suggested initial early clinical response is an independent prognostic factor for survival.16, 17, 18 Randomised studies have done clinical assessments of the tumour between 6 weeks and 12 weeks after completion of chemoradiotherapy.1, 2, 3, 4, 5, 6 The mainstay of clinical evaluation has relied on digital rectal examination and careful examination of the inguinal regions. The limitations of this approach include the subjectivity of the clinical examination and the absence of treatment response information for any more deeply sited, non-palpable disease, including the pelvic lymph nodes. Also, radiotherapy-induced acute local effects can cause problems in our interpretation of response. Severe skin reactions, oedema, residual fibrosis, or scar tissue can be difficult to distinguish from persistent, active, or recurrent disease.19, 20 The risk of radionecrosis should also be kept in mind when considering biopsy. A significant correlation has also been observed between high-grade acute toxicity in terms of skin reactions, cystitis, proctitis, or enteritis (National Cancer Institute Common Toxicity Criteria grade 3 or worse) during chemoradiotherapy and overall survival, locoregional control, and stoma-free survival.21

g biopsy. A significant correlation has also been observed between high-grade acute toxicity in terms of skin reactions, cystitis, proctitis, or enteritis (National Cancer Institute Common Toxicity Criteria grade 3 or worse) during chemoradiotherapy and overall survival, locoregional control, and stoma-free survival.21 Imaging in terms of endoanal ultrasound or MRI has an established role in initial locoregional staging of the primary and pelvic lymph nodes,8, 10 but their ability to assess response accurately is less welldefined.22 It can take 16–20 weeks to allow sufficient time for resolution of oedema to enable accurate classification of treatment response by ultrasound.23 MRI complements clinical assessment but can assign a worse stage prognosis to a patient, particularly lymph node status because enlarged lymph nodes can be recorded as node positive although many do not contain tumours. Metabolic and functional imaging techniques, such as diffusion-weighted MRI,24 or PET with 18F-fluorodeoxyglucose,25, 26, 27 might offer an alternative option to assess response at an early timepoint without the potential morbidity of a biopsy, but large prospective studies are required.

gh many do not contain tumours. Metabolic and functional imaging techniques, such as diffusion-weighted MRI,24 or PET with 18F-fluorodeoxyglucose,25, 26, 27 might offer an alternative option to assess response at an early timepoint without the potential morbidity of a biopsy, but large prospective studies are required. The reported 5 year overall survival with salvage abdominoperineal resection for persistent or locally recurrent anal cancer was only 24–64%;28, 29 hence, the rationale for proactive follow-up. For surgeons, there is always a tension between doing an early biopsy to define an early recurrence with limited locoregional disease, when surgical salvage is likely to be effective, and the risk of provoking necrosis from a biopsy after chemoradiotherapy. 209 (30%) of 691 patients were not in complete clinical response 4 weeks after chemoradiotherapy (assessment 1); however, 115 remained disease-free at later follow-up, hence, early surgical salvage would not have been appropriate for these patients. The RTOG-8704 trial mandated biopsy of the primary tumour at 4–6 weeks after a dose of 45 Gy to confirm complete pathological response in clinically residual disease,1 but routine biopsy is controversial in the monitoring of response to treatment, with some clinicians supporting random biopsies at a 3 month interval, while others support a biopsy only when there is a suspicious lesion.30

after a dose of 45 Gy to confirm complete pathological response in clinically residual disease,1 but routine biopsy is controversial in the monitoring of response to treatment, with some clinicians supporting random biopsies at a 3 month interval, while others support a biopsy only when there is a suspicious lesion.30 Our results suggest that early response assessments might not be reliable by failing to account for patients who are slow to respond to treatment. The data confirm that partial regression can be managed by close follow-up to confirm eventual complete regression. Others have recommended that patients deemed to be slow to respond (persistence of local disease with regression <50%) should be evaluated by at least two physicians at the end of radiotherapy, patients should be given an additional one or two courses of chemotherapy courses.31 However, our data previously showed no significant effect from an additional two cycles of maintenance chemotherapy (fluorouracil and cisplatin) after chemoradiotherapy (complete clinical response 82% vs 85% p=0·34).7 The ACT II chemoradiotherapy schedule achieved an excellent proportion of patients with an early complete clinical response, but our results show 151 (72%) of 209 patients not in complete clinical response at 4 weeks after chemoradiotherapy achieved complete clinical response at 26 weeks. The consistency of the overall treatment time (median 38 days) makes the validity of our data for timing of complete clinical response assessment even stronger.

ur results show 151 (72%) of 209 patients not in complete clinical response at 4 weeks after chemoradiotherapy achieved complete clinical response at 26 weeks. The consistency of the overall treatment time (median 38 days) makes the validity of our data for timing of complete clinical response assessment even stronger. Although we would advise careful monitoring from completion of treatment to facilitate timely surgical salvage therapy for progressive disease, it seems safe to observe a resolving tumour up to 26 weeks after the start of chemoradiotherapy, and some patients could thus avoid unnecessary surgery. It might even be safe to extend evaluation beyond this timepoint, as some studies suggest a few patients might require more than 10 months for complete regression, but prospective data are required to confirm this timeline. We propose guidelines should be revised, and that the assessment of response at 26 weeks be used in future trials and explored as a surrogate endpoint for overall survival and progression-free survival. Supplementary Material Supplementary appendix Acknowledgments Cancer Research UK provided funding and UCL sponsored the trial. The ACT II Trial Management Group comprised all authors plus Roger D James and Jonathan Ledermann. We thank all those who participated in this trial, clinicians and their staff, and patients.

Although we would advise careful monitoring from completion of treatment to facilitate timely surgical salvage therapy for progressive disease, it seems safe to observe a resolving tumour up to 26 weeks after the start of chemoradiotherapy, and some patients could thus avoid unnecessary surgery. It might even be safe to extend evaluation beyond this timepoint, as some studies suggest a few patients might require more than 10 months for complete regression, but prospective data are required to confirm this timeline. We propose guidelines should be revised, and that the assessment of response at 26 weeks be used in future trials and explored as a surrogate endpoint for overall survival and progression-free survival. Supplementary Material Supplementary appendix Acknowledgments Cancer Research UK provided funding and UCL sponsored the trial. The ACT II Trial Management Group comprised all authors plus Roger D James and Jonathan Ledermann. We thank all those who participated in this trial, clinicians and their staff, and patients. Contributors RG-J, DS-M, DC, LK, AH, HMM contributed to the design of this substudy, the figures, data collection, data interpretation, and writing. FA, KB, RJH, and JS collected data. LK and AH did or supervised analysis. SB and RB collected data and did data interpretation. All drafts were overseen by the corresponding author with input from all coauthors. All authors reviewed iterations of the manuscript and gave final approval.

data interpretation, and writing. FA, KB, RJH, and JS collected data. LK and AH did or supervised analysis. SB and RB collected data and did data interpretation. All drafts were overseen by the corresponding author with input from all coauthors. All authors reviewed iterations of the manuscript and gave final approval. Declaration of interests DC has received research funding from Amgen, AstraZeneca, Bayer, Celgene, Merrimack Medimmune, Merck, and Sanofi and RG-J has received grant funding and personal fees from Merck and Roche, and personal fees from Eisai, Amgen, and Servier. All other authors declare no competing interests. Figure 1 Treatment and assessment schedule in ACT II *Patients referred for surgical salvage as appropriate. Figure 2 Overall survival according to response at assessments 1, 2, and 3 among all 940 patients in whom response data were known cCR=complete clinical response. HR=hazard ratio. Figure 3 Progression-free survival according to response at assessments 1, 2, and 3 among all 940 patients in whom response data were known cCR=complete clinical response. HR=hazard ratio. Table 1 Complete clinical response at all three assessments in patients with primary tumour response data at all three assessments

cCR=complete clinical response. HR=hazard ratio. Figure 3 Progression-free survival according to response at assessments 1, 2, and 3 among all 940 patients in whom response data were known cCR=complete clinical response. HR=hazard ratio. Table 1 Complete clinical response at all three assessments in patients with primary tumour response data at all three assessments Overall (n=691) Mitomycin (n=345) Cisplatin (n=346) χ2 No maintenance (n=347)* Maintenance (n=305) χ2 Assessment 1 441 (64%); (61–67) 231 (67%); (62–72) 210 (61%); (56–66) p=0·09 224 (65%); (60–70) 187 (61%); (56–67) p=0·39† Assessment 2 556 (80%); (78–88) 273 (79%); (75–83) 283 (82%); (78–86) p=0·38 274 (79%); (75–83) 252 (83%); (78–87) p=0·24 Assessment 3 590 (85%); (83–88) 292 (85%); (81–88) 298 (86%); (82–90) p=0·58 294 (85%); (81–89) 264 (87%); (83–90) p=0·51 Data are shown as n (%); (95% CI), excluding eight patients who had salvage surgery. * 39 of 691 patients did not get randomly assigned to the maintenance therapy (at physicians' discretion), so they were not included in the maintenance analysis. The total number analysed for maintenance comparison is, therefore, 652 and not 691. † No difference in the proportion of patients with a complete response is expected between patients with and without maintenance therapy at assessment 1 because maintenance treatment would only start after this time. The p values shown were calculated with χ2 tests. Table 2 Distribution of patients and tumour response for patients who attended all three assessments (n=691)

† No difference in the proportion of patients with a complete response is expected between patients with and without maintenance therapy at assessment 1 because maintenance treatment would only start after this time. The p values shown were calculated with χ2 tests. Table 2 Distribution of patients and tumour response for patients who attended all three assessments (n=691) Patients with complete clinical response Patients without complete clinical response Patients with unknown response data* Assessment 1 441 209 41 Assessment 2 556 106 29 Assessment 3† 590 88 13 * Patients classified as “unknown” attended the assessment but had response data that were inconclusive. † 23 patients died before assessment 3. Some patients did not attend for more than one assessment or had missing response data for more than one assessment so it is not possible to sum these numbers over all three timepoints. Table 3 Distribution of patients and tumour response for all patients in the trial (n=940) Patients with complete clinical response Patients without complete clinical response Patients with unknown response data* Patients with missing data† Assessment 1 492 235 82 131 Assessment 2 665 137 50 88 Assessment 3‡ 730 119 22 69 * Patients classified as “unknown” attended the assessment but had response data that were inconclusive. † Patients classified as “missing” included those for whom response data were not reported and patients who did not attend clinic for assessment.

Patients with complete clinical response Patients without complete clinical response Patients with unknown response data* Patients with missing data† Assessment 1 492 235 82 131 Assessment 2 665 137 50 88 Assessment 3‡ 730 119 22 69 * Patients classified as “unknown” attended the assessment but had response data that were inconclusive. † Patients classified as “missing” included those for whom response data were not reported and patients who did not attend clinic for assessment. ‡ 23 patients died before assessment 3. Some patients did not attend for more than one assessment or had missing response data for more than one assessment so it is not possible to sum these numbers over all three timepoints. Table 4 Association between overall survival or progression-free survival and tumour response at three different assessment timepoints

‡ 23 patients died before assessment 3. Some patients did not attend for more than one assessment or had missing response data for more than one assessment so it is not possible to sum these numbers over all three timepoints. Table 4 Association between overall survival or progression-free survival and tumour response at three different assessment timepoints Overall survival Progression-free survival Crude Adjusted* Crude Adjusted* All 940 patients (using response data wherever available) 1 0·56 (0·40–0·77); p<0·0005 0·77 (0·50–1·18); p=0·22 0·59 (0·45–0·78); p<0·002 0·66 (0·46–0·95); p=0·02 2 0·30 (0·22–0·41); p<0·0001 0·40 (0·26–0·61); p<0·0001 0·37 (0·28–0·49); p<0·0001 0·43 (0·29–0·62); p<0·0001 3 0·17 (0·12–0·23); p<0·0001 0·22 (0·14–0·35); p<0·0001 0·16 (0·12–0·21); p<0·0001 0·15 (0·10–0·21); p<0·0001 691 patients with response data at all three timepoints† 1 0·55 (0·39–0·79); p=0·001 0·81 (0·51–1·29); p=0·38 0·61 (0·45–0·82); p=0·001 0·68 (0·46–0·99); p=0·05 2 0·30 (0·20–0·43); p<0·0001 0·44 (0·26–0·72); p=0·001 0·36 (0·26–0·50); p<0·0001 0·44 (0·29–0·66); p<0·0001 3 0·17 (0·12–0·24); p<0·0001 0·24 (0·14–0·40); p<0·0001 0·15 (0·11–0·21); p<0·0001 0·16 (0·10–0·24); p<0·0001 All 940 patients (missing response data assumed to be complete clinical response) 1 0·72 (0·54–0·96); p=0·028 0·98 (0·66–1·46); p=0·93 0·71 (0·55–0·92); p=0·01 0·79 (0·56–1·11); p=0·17 2 0·38 (0·28–0·52); p<0·0001 0·46 (0·30–0·70); p<0·0001 0·44 (0·33–0·58); p<0·001 0·46 (0·32–0·66); p<0·0001 3 0·25 (0·19–0·34); p<0·0001 0·31 (0·20–0·47); p<0·0001 0·21 (0·16–0·27); p<0·001 0·19 (0·13–0·27); p<0·0001 All 940 patients (missing response data assumed to be not complete clinical response) 1 0·51 (0·39–0·68); p<0·0001 0·68 (0·47–0·97); p=0·04 0·57 (0·45,0·72); p<0·0001 0·65 (0·48–0·89); p=0·01 2 0·31 (0·24–0·41); p<0·0001 0·40 (0·28–0·59); p<0·0001 0·39 (0·31,0·49); p<0·0001 0·49 (0·35–0·68); p<0·0001 3 0·16 (0·12–0·20); p<0·0001 0·19 (0·13–0·27); p<0·0001 0·16 (0·13–0·21); p<0·0001 0·17 (0·12–0·24); p<0·0001 Data are HR (95% cCI); p value. Complete clinical response is the complete disappearance of disease in the primary and nodes.

001 0·40 (0·28–0·59); p<0·0001 0·39 (0·31,0·49); p<0·0001 0·49 (0·35–0·68); p<0·0001 3 0·16 (0·12–0·20); p<0·0001 0·19 (0·13–0·27); p<0·0001 0·16 (0·13–0·21); p<0·0001 0·17 (0·12–0·24); p<0·0001 Data are HR (95% cCI); p value. Complete clinical response is the complete disappearance of disease in the primary and nodes. * Adjusted for potential confounding factors: age, sex, site of primary, tumour differentiation, histology, baseline white blood cell count, baseline platelets, baseline haemoglobin, tumour size, nodal stage, and trial treatment. † Excluding eight patients who had salvage surgery.

Introduction An important indicator of the public health impact of the National Health Service Breast Screening Programme (NHSBSP) in the UK is the participation rate, defined as the percentage of women invited for screening who are screened adequately within 180 days of invitation (usually referred to as uptake in official reports). In England, participation following routine invitation fell from 74·4% in 2004–05 to 71·3% in 2014–15,1 and we are seeing a decline for the fourth consecutive year in a row, approaching the national minimum standard of 70%. In particular, participation among women invited for their prevalent (first) round of screening has decreased by an even greater amount (from 70·1% in 2004–05 to 63·3% in 2014–15).1 Participation in breast cancer screening also tends to be lower in areas of socioeconomic deprivation than in wealthier areas.2, 3

of 70%. In particular, participation among women invited for their prevalent (first) round of screening has decreased by an even greater amount (from 70·1% in 2004–05 to 63·3% in 2014–15).1 Participation in breast cancer screening also tends to be lower in areas of socioeconomic deprivation than in wealthier areas.2, 3 The NHSBSP invites women aged 50–70 years to mammographic screening every 3 years. The invitation letter includes a screening appointment with a given date, time, and place. An age eligibility extension to invite women aged between 47 years and 73 years is currently being trialled.4 The usual practice for non-attenders of the first offered appointment is to send them a second invitation letter, which can vary: some centres supply open invitations, asking women to telephone to make an alternative appointment; whereas others routinely offer second timed appointments, with date, time, and place stipulated. The Department of Health has advised NHS England that the approach with second timed appointments should be used.5 Second timed appointments are NHSBSP policy, although this approach is not universally followed. Research in context Evidence before this study

The NHSBSP invites women aged 50–70 years to mammographic screening every 3 years. The invitation letter includes a screening appointment with a given date, time, and place. An age eligibility extension to invite women aged between 47 years and 73 years is currently being trialled.4 The usual practice for non-attenders of the first offered appointment is to send them a second invitation letter, which can vary: some centres supply open invitations, asking women to telephone to make an alternative appointment; whereas others routinely offer second timed appointments, with date, time, and place stipulated. The Department of Health has advised NHS England that the approach with second timed appointments should be used.5 Second timed appointments are NHSBSP policy, although this approach is not universally followed. Research in context Evidence before this study We searched the PubMed database with the keywords “breast cancer”, “breast screening”, “appointment”, and “non-attenders” for articles in English published between Jan 1, 1990, and Dec 31, 2016. This search retrieved eight papers, of which six were not considered relevant after title or abstract review. In 1998, Stead and colleagues compared second appointments with fixed date and time versus open second invitations for non-attenders of breast cancer screening in a randomised controlled trial in one breast screening centre in England, finding an increased participation rate with second timed appointments. Although the quality of the trial was good, its findings might not apply to current practice and target populations for screening since these have changed in terms of age and might have changed in terms of social support and employment status. In 2016, Hudson and colleagues reported the results of an observational study in north London comparing timed and non-timed second appointments, showing increased participation with timed appointments. We identified no trials of second timed appointments for non-attenders from outside the UK breast screening programme.

In 2016, Hudson and colleagues reported the results of an observational study in north London comparing timed and non-timed second appointments, showing increased participation with timed appointments. We identified no trials of second timed appointments for non-attenders from outside the UK breast screening programme. Added value of this study Our randomised controlled trial assessed the effects of sending invitations for breast cancer screening with a second timed appointment to women who did not attend their first offered appointment. Unlike previous studies on this subject, the trial was done at a national level (six centres across England) in 2014–15. We could therefore analyse the efficacy of the intervention depending on a woman's location and level of socioeconomic deprivation since the sites in the study covered a wide range of socioeconomic status levels. Implications of all the available evidence Our findings show the positive effects of second timed appointments on attendance for breast cancer screenings. The results are of policy interest for early detection of breast cancer, because a simple change in the procedure of addressing non-attenders of breast cancer screening invitations could result in more women being screened.

ow the positive effects of second timed appointments on attendance for breast cancer screenings. The results are of policy interest for early detection of breast cancer, because a simple change in the procedure of addressing non-attenders of breast cancer screening invitations could result in more women being screened. Although some women might not attend their screening appointment because they have made an informed choice not to do so, some of them will not attend for other reasons. These women might find a second timed appointment more beneficial than an open invitation because it does not require any effort to book a new appointment with the screening centre. Previous findings suggest that participation is greater when a second timed appointment is given to non-attenders,6, 7 but further investigations are needed to identify the women who would be most and least likely to respond to the second invitation. For example, someone who has not attended their last three screening appointments might not attend whatever the form of the second invitation.

ed appointment is given to non-attenders,6, 7 but further investigations are needed to identify the women who would be most and least likely to respond to the second invitation. For example, someone who has not attended their last three screening appointments might not attend whatever the form of the second invitation. In a randomised trial published in 1998, Stead and colleagues6 found that the effect of second timed appointments declined with increasing time since last screen, and in a more recent observational study, Hudson and colleagues7 noted the same association, at least in absolute terms, in north London. The efficacy of this approach has not been investigated in a randomised trial, or quantified with precision, in the current target population for screening. Therefore, we did a randomised trial of second timed appointments versus open invitations for non-attenders within the NHSBSP, powered to obtain significant results within subgroups of time since last screen.

vestigated in a randomised trial, or quantified with precision, in the current target population for screening. Therefore, we did a randomised trial of second timed appointments versus open invitations for non-attenders within the NHSBSP, powered to obtain significant results within subgroups of time since last screen. Methods Study design and participants This open, two-arm, randomised controlled trial was done in six screening sites in England (Derby, Hull, Plymouth, Sheffield, southeast London, and west London) for different time lengths between June 2, 2014, and Sept 30, 2015. We chose these sites because we already had links in terms of research on and evaluation of the screening programme. Our prior constraints were that we wanted substantial numbers both within and outside London and areas of varying socioeconomic status. We excluded three screening centres where we were already conducting another trial of pre-appointment reminders. The protocol is available in the appendix (pp 4–20).

reening programme. Our prior constraints were that we wanted substantial numbers both within and outside London and areas of varying socioeconomic status. We excluded three screening centres where we were already conducting another trial of pre-appointment reminders. The protocol is available in the appendix (pp 4–20). Women were invited to breast screening in batches of varying size but typically several hundreds. The date a batch was set up (ie, the list of women to be invited was compiled), was defined as the date the screening episode was opened for each woman in that list. For an individual woman, her screening episode is closed when she attends for screening or after 180 days if she does not attend. Batches of women invited to routine breast cancer screening were randomly assigned (1:1) to be sent either a second appointment with a fixed date and time (intervention) or an open invitation (control) in the event of non-attendance at the first offered appointment. After randomisation, the analysis was restricted to women aged 50–70 years. Women who self-referred for screening, women on an early recall protocol, and women who were invited because of a high risk of breast cancer were not randomised. Some women had more than one recorded invitation in the study period, which resulted in multiple records in the dataset. For these women, the first invitation date was used for reference and participation was based on the first attendance date (if any)—ie, only one of the multiple records was kept. Other exclusions that were judged to be necessary were women invited to screening outside the study period of the screening sites; observations with previous screening appointment more recent than the first offered appointment or date of the screening episode being opened; and observations with a date on the invitation letter for a previous round of screening more recent than the date on the current invitation letter. We also excluded women who participated but had missing dates of attendance, because in those cases it was not possible to determine whether they had attended within 90 days of their first offered appointment or within 180 days of their episode being opened (or neither).

than the date on the current invitation letter. We also excluded women who participated but had missing dates of attendance, because in those cases it was not possible to determine whether they had attended within 90 days of their first offered appointment or within 180 days of their episode being opened (or neither). Women were not informed of the study or asked to give consent for three reasons. First, the intervention was a minor variation in invitation practice, which was already standard procedure in some areas of England. Second, previous notification of a possible variation to the second letter of invitation might change the women's behaviour and defeat the purpose of the study. Finally, limiting the study to those interested in participating would render the results non-generalisable. The study was approved by the London Bloomsbury Research Ethics Committee.

ation of a possible variation to the second letter of invitation might change the women's behaviour and defeat the purpose of the study. Finally, limiting the study to those interested in participating would render the results non-generalisable. The study was approved by the London Bloomsbury Research Ethics Committee. Randomisation and masking Within each screening centre, every invited woman was allocated a unique number, known as the SX number. At the beginning of the study, a coin toss by the chief investigator (SWD) decided that women with an odd SX number would be allocated to the intervention group (second timed appointment), whereas women with an even SX number would be allocated to the control group (open second invitation). This is a pseudorandomisation approach. Thus, there was no masking; however, the endpoint required no subjective judgment. Women who received the wrong intervention for their group assignment (eg, women randomly assigned to the control group who received a second timed appointment letter by administrative error) were marked with an error code but were still included in analysis. Procedures Women who did not attend their first offered appointment were flagged as such by staff at that clinic on the National Breast Screening System (NBSS). Subsequently, administration staff identified the non-attenders in NBSS who were sent a second invitation within 2 weeks of the non-attended first appointment.

Procedures Women who did not attend their first offered appointment were flagged as such by staff at that clinic on the National Breast Screening System (NBSS). Subsequently, administration staff identified the non-attenders in NBSS who were sent a second invitation within 2 weeks of the non-attended first appointment. The intervention consisted of an invitation to a second appointment with fixed date and time. The control group received a second invitation that consisted of a letter with a telephone number that the women should call to rebook the missed screening appointment. Differences between the two letters were kept to a minimum so that women receiving second timed appointment letters did not feel pressurised into attending their screening appointment if they had made the decision not to participate. Second timed appointments had to be allocated to non-attenders within 90 days of the missed appointment. Because many practices already used second timed appointments for non-attenders of breast cancer screening, this study merely represented a minor variation in routine practice. Data were pseudonymised, removing identifying data items such as month and day of birth and postcode, before being sent to the Centre for Cancer Prevention (Wolfson Institute of Preventive Medicine, Queen Mary University of London), where analyses were done.

udy merely represented a minor variation in routine practice. Data were pseudonymised, removing identifying data items such as month and day of birth and postcode, before being sent to the Centre for Cancer Prevention (Wolfson Institute of Preventive Medicine, Queen Mary University of London), where analyses were done. Outcomes The primary endpoint was participation (ie, attendance) within 90 days of the first offered appointment. The key secondary endpoint was participation within 180 days of the screening episode being opened, used formally in the programme as a measure for calculating participation rates (usually referred to as uptake in reports). The endpoint was determined locally and objectively as whether or not the invitee attended for screening. Both endpoints were assessed once data collection was completed on March 1, 2016. Other secondary endpoints were subgroup analyses by prevalent (first) or incident (subsequent) screen status, by time since last attended screen, and by index of multiple deprivation and age. An economic analysis is also planned, but as a separate exercise, since this analysis will be a major analytic effort.

1, 2016. Other secondary endpoints were subgroup analyses by prevalent (first) or incident (subsequent) screen status, by time since last attended screen, and by index of multiple deprivation and age. An economic analysis is also planned, but as a separate exercise, since this analysis will be a major analytic effort. Statistical analysis On the assumption that 40% of women who received their first invitation letter would not attend their screening appointment, we required 90% power for a difference of 20% (intervention group) versus 15% (control group) of those re-invited participating within 90 days. We also assumed that 20% of invitees would be non-attenders at their last routine screening; that 15% would not have attended for three screening episodes or more; and that 10% would not have attended for four episodes or more. These proportions were approximations derived from Offman and colleagues' study8 and from the 10% never-attenders observed in the West Midlands screening histories project.9 We required 80% power in all the subgroups for a difference of 14% (intervention) versus 10% (control) in non-attenders at their last routine screening, of 10% versus 7% in those who had not attended for three screening episodes or more, and of 2% versus 1% in those who had not attended for four screening episodes or more. These proportions corresponded to requirements of 1252, 1085, 1422, and 2515 individuals per group in each of the four categories, respectively. Thus, we required at least 10 060 non-attenders per group in total (corresponding to approximately 50 300 women invited to first screening appointment in the two groups). We asked participating centres to recruit substantially more women than this number as a failsafe measure.

each of the four categories, respectively. Thus, we required at least 10 060 non-attenders per group in total (corresponding to approximately 50 300 women invited to first screening appointment in the two groups). We asked participating centres to recruit substantially more women than this number as a failsafe measure. The difference in participation between the two groups was compared with Poisson regression for the primary and key secondary endpoints, offset by the total numbers of invitees. This analysis yielded relative risks (RRs) and 95% CIs for participation, and likelihood ratio tests for significance. We also did prespecified subgroup analyses by prevalent or incident screen status, by time since last attended screen, and by the 2010 Index of Multiple Deprivation (IMD, based on a woman's postcode).10 National quintiles of IMD were used. Formal tests for heterogeneity of the effect of the intervention by prevalent or incident status or by socioeconomic status on attendance were also done. Primary analysis and subgroup analyses were by intention to treat, so that women who received the wrong type of letter for their trial group were retained in the analysis as if they had received the correct letter. Other women who were potentially excludable but were kept in the intention-to-treat analysis were those who were being screened at the time of extraction of the dataset, who were permanently or temporarily under care, who died, who moved away, who were not known at their recorded address, who attended for screening but for some reason were not screened, who had been recently screened, or whose reason for attendance or non-attendance was missing or coded as “other”. These potentially excludable women were more likely to be identified in the intervention group than in the control group, since the offer of a timed appointment was more likely to prompt the invitee to inform the service that she had moved away or had already been recently screened. We also estimated the effect of the intervention in each site separately, as a post-hoc exploratory analysis. All analyses were done with Stata/IC version 13.1. This trial is registered with Barts Health, number 009304QM.

t the invitee to inform the service that she had moved away or had already been recently screened. We also estimated the effect of the intervention in each site separately, as a post-hoc exploratory analysis. All analyses were done with Stata/IC version 13.1. This trial is registered with Barts Health, number 009304QM. Role of the funding source The Department of Health Policy Research Programme was given the opportunity to comment on this report before submission for publication. One author (JP) was employed by the funding source but only before the results became available to the authors. The sponsors and funders of the study had no role in data collection, data analysis, data interpretation, study design, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

ecame available to the authors. The sponsors and funders of the study had no role in data collection, data analysis, data interpretation, study design, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results The dataset received from the six screening centres had 33 146 records of women invited for screening at different times between June 2, 2014, and Sept 30, 2015, who did not attend their first appointment (see appendix p 3 for details of recruitment by centre). Women were followed up for more than 180 days after their episode in the current round of screening was opened, and the trial was ended when we estimated that the number of women recruited was larger than the one required by our power calculations. 7092 (21%) of the 33 146 records were excluded after randomisation because of the reasons stated in the Methods—eg, ineligible age or date of invitation, missing attendance record, or multiple records for the same woman (3520 in the intervention group vs 3572 in the control group; figure). Of the remaining 26 054 women, 12 807 (49%) had been randomly assigned to receive the intervention of a second timed appointment letter, and 13 247 (51%) to receive an open invitation letter. Characteristics of women included in the analysis were similar between groups (table 1). Most women in both groups were younger than 60 years of age and came from more deprived socioeconomic areas (IMD quintiles 1 and 2; table 1).Figure Trial profile

ntment letter, and 13 247 (51%) to receive an open invitation letter. Characteristics of women included in the analysis were similar between groups (table 1). Most women in both groups were younger than 60 years of age and came from more deprived socioeconomic areas (IMD quintiles 1 and 2; table 1).Figure Trial profile SX=a sequential unique identifier of each woman within the NHS Breast Screening Programme. *Some records had more than one reason for exclusion. Table 1 Characteristics of study groups Intervention group (n=12 807) Control group (n=13 247) Screen status Prevalent 6300 (49%) 6517 (49%) Incident 6507 (51%) 6730 (51%) Age (years) 50–59 7484 (58%) 7842 (59%) 60–70 5323 (42%) 5405 (41%) Median (IQR) 58 (53–64) 58 (53–63) IMD quintile* 1 3395 (27%) 3623 (27%) 2 3645 (28%) 3703 (28%) 3 2864 (22%) 2978 (22%) 4 1946 (15%) 2028 (15%) 5 939 (7%) 891 (7%) Missing 18 (<1%) 24 (<1%) Site Derby 2398 (19%) 2355 (18%) Hull 2120 (17%) 2233 (17%) Plymouth 2456 (19%) 2611 (20%) Southeast London 1595 (12%) 1638 (12%) Sheffield 1207 (9%) 1185 (9%) West London 3031 (24%) 3225 (24%) Data are n (%) unless otherwise stated. * 2010 Index of Multiple Deprivation (IMD) from most deprived to most affluent.

Intervention group (n=12 807) Control group (n=13 247) Screen status Prevalent 6300 (49%) 6517 (49%) Incident 6507 (51%) 6730 (51%) Age (years) 50–59 7484 (58%) 7842 (59%) 60–70 5323 (42%) 5405 (41%) Median (IQR) 58 (53–64) 58 (53–63) IMD quintile* 1 3395 (27%) 3623 (27%) 2 3645 (28%) 3703 (28%) 3 2864 (22%) 2978 (22%) 4 1946 (15%) 2028 (15%) 5 939 (7%) 891 (7%) Missing 18 (<1%) 24 (<1%) Site Derby 2398 (19%) 2355 (18%) Hull 2120 (17%) 2233 (17%) Plymouth 2456 (19%) 2611 (20%) Southeast London 1595 (12%) 1638 (12%) Sheffield 1207 (9%) 1185 (9%) West London 3031 (24%) 3225 (24%) Data are n (%) unless otherwise stated. * 2010 Index of Multiple Deprivation (IMD) from most deprived to most affluent. 586 (5%) of 12 807 women in the intervention group and 52 (<1%) of 13 247 women in the control group received the incorrect invitation letter for their group but were still included in the intention-to-treat analysis as if they had received the correct letter. 455 (4%) women in the intervention group and 119 (<1%) women in the control group were judged to be potentially excludable, because of the nature of the intervention (as noted in the Methods section). The majority of these women were either recently screened and so invited in error (121 [21%] of 574), had missing reason for non-attendance (162 [28%]), or were not known to be or no longer living at the address held (194 [34%]). These women were also retained in our intention-to-treat analyses.

n the Methods section). The majority of these women were either recently screened and so invited in error (121 [21%] of 574), had missing reason for non-attendance (162 [28%]), or were not known to be or no longer living at the address held (194 [34%]). These women were also retained in our intention-to-treat analyses. In total, 4493 (17%) of 26 054 women participated in breast cancer screening within 90 days of the date of their first offered appointment. A significantly higher proportion of women in the intervention group participated within 90 days than did women in the control group (2861 [22%] of 12 807 vs 1632 [12%] of 13 247; RR 1·81 [95% CI 1·70–1·93], p<0·0001; table 2). A higher proportion of women in the intervention group than in the control group also met the secondary endpoint of participation within 180 days of the screening episode being opened (RR 1·77 [95% CI 1·67–1·88], p<0·0001). Similar results were obtained in a post-hoc per-protocol analysis and an analysis excluding the women deemed potentially excludable (data not shown).Table 2 Participation at second invitation for all women in the trial Intervention group (n=12 807) Control group (n=13 247) Absolute difference in attendance RR (95% CI) p value Within 90 days of first offered appointment 2861 (22%) 1632 (12%) 10% 1·81 (1·70–1·93) <0·0001 Within 180 days of episode opened 3054 (24%) 1784 (13%) 10% 1·77 (1·67–1·88) <0·0001 Data are number who participated (%) unless otherwise stated. RR=relative risk. Percentages have been rounded up.

ndance RR (95% CI) p value Within 90 days of first offered appointment 2861 (22%) 1632 (12%) 10% 1·81 (1·70–1·93) <0·0001 Within 180 days of episode opened 3054 (24%) 1784 (13%) 10% 1·77 (1·67–1·88) <0·0001 Data are number who participated (%) unless otherwise stated. RR=relative risk. Percentages have been rounded up. Overall, 12 817 (49%) of 26 054 women were offered a prevalent screen (6300 in the intervention group vs 6517 in the control group) and 13 237 (51%) women were offered an incident screen (6507 in the intervention group vs 6730 in the control group). To take into account the fact that the prevalent screen includes women who have been invited before but have never attended, we first analysed younger women in the prevalent round (ie, those aged 50–52 years). In this subgroup, participation within 90 days of the first offered appointment was significantly greater for women in the intervention group than in the control group (table 3); participation within 180 days of the episode being opened supported this result. Next, we analysed data for prevalent screen women aged 53–70 years, who had never previously attended. Although numbers were small, participation at second invitation was still significantly higher in the intervention group than in the control group (table 3).Table 3 Participation at second invitation for prevalent screen women, by age group

for prevalent screen women aged 53–70 years, who had never previously attended. Although numbers were small, participation at second invitation was still significantly higher in the intervention group than in the control group (table 3).Table 3 Participation at second invitation for prevalent screen women, by age group Intervention group Control group Absolute difference in attendance RR (95% CI) p value 50–52 years Number invited 2017 2072 .. .. .. Participation in screening Within 90 days of first offered appointment 347 (17%) 147 (7%) 10% 2·42 (1·99–2·95) <0·0001 Within 180 days of episode opened 369 (18%) 163 (8%) 10% 2·33 (1·93–2·80) <0·0001 53–70 years Number invited 4283 4445 .. .. .. Participation in screening Within 90 days of first offered appointment 283 (7%) 82 (2%) 5% 3·58 (2·80–4·58) <0·0001 Within 180 days of episode opened 307 (7%) 97 (2%) 5% 3·28 (2·61–4·13) <0·0001 Data are number who participated (%) unless otherwise stated. RR=relative risk. Percentages for the difference in attendance have been rounded up.

ening Within 90 days of first offered appointment 283 (7%) 82 (2%) 5% 3·58 (2·80–4·58) <0·0001 Within 180 days of episode opened 307 (7%) 97 (2%) 5% 3·28 (2·61–4·13) <0·0001 Data are number who participated (%) unless otherwise stated. RR=relative risk. Percentages for the difference in attendance have been rounded up. Participation data for incident screen women aged 53–70 years who had attended any time previously are shown in table 4 by time since last attended screen before the date of the first offered appointment for this screening episode. Age intervals were adjusted accordingly (eg, only women aged 56–70 years were included in the group who last attended their screening 6–9 years before their first offered appointment). Despite numbers of women participating diminishing with increasing time since last attendance, all results were significantly in favour of the intervention, even for women who had attended previously but 9 years or more before their first offered appointment (table 4). For women who had last attended 1–3 years previously, the expected proportion of women participating in screening within 180 days in the intervention group if there had been no effect of the intervention is 32% (attendance in the control group) of 2853 (number invited in the intervention group, n=912). Therefore, the effect of 2853 second timed appointments was generating 495 (ie, 1407 [the number of attendees in the intervention group]–912 [the number expected of attendees]) attended screens. Thus, in this group about six (2853/495) second timed appointments would have to be offered per additional participant attending screening. The corresponding numbers of second timed appointments required per additional participant are six, 15, and 26 for women whose last attendance was 3–6 years, 6–9 years, and 9 or more years before their current first offered appointment, respectively, calculated as for those attending 1–3 years previously.Table 4 Participation at second invitation for incident screen women, overall and by time since last attendance

x, 15, and 26 for women whose last attendance was 3–6 years, 6–9 years, and 9 or more years before their current first offered appointment, respectively, calculated as for those attending 1–3 years previously.Table 4 Participation at second invitation for incident screen women, overall and by time since last attendance Intervention group Control group Absolute difference in attendance RR (95% CI) p value Attended any time previously Number invited 6507 6730 .. .. .. Participation in screening Within 90 days of first offered appointment 2231 (34%) 1403 (21%) 13% 1·64 (1·53–1·78) <0·0001 Within 180 days of episode opened 2378 (37%) 1524 (23%) 14% 1·61 (1·51–1·73) <0·0001 Aged 51–70 years who attended 1 to <3 years previously Number invited 2853 2992 .. .. .. Participation in screening Within 90 days of first offered appointment 1307 (46%) 876 (29%) 17% 1·56 (1·43–1·71) <0·0001 Within 180 days of episode opened 1407 (49%) 956 (32%) 17% 1·54 (1·42–1·68) <0·0001 Aged 53–70 years who attended 3 to <6 years previously Number invited 1633 1638 .. .. .. Participation in screening Within 90 days of first offered appointment 568 (35%) 306 (19%) 16% 1·86 (1·62–2·14) <0·0001 Within 180 days of episode opened 590 (36%) 327 (20%) 16% 1·81 (1·58–2·08) <0·0001 Aged 56–70 years who attended 6 to <9 years previously Number invited 529 582 .. .. .. Participation in screening Within 90 days of first offered appointment 71 (13%) 39 (7%) 7% 2·00 (1·35–2·97) <0·0001 Within 180 days of episode opened 76 (14%) 45 (8%) 7% 1·86 (1·28–2·69) <0·0001 Aged 59–70 years who attended ≥9 years previously Number invited 471 453 .. .. .. Participation in screening Within 90 days of first offered appointment 35 (7%) 16 (4%) 4% 2·10 (1·16–3·81) 0·01 Within 180 days of episode opened 37 (8%) 18 (4%) 4% 1·98 (1·12–3·48) 0·02 Data are number who participated (%) unless otherwise stated. RR=relative risk. Percentages for the difference in attendance have been rounded up.

in screening Within 90 days of first offered appointment 35 (7%) 16 (4%) 4% 2·10 (1·16–3·81) 0·01 Within 180 days of episode opened 37 (8%) 18 (4%) 4% 1·98 (1·12–3·48) 0·02 Data are number who participated (%) unless otherwise stated. RR=relative risk. Percentages for the difference in attendance have been rounded up. Formal tests for heterogeneity of the effect of the intervention by prevalent or incident status were significant (p<0·0001 for participation within 90 days of the first offered appointment and within 180 days of the episode being opened). Separate results for prevalent and incident screens can be seen in Table 3, Table 4. Although the relative effects are larger in the prevalent screen women, the absolute differences in participation are larger in the incident screen women (Table 3, Table 4).

appointment and within 180 days of the episode being opened). Separate results for prevalent and incident screens can be seen in Table 3, Table 4. Although the relative effects are larger in the prevalent screen women, the absolute differences in participation are larger in the incident screen women (Table 3, Table 4). Results did not vary substantially by age group (data not shown). Results by national IMD quintile are shown in table 5. We excluded 42 women who had missing IMD data from this analysis (18 in the intervention group and 24 in the control group). The first two quintiles (1 and 2), corresponding to the most deprived populations, have higher RRs than the other quintiles for participation within 90 days of the first offered appointment and participation within 180 days of the episode being opened. From the third to fifth quintiles, RRs decrease for more affluent women. However, it should be noted that the absolute differences in participation between the two groups were similar, at around 10%, in all quintiles. Results were highly significant in all quintiles (p<0·0001 in all cases).Table 5 Participation at second invitation for all national IMD quintiles (from most to least deprived)

owever, it should be noted that the absolute differences in participation between the two groups were similar, at around 10%, in all quintiles. Results were highly significant in all quintiles (p<0·0001 in all cases).Table 5 Participation at second invitation for all national IMD quintiles (from most to least deprived) Intervention group Control group Absolute difference in attendance RR (95% CI) p value IMD quintile 1 (most deprived) Number invited 3395 3623 .. .. .. Participation in screening Within 90 days of first offered appointment 639 (19%) 353 (10%) 9% 1·93 (1·69–2·20) <0·0001 Within 180 days of episode opened 682 (20%) 386 (11%) 9% 1·89 (1·66–2·14) <0·0001 IMD quintile 2 Number invited 3645 3703 .. .. .. Participation in screening Within 90 days of first offered appointment 768 (21%) 398 (11%) 10% 1·96 (1·73–2·22) <0·0001 Within 180 days of episode opened 825 (23%) 434 (12%) 11% 1·93 (1·71–2·17) <0·0001 IMD quintile 3 Number invited 2864 2978 .. .. .. Participation in screening Within 90 days of first offered appointment 686 (24%) 402 (13%) 10% 1·77 (1·56–2·01) <0·0001 Within 180 days of episode opened 734 (26%) 442 (15%) 11% 1·73 (1·53–1·95) <0·0001 IMD quintile 4 Number invited 1946 2028 .. .. .. Participation in screening Within 90 days of first offered appointment 488 (25%) 297 (15%) 10% 1·71 (1·48–1·98) <0·0001 Within 180 days of episode opened 519 (27%) 324 (16%) 11% 1·67 (1·45–1·92) <0·0001 IMD quintile 5 (least deprived) Number invited 939 891 .. .. .. Participation in screening Within 90 days of first offered appointment 277 (29%) 178 (20%) 10% 1·48 (1·22–1·79) <0·0001 Within 180 days of episode opened 290 (31%) 194 (22%) 9% 1·42 (1·18–1·71) <0·0001 Data are number who participated (%) unless otherwise stated. IMD=2010 Index of Multiple Deprivation. RR=relative risk. Percentages fhave been rounded up.

days of first offered appointment 277 (29%) 178 (20%) 10% 1·48 (1·22–1·79) <0·0001 Within 180 days of episode opened 290 (31%) 194 (22%) 9% 1·42 (1·18–1·71) <0·0001 Data are number who participated (%) unless otherwise stated. IMD=2010 Index of Multiple Deprivation. RR=relative risk. Percentages fhave been rounded up. In our post-hoc exploratory analyses, the intervention significantly increased participation at all study sites (p<0·0001 in all centres; appendix p 2) compared with that in the control group. The effect of the intervention was highest in southeast London (RR for participation within 90 days of the first offered appointment 2·27 [95% CI 1·90–2·72]), which had the highest proportion of women in the two most deprived IMD quintiles (3249 [87%] of 3738), and only 53 (1%) women in the two most affluent quintiles. By contrast, the effect of the intervention was smallest in Plymouth (RR 1·55 [95% CI 1·36–1·77]), where 3595 (50%) of 7190 women were in the two most deprived IMD quintiles and 1706 (24%) were in the two most affluent quintiles. Discussion In this study, the intervention of inviting non-attenders of breast cancer screening to a second appointment with a fixed date and time caused an absolute increase in participation of 10·3% compared with an open invitation, which would translate to an increase in participation of 3%, since around 30% of invitees to a first appointment were non-attenders.1 Most women included in our analysis were younger than 60 years of age and came from more deprived socioeconomic areas.

ncrease in participation of 10·3% compared with an open invitation, which would translate to an increase in participation of 3%, since around 30% of invitees to a first appointment were non-attenders.1 Most women included in our analysis were younger than 60 years of age and came from more deprived socioeconomic areas. A limitation of this study was that more than 20% of women were excluded after randomisation. Most excluded women were outside the screening age range, so this factor is unlikely to be a source of bias. Moreover, although the six sites were spread across England, they might not be representative of the English population overall. However, we have no reason to believe that this is the case, and the primary result was similar in all centres. Also, we used pseudorandomisation, allocating to trial group by whether the SX number was odd or even. However, because SX numbers are not assigned in a systematic way by screening centres, the practice should allow valid comparison of trial groups. The control group was larger than the intervention group—ie, there were more women with even SX numbers. Since the SX numbers used for allocation in this trial are assigned strictly sequentially at first invitation to the programme, we are unsure why this difference occurred. We have been unable to identify a systematic factor or staff action that could have caused this imbalance, although centres with more error codes for the sending of the wrong letter were also those with the larger imbalances between the sizes of the two groups. However, analysis was by intention to treat. Notably, the previous trial of second timed appointments for non-attenders had a similar imbalance, although in the other direction (more odd than even SX numbers).6

e sending of the wrong letter were also those with the larger imbalances between the sizes of the two groups. However, analysis was by intention to treat. Notably, the previous trial of second timed appointments for non-attenders had a similar imbalance, although in the other direction (more odd than even SX numbers).6 Nearly twice as many women in the intervention group than the control group participated in screening at second invitation. Although the intention-to-treat analysis led to more diluted results than less statistically cautious approaches, the improvement in participation in the intervention group was still substantial, as shown by our primary analysis. These results, therefore, support the NHSBSP policy of second timed appointments. The greater effect at prevalent screen than incident screen is important, since a stronger decline in participation over time has been reported at first invitation than at subsequent invitations.1 The universal adoption of second timed appointments for non-attenders at first invitation could go some way to rectifying this trend.

r effect at prevalent screen than incident screen is important, since a stronger decline in participation over time has been reported at first invitation than at subsequent invitations.1 The universal adoption of second timed appointments for non-attenders at first invitation could go some way to rectifying this trend. Evidence at the ecological and individual level from Europe and North America suggests that socioeconomic deprivation, often in conjunction with specific ethnicity, is strongly associated with non-participation in breast cancer screening.11, 12, 13, 14, 15, 16 Transport issues and difficulties in getting to the screening appointment are cited as reasons for non-participation.12, 14 Further detailed surveys of non-participants are needed to improve our understanding of the barriers to participation. In our trial, women living in areas of higher deprivation showed a better response to the intervention than did women living in areas of lower deprivation. Thus, the practice of second timed appointments for non-attenders could address in part the socioeconomic gradient in delivery of the breast screening service.2, 3 At the very least, this practice should not exacerbate the problem. This gradient needs to be addressed because the clinical stage of presentation tends to be later for women of lower socioeconomic status than for women of higher status, and is a factor that contributes to worse treatment outcomes.11 Arguably, a key factor responsible for women not attending their breast screening invitation might be car ownership, which is strongly correlated with socioeconomic status and positively associated with breast screening coverage.12 A comparison of car ownership within the same IMD quintile between women who responded to their second timed appointment letter and women who did not could be interesting.

nvitation might be car ownership, which is strongly correlated with socioeconomic status and positively associated with breast screening coverage.12 A comparison of car ownership within the same IMD quintile between women who responded to their second timed appointment letter and women who did not could be interesting. The greatest effect of the intervention was seen in southeast London, which had the highest proportion of women in the two lowest IMD quintiles. Plymouth, with the second largest proportion of study participants in the two lowest IMD quintiles, showed the smallest (albeit still substantial) effect of the intervention. In view of our findings, and the internationally observed socioeconomic gradient in participation in breast cancer screening,14, 15, 16 our results could have relevance to other countries with organised screening programmes.

owest IMD quintiles, showed the smallest (albeit still substantial) effect of the intervention. In view of our findings, and the internationally observed socioeconomic gradient in participation in breast cancer screening,14, 15, 16 our results could have relevance to other countries with organised screening programmes. In all time-interval categories into which women were divided, the proportions of women who participated in screening increased in the intervention group versus the control by around the same relative factor for second timed appointments and there did not seem to be a trend between time since last screen and efficacy of the intervention, by contrast with findings reported by Stead and colleagues.6 However, the absolute effect declined with time since last attended screen, as has been reported previously, and reflecting generally similar or larger relative effects but smaller absolute effects in those with a lower baseline participation. The number of second timed appointment letters that need to be offered per additional participant increases for women whose last attendance at breast cancer screening was more than 6 years before their current first offered appointment. Of course, fewer second timed appointment letters have to be issued than open letters per participant; however, reservation of the appointment time is the call on resources.

icipant increases for women whose last attendance at breast cancer screening was more than 6 years before their current first offered appointment. Of course, fewer second timed appointment letters have to be issued than open letters per participant; however, reservation of the appointment time is the call on resources. The economic implications of compulsorily extending the intervention to all women in the breast cancer screening programme will have to be examined before such a change in policy is made, because allocating time slots for fixed timed appointments has a cost in terms of resources. In most cases, overbooking for screening appointments is advised to minimise the waste from unused slots; our results suggest that increasing the overbooking ratio for second timed appointments for previous non-attenders would be safe. However, with the higher variability of the likelihood of participation in the second timed appointments, it would be prudent to mix small numbers of these within sessions with a majority of first offered appointments. Booking software (eg, Smart Clinics) is now available that overbooks by a magnitude depending on the likelihood of attendance. Offering second timed appointments only to those who have attended in the 6 years previous to their first offered appointment might be a cost-effective approach, since it leads to fewer wasted appointments than in those who have not attended for a longer period. This strategy, however, raises questions of ethics and equity and should be considered further by the Department of Health and Public Health England to determine the appropriate policy for the programme. Scope also exists for qualitative research into the public acceptability of the policy of second timed appointments.

This strategy, however, raises questions of ethics and equity and should be considered further by the Department of Health and Public Health England to determine the appropriate policy for the programme. Scope also exists for qualitative research into the public acceptability of the policy of second timed appointments. In other countries, such as Italy,17, 18 fixed appointments rather than open invitations have been associated with improved participation in screening in general, not only for non-attenders. Other methods that can successfully increase participation in breast cancer screening are text message,19, 20 postal,21 and telephone reminders;22 general practitioner endorsement;23, 24 and the possibility to change appointments to out-of-office hours.8 Second timed appointments for non-attenders could be implemented with these methods (eg, with the second timed appointment letter offering the opportunity to change to an out-of-hours time slot, or having primary care endorsement) and other interventions to improve delivery to all eligible women, with the ultimate goal of improving early detection of breast cancer in England and worldwide. Supplementary Material Supplementary appendix

In other countries, such as Italy,17, 18 fixed appointments rather than open invitations have been associated with improved participation in screening in general, not only for non-attenders. Other methods that can successfully increase participation in breast cancer screening are text message,19, 20 postal,21 and telephone reminders;22 general practitioner endorsement;23, 24 and the possibility to change appointments to out-of-office hours.8 Second timed appointments for non-attenders could be implemented with these methods (eg, with the second timed appointment letter offering the opportunity to change to an out-of-hours time slot, or having primary care endorsement) and other interventions to improve delivery to all eligible women, with the ultimate goal of improving early detection of breast cancer in England and worldwide. Supplementary Material Supplementary appendix Acknowledgments We thank all the staff who worked at the screening centres that took part to this project. The fieldwork was funded by the National Health Service Cancer Screening Programmes. SWD and RM contributed to this study as part of the programme of the Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis, which receives funding for a research programme from the Department of Health Policy Research Programme. This programme is a collaboration between researchers from seven institutions (Queen Mary University of London, University College London, King's College London, London School of Hygiene & Tropical Medicine, Hull York Medical School, Durham University, and Peninsula Medical School).

of Health Policy Research Programme. This programme is a collaboration between researchers from seven institutions (Queen Mary University of London, University College London, King's College London, London School of Hygiene & Tropical Medicine, Hull York Medical School, Durham University, and Peninsula Medical School). Contributors PCA contributed to study design, data analysis, data interpretation, and drafting the report. RM contributed to data analysis, data interpretation, and drafting the report. SH was responsible for informatics management and contributed to editing the report. JO contributed to study design and editing the report. AET, LP, JS, GK, CEI, and JS were responsible for leading the study at the screening sites and contributed to editing the report. CF contributed to the informatics management at one of the sites and to editing the report. AGT, RG, and AJM contributed to study design and editing the report. JP was responsible for study concept and design, and contributed to editing the report. SWD was the chief investigator of the study. He was responsible jointly with JP for study concept and design, supervised the statistical analysis and data interpretation, and contributed to drafting the report. Declaration of interests We declare no competing interests.

Research in context Evidence before this study The role of thoracic radiotherapy is well established in the management of limited-stage small-cell lung cancer, and the standard of care in patients with good performance status is concurrent chemoradiotherapy. However, the optimal radiotherapy dose and fractionation remains controversial. One standard of care is twice-daily radiotherapy, which was shown to be superior to once-daily radiotherapy in a landmark Intergroup 0096 study in 1999. We searched PubMed and the abstracts of major conferences (such as the American Society of Clinical Oncology) with the terms “small cell lung cancer”, “limited-stage”, “radiotherapy (or irradiation)”, and “chemotherapy”, with no constraints imposed on the timeframe for the search, for randomised evidence to support this practice. We found only one relevant randomised clinical trial comparing once-daily and twice-daily radiotherapy. Added value of this study

The role of thoracic radiotherapy is well established in the management of limited-stage small-cell lung cancer, and the standard of care in patients with good performance status is concurrent chemoradiotherapy. However, the optimal radiotherapy dose and fractionation remains controversial. One standard of care is twice-daily radiotherapy, which was shown to be superior to once-daily radiotherapy in a landmark Intergroup 0096 study in 1999. We searched PubMed and the abstracts of major conferences (such as the American Society of Clinical Oncology) with the terms “small cell lung cancer”, “limited-stage”, “radiotherapy (or irradiation)”, and “chemotherapy”, with no constraints imposed on the timeframe for the search, for randomised evidence to support this practice. We found only one relevant randomised clinical trial comparing once-daily and twice-daily radiotherapy. Added value of this study Although twice-daily radiotherapy has produced the best outcomes in these patients so far, concerns about its toxicity, logistical issues in the delivery of twice-daily radiotherapy, and the low radiation dose used in the control group of the Intergroup 0096 study have resulted in the poor adoption of this regimen and no consensus on the standard treatment to use in the routine setting. The CONVERT trial provides further evidence supporting the use of twice-daily radiotherapy in the routine setting and will help to standardise patient care. Furthermore, the results of this study show that in the era of modern radiotherapy techniques, the frequency and severity of acute and late radiation toxicities are lower than previously reported.

r evidence supporting the use of twice-daily radiotherapy in the routine setting and will help to standardise patient care. Furthermore, the results of this study show that in the era of modern radiotherapy techniques, the frequency and severity of acute and late radiation toxicities are lower than previously reported. Implications of all the available evidence Results from this study showed that twice-daily radiotherapy should be considered standard-of-care in patients with limited-stage small-cell lung cancer. The implication for future research is that overall treatment duration of radiotherapy should be kept short when combined with chemotherapy. This Article provides updated information on expected treatment toxicity that clinicians can relay to their patients.

f-care in patients with limited-stage small-cell lung cancer. The implication for future research is that overall treatment duration of radiotherapy should be kept short when combined with chemotherapy. This Article provides updated information on expected treatment toxicity that clinicians can relay to their patients. Introduction Small-cell lung cancer is characterised by its rapid tumour doubling time, early dissemination, and high response rate to both chemotherapy and radiotherapy. Of the 42 000 patients in the UK and 225 000 in the USA diagnosed with lung cancer every year, 15% have small-cell lung cancer and 30% of those have limited-stage disease that can be encompassed within a tolerable radiotherapy field.1 Even in this early-stage disease, outcomes are poor, with median survival of 16–24 months after curative intent treatment and 2-year survival of less than 50%.2, 3, 4 Combined chemotherapy and thoracic radiotherapy is the standard treatment for limited-stage small-cell lung cancer. Results from two meta-analyses5, 6 showed that the addition of radiotherapy to chemotherapy improves median survival, 3-year survival, and local control. Subsequently, meta-analyses of clinical trials investigating the optimal timing and sequencing of chemoradiotherapy have shown an advantage for early concurrent thoracic radiotherapy.7, 8, 9, 10, 11 Furthermore, twice-daily radiotherapy was superior to once-daily radiotherapy in the landmark Intergroup 0096 study.4 In that study, patients were randomly assigned to receive either 45 Gy once-daily (1·8 Gy per fraction) for 5 weeks or 45 Gy twice-daily (1·5 Gy per fraction) for 3 weeks. In both groups, radiotherapy was given concurrently, starting with the first cycle of chemotherapy. Twice-daily radiotherapy significantly improved 5-year overall survival compared with once-daily radiotherapy (26% vs 16%) and reduced the risk of thoracic relapse (36% vs 52%) but at the cost of increased severe radiation oesophagitis (32% vs 16%). Consequently, twice-daily radiotherapy concurrently with chemotherapy was adopted as a standard of care for limited-stage small-cell lung cancer.12 However, it is unclear whether twice-daily radiotherapy resulted in better outcomes because of the increase in the biologically effective dose of radiation or because of shorter overall treatment time, which is important in this rapidly proliferating disease.

standard of care for limited-stage small-cell lung cancer.12 However, it is unclear whether twice-daily radiotherapy resulted in better outcomes because of the increase in the biologically effective dose of radiation or because of shorter overall treatment time, which is important in this rapidly proliferating disease. Radiotherapy techniques have evolved since the Intergroup 0096 study was designed in the late 1980s; specifically, the use of CT-planned conformal treatment and the omission of elective nodal irradiation to reduce normal tissue exposure and toxicity, particularly oesophagitis. Although twice-daily radiotherapy concurrently with chemotherapy has produced the best outcomes so far, concerns about its toxicity, logistical issues in its delivery, and the low radiation dose in the control group of the Intergroup 0096 study, resulting in a very high (52%) local failure rate, have resulted in the poor adoption of this regimen and no consensus on the standard treatment to use in the routine setting.13 The authors of one study14 suggested that the local control could be improved with a higher dose of once-daily radiotherapy. The CONVERT trial was therefore designed as a superiority trial to improve on the standard of care for limited-stage small-cell lung cancer by comparing twice-daily radiotherapy to a higher dose of radiotherapy delivered once daily, given concurrently with chemotherapy.

ed with a higher dose of once-daily radiotherapy. The CONVERT trial was therefore designed as a superiority trial to improve on the standard of care for limited-stage small-cell lung cancer by comparing twice-daily radiotherapy to a higher dose of radiotherapy delivered once daily, given concurrently with chemotherapy. Methods Study design and participants The CONVERT trial was an international, multicentre, open-label, randomised phase 3 superiority trial. Details of the trial design have been published previously.15 Patients were recruited at 73 centres in eight countries (Belgium, Canada, France, Netherlands, Poland, Slovenia, Spain, and the UK; appendix pp 1–2).

ants The CONVERT trial was an international, multicentre, open-label, randomised phase 3 superiority trial. Details of the trial design have been published previously.15 Patients were recruited at 73 centres in eight countries (Belgium, Canada, France, Netherlands, Poland, Slovenia, Spain, and the UK; appendix pp 1–2). Eligible patients were aged 18 years or older; had histologically or cytologically confirmed small-cell lung cancer with limited disease (as defined by the Veterans Administration Lung Cancer Study Group—ie, patients whose disease can be encompassed within a radical radiation portal);16 had an Eastern Cooperative Oncology Group performance status of 0–117 or performance status of 2 due to disease-related symptoms and not comorbidities (since small-cell lung cancer is characterised by rapid doubling time and central disease location, which can be associated with a sudden change in performance status); had no malignant pleural or pericardial effusions; and had acceptable radiotherapy target volume (according to the local radiotherapist). Eligible patients had a maximum of one adverse biochemical factor (concentrations of serum alkaline phosphatase >1·5-times the upper limit of normal, serum sodium <lower limit of normal, and serum lactate dehydrogenase >the upper limit of normal), forced expiratory volume in 1 s greater than 1 L or 40% predicted value, and transfer factor for carbon monoxide greater than 40% predicted value. Patients with a previous history of malignancy in the past 5 years (except for non-melanomatous skin or in-situ cervix carcinoma) and those with previous or concomitant illness or treatment that, in the opinion of the investigator, would interfere with the trial treatments or comparisons were excluded.

predicted value. Patients with a previous history of malignancy in the past 5 years (except for non-melanomatous skin or in-situ cervix carcinoma) and those with previous or concomitant illness or treatment that, in the opinion of the investigator, would interfere with the trial treatments or comparisons were excluded. Participants gave written informed consent and the study was done according to the Declaration of Helsinki and Good Clinical Practice Guidelines. The trial was reviewed in the UK by the National Research Ethics Service Committee North West–Greater Manchester Central, which granted ethics approval for the study on Dec 21, 2007 (REC reference: 07/H1008/229). The protocol was also approved by the institutional review board or research ethics committee in each country and at each study centre. Randomisation and masking Patients were randomly assigned (1:1) to one of the two treatment groups (twice-daily vs once-daily radiotherapy). Allocation to treatment group was done by phone call or fax from the recruiting centre to the Manchester Academic Health Science Centre Trials Coordination Unit. The allocation method used was minimisation with a random element using a bespoke computer application. The factors controlled for in the allocation were institution, planned number of chemotherapy cycles (four vs six), and performance status (0–1 vs 2). Patients and investigators were not masked to treatment allocation.

e allocation method used was minimisation with a random element using a bespoke computer application. The factors controlled for in the allocation were institution, planned number of chemotherapy cycles (four vs six), and performance status (0–1 vs 2). Patients and investigators were not masked to treatment allocation. Procedures At baseline, all patients underwent baseline investigations, which included physical examination, chest radiograph, CT scan of the thorax and upper abdomen, CT or MRI of the brain, full blood count, biochemical profile, and lung function tests. PET/CT scans were allowed but not mandatory. Staging was done using the Union for International Cancer Control/American Joint Committee on Cancer classification system.18

n, chest radiograph, CT scan of the thorax and upper abdomen, CT or MRI of the brain, full blood count, biochemical profile, and lung function tests. PET/CT scans were allowed but not mandatory. Staging was done using the Union for International Cancer Control/American Joint Committee on Cancer classification system.18 Patients were randomly assigned to receive radiotherapy either twice-daily (45 Gy in 30 twice-daily fractions of 1·5 Gy, with a minimum of 6 h between fractions, over 19 days, on 5 consecutive days a week) or once-daily (66 Gy in 33 daily fractions of 2 Gy over 45 days, on 5 consecutive days a week), concurrently with chemotherapy. Chemotherapy was started within 4 weeks of randomisation and consisted of four to six cycles of cisplatin and etoposide every 3 weeks in both groups (etoposide 100 mg/m2 intravenously on days 1–3 and cisplatin 75 mg/m2 intravenously on day 1, or etoposide 100 mg/m2 intravenously on days 1–3 and cisplatin 25 mg/m2 intravenously on days 1–3). Each centre had to elect to prescribe four or six cycles for all eligible trial patients. The first cycle of chemotherapy was given before radiotherapy and the second was given concurrently with radiotherapy if no delay with chemotherapy occurred. No later than 6 weeks after the last cycle of chemotherapy, patients without evidence of progressive disease on the CT scan and with no clinical evidence of brain metastases were offered prophylactic cranial irradiation.

apy and the second was given concurrently with radiotherapy if no delay with chemotherapy occurred. No later than 6 weeks after the last cycle of chemotherapy, patients without evidence of progressive disease on the CT scan and with no clinical evidence of brain metastases were offered prophylactic cranial irradiation. Radiotherapy commenced on day 22 of cycle one of chemotherapy, coinciding with cycle two of chemotherapy in patients not experiencing chemotherapy delay due to toxicity. 3D conformal radiotherapy was mandatory and elective nodal irradiation was not permitted. The total dose was prescribed at the International Commission on Radiation Units and Measurements reference point. Intensity-modulated radiotherapy and PET/CT planning was permitted but not mandated. The protocol specified that if dose constraints to the organs at risk could not be met, the dose delivered could be decreased accordingly. The policy for chemotherapy was to delay and give at full dose later, rather than give at a reduced dose. However, we recommended a chemotherapy treatment delay of more than 7 days for grade 4 febrile neutropenia, grade 4 thrombocytopenia requiring medical intervention, or grade 2 or worse bleeding with thrombocytopenia; for the first episode of such an event, we recommended full-dose chemotherapy and granulocyte colony-stimulating factor support, or a 20% dose reduction. In case of a second episode, we recommended a 30% dose reduction. If a third episode occurred, the patient was removed from the trial.

bleeding with thrombocytopenia; for the first episode of such an event, we recommended full-dose chemotherapy and granulocyte colony-stimulating factor support, or a 20% dose reduction. In case of a second episode, we recommended a 30% dose reduction. If a third episode occurred, the patient was removed from the trial. A radiotherapy quality assurance programme was set up to ensure the robustness of the radiotherapy procedures, and the details of the programme have been reported previously.15 The programme was managed by the UK National Cancer Research Institute Radiotherapy Trials Quality Assurance Team. On completion of study treatment, patients were followed up weekly until the resolution of acute side-effects, then every 3 months until 1 year, and every 6 months for 5 years. A CT scan of the thorax and abdomen was done at 4 weeks after cycle four (even if six cycles were given). Subsequently, during follow-up at 6 and 12 months after randomisation, investigations included physical evaluation, reporting of adverse events, and a CT scan of the thorax and abdomen. Follow-up investigations were done according to local policy.

en was done at 4 weeks after cycle four (even if six cycles were given). Subsequently, during follow-up at 6 and 12 months after randomisation, investigations included physical evaluation, reporting of adverse events, and a CT scan of the thorax and abdomen. Follow-up investigations were done according to local policy. Outcomes The primary outcome of the study was overall survival, defined as time from randomisation until death from any cause. Secondary outcomes included compliance with chemotherapy and radiotherapy (defined as dose intensity delivered), acute toxicity (defined as toxicity occurring between the start of treatment and up to 3 months after completion of treatment, and assessed according to the Common Terminology Criteria for Adverse Events [version 3.0]), late toxicity (according to the Common Terminology Criteria for Adverse Events [version 3.0]),19 and local and metastatic progression-free survival (calculated from date of randomisation to date of first clinical or radiological evidence of progressive disease at the primary site or distant sites). With regard to toxicity, frequencies of worst recorded grade of toxicity in the respective time periods were recorded. Response rate was another secondary outcome but it was not analysed because interpretation of CT imaging would have been too complex after concurrent chemoradiotherapy. The study also had post-hoc exploratory translational objectives, which will be reported at a later date. All serious adverse events were reported to the trial coordinating centre and were assessed for causality and expectedness, both locally by the Principal Investigator and centrally by the Chief Investigator.

erapy. The study also had post-hoc exploratory translational objectives, which will be reported at a later date. All serious adverse events were reported to the trial coordinating centre and were assessed for causality and expectedness, both locally by the Principal Investigator and centrally by the Chief Investigator. Statistical analysis Our hypothesis was that overall survival in the once-daily chemoradiotherapy group would be superior to that of the twice-daily group. A 12% higher overall survival at 2 years in the once-daily group versus the twice-daily group was considered to be clinically significant to show superiority of the once-daily regimen. Overall and progression-free survival were estimated with the Kaplan-Meier method, and between-group comparisons evaluated by the log-rank test with stratification for institution, planned number of chemotherapy cycles (four vs six), and performance status (0–1 vs 2). The number of events required to detect a hazard ratio (HR) for death of 0·7 with an α level (two-sided) of 0·05 and 80% power (ie, an increase in 2-year survival from 44% in the twice-daily radiotherapy group to 56% in the once-daily radiotherapy group) was 247. An additional 5% was added to the sample size of 506 patients to allow for ineligible patients, giving a total recruitment target of 532 patients. The primary survival outcome was analysed using the modified intention-to-treat principle, because four cases provided no follow-up data and hence were censored at time zero. Further details about the statistical analysis are available in the protocol.15 All randomly assigned patients who were treated with at least one study dose of chemotherapy and who were alive at the time of the first toxicity assessment were included in the safety analysis. Data were collected at each study site and monitored by the independent data monitoring committee. We submitted reports to the independent data monitoring committee on an annual basis, commencing 12 months after the first patient was randomly assigned. The statistical package used for the analyses was Stata (version 13.1).

were collected at each study site and monitored by the independent data monitoring committee. We submitted reports to the independent data monitoring committee on an annual basis, commencing 12 months after the first patient was randomly assigned. The statistical package used for the analyses was Stata (version 13.1). This trial is registered with ISRCTN, number 91927162, and ClinicalTrials.gov, number NCT00433563. Role of the funding source Cancer Research UK reviewed and approved the study design. None of the funders had a role in data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

ource Cancer Research UK reviewed and approved the study design. None of the funders had a role in data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Between April 7, 2008, and Nov 29, 2013, we recruited 547 patients from 73 centres in eight countries. We randomly assigned 274 patients to receive twice-daily chemoradiotherapy and 273 to receive once-daily chemoradiotherapy. The modified intention-to-treat survival analysis included 543 patients (273 in the twice-daily chemoradiotherapy group and 270 in the once-daily chemoradiotherapy group) because four patients were lost to follow-up (centres did not return their case report forms; (figure 1). Table 1 shows the baseline characteristics of the participants. The median age at randomisation was 62 years (IQR 29–84) in the twice-daily group and 63 years (34–81) in the once-daily group, with 83 (15%) of 547 patients being older than 70 years (32 [12%] in the twice-daily group and 51 [19%] in the once-daily group). More than 95% of patients overall had a performance status of 0–1. Less than 2% of patients were never smokers, almost two-thirds were former smokers, and just over a third were current smokers (table 1). 312 (57%) of 547 patients were staged with PET/CT, and 426 (78%) of 547 were stage III according to the Union for International Cancer Control classification (table 1).Figure 1 Trial profile

n 2% of patients were never smokers, almost two-thirds were former smokers, and just over a third were current smokers (table 1). 312 (57%) of 547 patients were staged with PET/CT, and 426 (78%) of 547 were stage III according to the Union for International Cancer Control classification (table 1).Figure 1 Trial profile *One patient withdrew consent for twice-daily radiotherapy. †Dose constraints to organs at risk not met in four patients and twice-daily radiotherapy given in error to two patients. ‡Six patients did not receive any chemotherapy and two patients died during cycle one before toxicity assessment. ¶Seven patients did not receive any chemotherapy and three patients died during cycle one before toxicity assessment. Numbers assessed and ineligible are unavailable because screening logs were not completed by all centres. Table 1 Baseline characteristics

*One patient withdrew consent for twice-daily radiotherapy. †Dose constraints to organs at risk not met in four patients and twice-daily radiotherapy given in error to two patients. ‡Six patients did not receive any chemotherapy and two patients died during cycle one before toxicity assessment. ¶Seven patients did not receive any chemotherapy and three patients died during cycle one before toxicity assessment. Numbers assessed and ineligible are unavailable because screening logs were not completed by all centres. Table 1 Baseline characteristics Twice-daily radiotherapy (n=274) Once-daily radiotherapy (n=273) Age (years) 62 (29–84) 63 (34–81) Sex Male 147 (54%) 150 (55%) Female 127 (46%) 123 (45%) Ethnicity White 262 (96%) 265 (97%) African 1 (<1%) 1 (<1%) Asian 1 (<1%) 4 (2%) Other 6 (2%) 3 (1%) Not known 4 (2%) 0 Eastern Cooperative Oncology Group performance status 0 125 (46%) 123 (45%) 1 137 (50%) 142 (52%) 2 9 (3%) 8 (3%) Not known* 3 (1%) 0 Smoking history† Never smoker 3 (1%) 4 (2%) Former smoker 174 (64%) 163 (60%) Current smoker 94 (34%) 106 (39%) Not known 3 (1%) 0 Adverse biochemical factors Elevated lactate dehydrogenase 69 (25%) 60 (22%) Hyponatraemia 57 (21%) 53 (19%) Elevated alkaline phosphate 5 (2%) 6 (2%) PET/CT staging Yes 157 (57%) 155 (57%) No 113 (41%) 118 (43%) Not known 4 (2%) 0 UICC/AJCC stage18 I 1 (<1%) 3 (1%) II 34 (12%) 48 (18%) III 219 (80%) 207 (76%) Not known 20 (7%) 15 (6%) Gross tumour volume (cc) 81·6 (1·6–635·1) 85·6 (0·5–593·0) Number of chemotherapy cycles planned Four 188 (69%) 183 (67%) Six 86 (31%) 90 (33%) Data are median (IQR) or n (%). UICC/AJCC=Union for International Cancer Control/American Joint Committee on Cancer.

48 (18%) III 219 (80%) 207 (76%) Not known 20 (7%) 15 (6%) Gross tumour volume (cc) 81·6 (1·6–635·1) 85·6 (0·5–593·0) Number of chemotherapy cycles planned Four 188 (69%) 183 (67%) Six 86 (31%) 90 (33%) Data are median (IQR) or n (%). UICC/AJCC=Union for International Cancer Control/American Joint Committee on Cancer. * Eastern Cooperative Oncology Group Performance Status was not recorded on the source documentation and case report form in three cases at baseline; in all three cases, the performance score was recorded as 0–1 on the randomisation form. † Never smokers defined as adults who have never smoked a cigarette or who smoked fewer than 100 cigarettes in their entire lifetime; former smokers defined as adults who have smoked at least 100 cigarettes in their lifetime but say they currently do not smoke; current smokers defined as adults who have smoked 100 cigarettes in their lifetime and currently smoke cigarettes every day (daily) or on some days (non-daily). The number of planned cycles of chemotherapy was four for most patients (table 1). Almost 60% of patients actually received four cycles and a further 20% received six cycles of chemotherapy (table 2).Table 2 Treatment delivered

† Never smokers defined as adults who have never smoked a cigarette or who smoked fewer than 100 cigarettes in their entire lifetime; former smokers defined as adults who have smoked at least 100 cigarettes in their lifetime but say they currently do not smoke; current smokers defined as adults who have smoked 100 cigarettes in their lifetime and currently smoke cigarettes every day (daily) or on some days (non-daily). The number of planned cycles of chemotherapy was four for most patients (table 1). Almost 60% of patients actually received four cycles and a further 20% received six cycles of chemotherapy (table 2).Table 2 Treatment delivered Twice-daily radiotherapy (n=274) Once-daily radiotherapy (n=273) p value* Chemotherapy cycles delivered (all patients) 0·89 0 6 (2%) 7 (3%) 1 15 (6%) 15 (6%) 2 8 (3%) 6 (2%) 3 23 (8%) 24 (9%) 4 161 (59%) 156 (57%) 5 5 (2%) 12 (4%) 6 56 (20%) 53 (19%) Radiotherapy treatment 0·60 Concurrent chemoradiotherapy 249 (91%) 240 (88%) Sequential chemoradiotherapy 5 (2%) 6 (2%) No radiotherapy 20 (7%) 26 (10%) Not known .. 1 (<1%) Chemotherapy cycles delivered in patients who received concurrent chemoradiotherapy† 1 3/249 (1%) 1/240 (<1%) 2 5/249 (2%) 5/240 (2%) 3 21/249 (8%) 20/240 (8%) 4 161/249 (66%) 150/240 (63%) 5 5/249 (2%) 12/240 (5%) 6 54/249 (21%) 52/240 (22%) Intensity-modulated radiotherapy 0·59 Yes 40/254‡ (16%) 43/247‡ (17%) Not known .. 1 (<1%) Prophylactic cranial irradiation 229 (84%) 220 (81%) 0·36 Data are n (%) or n/N (%). * All p values were calculated with χ2 tests (except for number of cycles, which is a Wilcoxon rank sum test).

Twice-daily radiotherapy (n=274) Once-daily radiotherapy (n=273) p value* Chemotherapy cycles delivered (all patients) 0·89 0 6 (2%) 7 (3%) 1 15 (6%) 15 (6%) 2 8 (3%) 6 (2%) 3 23 (8%) 24 (9%) 4 161 (59%) 156 (57%) 5 5 (2%) 12 (4%) 6 56 (20%) 53 (19%) Radiotherapy treatment 0·60 Concurrent chemoradiotherapy 249 (91%) 240 (88%) Sequential chemoradiotherapy 5 (2%) 6 (2%) No radiotherapy 20 (7%) 26 (10%) Not known .. 1 (<1%) Chemotherapy cycles delivered in patients who received concurrent chemoradiotherapy† 1 3/249 (1%) 1/240 (<1%) 2 5/249 (2%) 5/240 (2%) 3 21/249 (8%) 20/240 (8%) 4 161/249 (66%) 150/240 (63%) 5 5/249 (2%) 12/240 (5%) 6 54/249 (21%) 52/240 (22%) Intensity-modulated radiotherapy 0·59 Yes 40/254‡ (16%) 43/247‡ (17%) Not known .. 1 (<1%) Prophylactic cranial irradiation 229 (84%) 220 (81%) 0·36 Data are n (%) or n/N (%). * All p values were calculated with χ2 tests (except for number of cycles, which is a Wilcoxon rank sum test). † The denominator in each group is the number of patients who received concurrent chemoradiotherapy. ‡ The denominator in each group is the number of patients who received radiotherapy. At the data analysis cutoff in March 1, 2016, the median follow-up was 45 months (IQR 35–58) for those still alive. 164 (60%) of 273 patients in the twice-daily group had died, compared with 176 (65%) of 270 patients in the once-daily group.

‡ The denominator in each group is the number of patients who received radiotherapy. At the data analysis cutoff in March 1, 2016, the median follow-up was 45 months (IQR 35–58) for those still alive. 164 (60%) of 273 patients in the twice-daily group had died, compared with 176 (65%) of 270 patients in the once-daily group. In our survival analysis (which included 273 patients in the twice-daily group and 270 in the once-daily group), median overall survival was 30 months (95% CI 24–34) in the twice-daily group and 25 months (21–31) in the once-daily group (hazard ratio 1·18 [95% CI 0·95–1·45]; p=0·14; figure 2A). 2-year overall survival was 56% (95% CI 50–62) in the twice-daily group and 51% (45–57) in the once-daily group (absolute difference between the treatment groups 5·3% [95% CI −3·2% to 13·7%]). 5-year overall survival was 34% (95% CI 27–41) in the twice-daily group and 31% (25–37) in the once-daily group (absolute difference 2·8% [95% CI −6·4% to 12·0%]). In the twice-daily group versus the once-daily group, causes of death were lung cancer (152 vs 146), intercurrent deaths (six vs 14), treatment-related (three vs eight), and cardiovascular (three vs eight); causes of the 12 treatment-related deaths were radiation pneumonitis (one vs two), dementia possibly related to prophylactic cranial irradiation (none vs one), neutropenic sepsis (one vs three), septic shock (one vs none), bronchial pneumonia (none vs two), and peripheral vascalar ischaemia (one vs none).Figure 2 Overall and progression-free survival

ed deaths were radiation pneumonitis (one vs two), dementia possibly related to prophylactic cranial irradiation (none vs one), neutropenic sepsis (one vs three), septic shock (one vs none), bronchial pneumonia (none vs two), and peripheral vascalar ischaemia (one vs none).Figure 2 Overall and progression-free survival (A) Overall survival. (B) Local progression-free survival. (C) Metastatic progression-free survival. HR=hazard ratio.

ed deaths were radiation pneumonitis (one vs two), dementia possibly related to prophylactic cranial irradiation (none vs one), neutropenic sepsis (one vs three), septic shock (one vs none), bronchial pneumonia (none vs two), and peripheral vascalar ischaemia (one vs none).Figure 2 Overall and progression-free survival (A) Overall survival. (B) Local progression-free survival. (C) Metastatic progression-free survival. HR=hazard ratio. 25 (9%) of 273 patients in the twice-daily radiotherapy group and 33 (12%) of 270 in the once-daily radiotherapy group did not receive concurrent chemoradiotherapy (figure 1), giving compliance rates of 91% in the twice-daily group and 88% in the once-daily group. Less than 10% of patients did not receive any radiotherapy (20 [7%] in the twice-daily group and 26 [10%] in the once-daily group; figure 1, table 2). Of the patients who received radiotherapy, intensity-modulated radiotherapy was delivered to 40 (16%) of 254 participants in the twice-daily group versus 43 (17%) of 247 participants in the once-daily group. Prophylactic cranial irradiation was delivered to 229 (84%) of 274 vs 220 (81%) of 273 participants (table 2). More patients received the full dose of radiotherapy in the twice-daily group than in the once-daily group (p<0·0001; table 3). The optimal number of fractions, as defined in the protocol,15 (30 fractions in the twice daily group and 33 in the once daily group) were delivered in 213 (86%) of 249 patients in the twice-daily group and 192 (80%) of 240 patients in the once-daily group (p=0·10). Radiotherapy was delivered over the planned overall treatment time of 19 days in 158 (63%) of 249 patients in the twice-daily group and over the planned overall treatment time of 45 days in 114 (48%) of 240 patients in the once-daily group (p=0·0004). Protocol deviations and violations were mainly due to logistical reasons, such as public holidays.Table 3 Radiotherapy treatment delivered in patients receiving concurrent chemoradiotherapy (as per protocol)

r the planned overall treatment time of 45 days in 114 (48%) of 240 patients in the once-daily group (p=0·0004). Protocol deviations and violations were mainly due to logistical reasons, such as public holidays.Table 3 Radiotherapy treatment delivered in patients receiving concurrent chemoradiotherapy (as per protocol) Dose (Gy) Number of fractions Overall treatment time (days) <44 44–46* >46 <60 60–62 64–68* <28 28–29 30† >30 <30 30–32 33† >33 <19 19‡ 20–21§ >21¶ <45 45‡ 46–47§ >47¶ Twice-daily radiotherapy (n=249) 1 (<1%) 245 (98%) 3 (1%) .. .. .. 12 (5%) 23 (9%) 213 (86%) 1 (<1%) .. .. .. .. 15 (6%) 158 (63%) 24 (10%) 52 (20%) .. .. .. .. Once-daily radiotherapy (n=240) .. .. .. 22 (9%) 19 (8%) 199 (83%) .. .. .. .. 16 (7%) 31 (13%) 192 (80%) 1 (<1%) .. .. .. .. 41 (17%) 114 (48%) 43 (18%) 42 (18%) Data are n (%). * Full dose. † Optimal number of fractions, as defined in the protocol. ‡ Planned overall treatment time. § Deviation. ¶ Violation.

Dose (Gy) Number of fractions Overall treatment time (days) <44 44–46* >46 <60 60–62 64–68* <28 28–29 30† >30 <30 30–32 33† >33 <19 19‡ 20–21§ >21¶ <45 45‡ 46–47§ >47¶ Twice-daily radiotherapy (n=249) 1 (<1%) 245 (98%) 3 (1%) .. .. .. 12 (5%) 23 (9%) 213 (86%) 1 (<1%) .. .. .. .. 15 (6%) 158 (63%) 24 (10%) 52 (20%) .. .. .. .. Once-daily radiotherapy (n=240) .. .. .. 22 (9%) 19 (8%) 199 (83%) .. .. .. .. 16 (7%) 31 (13%) 192 (80%) 1 (<1%) .. .. .. .. 41 (17%) 114 (48%) 43 (18%) 42 (18%) Data are n (%). * Full dose. † Optimal number of fractions, as defined in the protocol. ‡ Planned overall treatment time. § Deviation. ¶ Violation. At the time of analysis, 181 (66%) of 273 patients in the twice-daily group and 189 (70%) of 270 patients in the once-daily group had disease progression (p=0·26). Median progression-free survival was 15·4 months (95% CI 13·7–19·8) in the twice-daily group and 14·3 months (12·0–17·0) in the once-daily group (hazard ratio 1·12 [95% CI 0·92–1·38]; p=0·26). Median local progression-free survival was 20·7 months (95% CI 16·1–27·9) in the twice-daily group versus 17·9 months (15·3–21·7) in the once daily group (figure 2B) and median metastatic progression-free survival was 20·2 months (95% CI 15·9–25·3) versus 16·6 months (13·7–21·8). The difference between groups for local progression-free survival (p=0·20) and metastatic progression-free survival (p=0·24) was not significant (figure 2). There was no notable difference between groups in treatment received at the time of progression (appendix p 4).

(95% CI 15·9–25·3) versus 16·6 months (13·7–21·8). The difference between groups for local progression-free survival (p=0·20) and metastatic progression-free survival (p=0·24) was not significant (figure 2). There was no notable difference between groups in treatment received at the time of progression (appendix p 4). Chemotherapy toxicity was assessed in 266 (97%) of 273 patients in the twice-daily group and 263 (97%) of 270 patients in the once-daily group, who had received at least one cycle of chemotherapy and who were alive at the time of the first toxicity assessment (figure 1, table 4). Radiotherapy toxicity was assessed in 254 (93%) of 273 patients in the twice-daily group and 246 (91%) of 270 patients in the once-daily group who had received either concurrent or sequential chemoradiotherapy (figure 1, table 4).Table 4 Acute adverse events (≤3 months after completion of study treatment)

1, table 4). Radiotherapy toxicity was assessed in 254 (93%) of 273 patients in the twice-daily group and 246 (91%) of 270 patients in the once-daily group who had received either concurrent or sequential chemoradiotherapy (figure 1, table 4).Table 4 Acute adverse events (≤3 months after completion of study treatment) Twice-daily group Once-daily group p value Grade 1–2 Grade 3 Grade 4 Grade 5 Grade 1–2 Grade 3 Grade 4 Grade 5 Adverse events in the population assessed for chemotherapy toxicity (n=266 in the twice-daily group; n=263 in the once-daily group) Nausea 172 (65%) 23 (9%) .. .. 171 (65%) 26 (10%) .. .. 0·63 Vomiting 105 (40%) 13 (5%) .. .. 95 (36%) 13 (5%) .. .. 0·99 Mucositis 88 (33%) 3 (1%) .. .. 87 (33%) 5 (2%) 1 (<1%) .. 0·34 Fatigue 212 (80%) 31 (12%) .. .. 216 (82%) 31 (12%) 2 (1%) .. 0·77 Neuropathy (motor) 12 (5%) 1 (<1%) .. .. 15 (6%) 2 (1%) .. .. 0·62 Neuropathy (sensory) 63 (24%) 3 (1%) 1 (<1%) 1 (<1%) 61 (23%) 5 (2%) .. .. >0·99 Infection 43 (16%) 27 (10%) 7 (3%) .. 52 (20%) 27 (10%) 2 (1%) 2 (1%) 0·52 Anaemia 194 (73%) 32 (12%) 1 (<1%) .. 184 (70%) 34 (13%) 1 (<1%) .. 0·72 Febrile neutropenia NA 49 (18%) 13 (5%) 1 (<1%) NA 38 (14%) 8 (3%) 3 (<1%) 0·13 Neutropenia 38 (14%) 68 (26%) 129 (49%) .. 47 (18%) 69 (26%) 101 (38%) .. 0·05 Anorexia 135 (51%) 18 (7%) .. .. 129 (49%) 21 (8%) .. .. 0·60 Other* 150 (57%) 65 (24%) 9 (3%) 1 (<1%)† 177 (67%) 44 (17%) 8 (3%)‡ 1 (<1%) 0·02 Adverse events in the population assessed for radiotherapy toxicity (n=254 in the twice-daily group; n=246 in the once-daily group) Oesophagitis 159 (63%) 46 (18%) 1 (<1%) .. 135 (54%) 47 (19%) .. .. 0·85 Pneumonitis 51 (20%) 3 (1%) 1 (<1%) 1 (<1%) 49 (19%) 3 (1%) 1 (<1%) 2 (1%) 0·70 Data are n (%). The radiotherapy toxicity population was used to analyse the prevalence of these adverse events because it would not be possible to report radiotherapy-related toxicity in patients who did not receive radiotherapy. NA=not applicable.

is 51 (20%) 3 (1%) 1 (<1%) 1 (<1%) 49 (19%) 3 (1%) 1 (<1%) 2 (1%) 0·70 Data are n (%). The radiotherapy toxicity population was used to analyse the prevalence of these adverse events because it would not be possible to report radiotherapy-related toxicity in patients who did not receive radiotherapy. NA=not applicable. * Other grade 3 reported toxicities included diarrhoea (n=7), hyponatremia (n=1), urinary retention (n=5), dysphagia (n=5), and lymphopenia (n=6) in the once-daily group; and diarrhoea (n=3), constipation (n=7), hyponatremia (n=1), dysphagia (n=8), lymphopenia (n=8), dyspnoea (n=8), and leucopenia (n=4) in the twice-daily group. Other grade 4 reported toxicities included pulmonary embolism (n=4), hyponatremia (n=2), dyspnoea (n=1), and myocardial infarction (n=1) in the once-daily group; and pulmonary embolism (n=2), hyponatremia (n=3), lymphopenia (n=3), and fast atrial fibrillation (n=1) in the once-daily group). † Two deaths (peripheral arterial ischaemia [n=1] and septic shock [n=1]). ‡ Two deaths (peripheral arterial ischaemia [n=1] in the twice-daily group and dementia possibly related to prophylactic cranial irradiation [n=1] in the once-daily group).

* Other grade 3 reported toxicities included diarrhoea (n=7), hyponatremia (n=1), urinary retention (n=5), dysphagia (n=5), and lymphopenia (n=6) in the once-daily group; and diarrhoea (n=3), constipation (n=7), hyponatremia (n=1), dysphagia (n=8), lymphopenia (n=8), dyspnoea (n=8), and leucopenia (n=4) in the twice-daily group. Other grade 4 reported toxicities included pulmonary embolism (n=4), hyponatremia (n=2), dyspnoea (n=1), and myocardial infarction (n=1) in the once-daily group; and pulmonary embolism (n=2), hyponatremia (n=3), lymphopenia (n=3), and fast atrial fibrillation (n=1) in the once-daily group). † Two deaths (peripheral arterial ischaemia [n=1] and septic shock [n=1]). ‡ Two deaths (peripheral arterial ischaemia [n=1] in the twice-daily group and dementia possibly related to prophylactic cranial irradiation [n=1] in the once-daily group). The most common grade 3–4 adverse event was neutropenia (affecting 197 [74%] of 266 patients in the twice-daily group vs 170 [65%] of 263 in the once-daily group). The frequencies of most adverse events recorded were similar in both groups, with the exception that significantly more grade 4 neutropenia was recorded in the twice-daily group than in the once-daily group (129 [49%] vs 101 [38%]; p=0·05). However, grade 3–5 febrile neutropenia did not differ significantly between the two groups (table 4). Acute radiotherapy toxicity was similar in both groups: grade 3–4 oesophagitis was reported in 47 (18%) of 254 patients in the twice-daily group and 47 (19%) of 246 patients in the once-daily group. 11 patients developed grade 3–5 radiation pneumonitis (five in the twice daily group and six in the once daily group), of whom three patients died within 3 months of radiotherapy (two in the once-daily group and one in the twice-daily group, one of whom received sequential rather than concurrent radiotherapy; table 4, appendix p 3). Regarding late toxicity, four patients in the once-daily group developed grade 3 oesophagitis, one of whom had an oesophageal stricture. Six patients in each group developed grade 3–4 pneumonitis, and five patients (three in the twice-daily group and two in the once-daily group) developed grade 3 pulmonary fibrosis (table 5).Table 5 Late adverse events (>3 months after study treatment)

oped grade 3 oesophagitis, one of whom had an oesophageal stricture. Six patients in each group developed grade 3–4 pneumonitis, and five patients (three in the twice-daily group and two in the once-daily group) developed grade 3 pulmonary fibrosis (table 5).Table 5 Late adverse events (>3 months after study treatment) Twice-daily group (n=248) Once-daily group (n=233) p value* Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 Dermatitis 15 (6%) .. .. 17 (7%) .. .. .. Oesophagitis 29 (12%) .. .. 39 (17%) 4 (2%) .. 0·06 Oesophageal stricture or fistula 8 (3%) .. .. 6 (3%) 1 (<1%) .. 0·48 Pulmonary fibrosis 119 (48%) 3 (1%) .. 106 (46%) 2 (1%) .. >0·99 Pneumonitis 71 (29%) 5 (2%) 1 (<1%) 70 (30%) 5 (2%) 1 (<1%) 0·90 Myelitis 1 (<1%)† .. .. 8 (3%)† .. .. .. Other 131 (53%) 20 (8%) 3 (1%) 113 (49%) 18 (8%) 2 (1%) 0·78 Data are n (%). * p values calculated for grade 3–4 adverse events. † All cases of myelitis were grade 1 adverse events. Discussion Our results show that once-daily radiotherapy did not improve overall survival in patients with limited-stage small-cell lung cancer and good performance status, compared with twice-daily radiotherapy, when given concurrently with chemotherapy. Radiotherapy treatment delivery was superior in the twice-daily group. Furthermore, both acute and late toxicities were similar and lower than expected with both regimens.

limited-stage small-cell lung cancer and good performance status, compared with twice-daily radiotherapy, when given concurrently with chemotherapy. Radiotherapy treatment delivery was superior in the twice-daily group. Furthermore, both acute and late toxicities were similar and lower than expected with both regimens. However, although results are unable to show superiority of the once-daily radiotherapy regimen, the CONVERT trial should have a major effect on the standardisation of chemoradiotherapy in this disease group—a treatment that has been the subject of controversy since the publication of the Intergroup 0096 study.4, 13 Overall survival with both regimens were higher than the survival results reported in the Intergroup 0096 study. In CONVERT, 2-year survival for twice-daily and once-daily radiotherapy was 56% and 51%, versus 47% and 41% in the Intergroup 0096 study.4 CONVERT was not an equivalence study (and was not powered for equivalence) so it cannot be concluded that the two regimens have the same efficacy. Furthermore, the 2-year survival of 56% achieved in the control group with twice-daily radiotherapy is the same survival that was projected for the experimental group. The better-than-expected performance of both groups might be explained by several changes in the management of small-cell lung cancer since the publication of the Intergroup study, including PET/CT staging in more than half of patients, the use of modern and precise radiotherapy techniques, and improvements in supportive care. These results, together with several meta-analyses and systematic overviews, support the use of short overall radiotherapy treatment time to avoid early cancer cell repopulation.7, 8, 9, 10, 11 One of the systematic overviews also identified that time from the start of any treatment to completion of radiotherapy is a key variable in predicting outcome.20 Although not significant, 2-year overall survival was slightly higher in the twice-daily group than in the once-daily group, which could possibly be a result of improved delivery of treatment in the twice-daily group, with more patients receiving full-dose radiotherapy, the optimal planned number of fractions, and treatment delivered over the optimal treatment time.

l was slightly higher in the twice-daily group than in the once-daily group, which could possibly be a result of improved delivery of treatment in the twice-daily group, with more patients receiving full-dose radiotherapy, the optimal planned number of fractions, and treatment delivered over the optimal treatment time. Another reason why treatment delivery was superior in the twice-daily group is because of the lower overall dose of radiotherapy in this group, which meant it was possible to achieve the protocol dose constraints for organs at risk, such as lung and spinal cord, in a greater proportion of patients than in the once-daily group. A further advantage of the twice-daily regime is that it halves the radiotherapy treatment time (from 45 days to 19 days) and reduces the number of fractions (from 33 to 30) compared with the once-daily regimen. Although no formal health economic analysis has been done as part of this study, the delivery of twice-daily radiotherapy could lead to cost savings, especially if patients require hospital transport to attend radiotherapy appointments.

duces the number of fractions (from 33 to 30) compared with the once-daily regimen. Although no formal health economic analysis has been done as part of this study, the delivery of twice-daily radiotherapy could lead to cost savings, especially if patients require hospital transport to attend radiotherapy appointments. Overall, the frequency and severity of acute and late radiation toxicities were lower than expected, probably because of the use of modern radiotherapy techniques, including 3D radiotherapy or intensity-modulated radiotherapy, and treatment of involved fields with regard to nodal disease. In the Intergroup 0096 trial,4 patients were treated with outdated radiotherapy techniques including elective nodal irradiation, which would have resulted in higher radiation exposure of normal tissues than in this trial. Indeed, the high rate of severe acute oesophagitis (32% with twice-daily radiotherapy) in the Intergroup study has been cited as the main reason for poor adoption of twice-daily radiotherapy.13 By contrast, less than 20% of patients had severe oesophagitis in the CONVERT study and only one patient developed an oesophageal stricture requiring intervention. Radiation pneumonitis was not specifically reported in the Intergroup 0096 study, but in this trial very few (<3%) patients had severe radiation pneumonitis or severe pulmonary fibrosis. The lower than anticipated toxicity rates and rates of local failure reported in this study suggest that radiotherapy treatment delivered concurrently with chemotherapy could be intensified further—for example, by means of dose escalation or hypofractionation.

severe radiation pneumonitis or severe pulmonary fibrosis. The lower than anticipated toxicity rates and rates of local failure reported in this study suggest that radiotherapy treatment delivered concurrently with chemotherapy could be intensified further—for example, by means of dose escalation or hypofractionation. A limitation of this study is that although we did not mandate an upper age limit—with the aim to gather much needed evidence about the outcome of elderly patients treated with concurrent chemoradiotherapy—only 15% of the patients included were older than 70 years. Data for patients older than 70 years participating in CONVERT was presented at the International Association for the Study of Lung Cancer 17th World Conference on Lung Cancer in Vienna, Austria, in 2016, and the results of this analysis will be presented in a separate report. Elderly patients have been reported to be less likely to receive concurrent chemoradiotherapy than their younger counterparts, which is mainly due to insufficient high-quality evidence to support the use of this potentially toxic treatment.21 Another limitation is that the majority of patients enrolled in both groups were white, and therefore the results of the study might not be applicable to other ethnicities.

han their younger counterparts, which is mainly due to insufficient high-quality evidence to support the use of this potentially toxic treatment.21 Another limitation is that the majority of patients enrolled in both groups were white, and therefore the results of the study might not be applicable to other ethnicities. To our knowledge, CONVERT is the largest study completed investigating thoracic radiotherapy in limited-stage small-cell lung cancer, and the first clinical trial in this group of patients to report on the outcome of patients treated with modern radiotherapy techniques incorporating a quality assurance programme. It was possible to complete this study because of the interest, enthusiasm, and collaborative efforts of a large number of investigators from many different countries. The key to completing accrual was to include a large number of recruiting sites. Furthermore, by contrast with US practice, concurrent chemoradiotherapy is not always adopted as the standard of care for limited-stage small-cell lung cancer in Europe, and the study provided an incentive for centres to adopt and set up concurrent treatment protocols.

was to include a large number of recruiting sites. Furthermore, by contrast with US practice, concurrent chemoradiotherapy is not always adopted as the standard of care for limited-stage small-cell lung cancer in Europe, and the study provided an incentive for centres to adopt and set up concurrent treatment protocols. Given the importance of keeping overall treatment time as short as possible, future studies could investigate dose-escalated twice-daily or hypofractionated radiotherapy concurrently with chemotherapy. Further data for the outcome of patients treated with high-dose 2 Gy per fraction treatment will be provided by the ongoing CALGB 30610/RTOG 0538 study (NCT00632853). The upcoming analysis of the CONVERT translational studies, including the prognostic role of baseline circulating tumour cells, could provide data for relevant biological stratification factors that can be used prospectively in future studies.

t will be provided by the ongoing CALGB 30610/RTOG 0538 study (NCT00632853). The upcoming analysis of the CONVERT translational studies, including the prognostic role of baseline circulating tumour cells, could provide data for relevant biological stratification factors that can be used prospectively in future studies. In conclusion, the results of CONVERT show that there were no significant differences in survival and no major differences in toxicity between twice-daily and once-daily radiotherapy. However, since the trial was designed to show superiority of once-daily radiotherapy and not powered to show equivalence, twice-daily radiotherapy should continue to be considered standard-of-care. Furthermore, twice-daily radiotherapy concurrently with chemotherapy is well tolerated, with better compliance and shorter treatment time than once-daily treatment. From a pragmatic perspective, once-daily radiotherapy could be considered when delivery of twice-daily radiotherapy is impossible because of departmental logistics or patient choice. Supplementary Material Supplementary appendix

In conclusion, the results of CONVERT show that there were no significant differences in survival and no major differences in toxicity between twice-daily and once-daily radiotherapy. However, since the trial was designed to show superiority of once-daily radiotherapy and not powered to show equivalence, twice-daily radiotherapy should continue to be considered standard-of-care. Furthermore, twice-daily radiotherapy concurrently with chemotherapy is well tolerated, with better compliance and shorter treatment time than once-daily treatment. From a pragmatic perspective, once-daily radiotherapy could be considered when delivery of twice-daily radiotherapy is impossible because of departmental logistics or patient choice. Supplementary Material Supplementary appendix Acknowledgments This study was funded by the Cancer Research UK Clinical Trials Awards and Advisory Committee (grant reference number C17052/A8154), French Ministry of Health, Programme Hospitalier de Recherché Clinique (grant reference number NAT 2007-28-01), Canadian Cancer Society Research Institute (grant reference number 021039), and the European Organisation for the Research and Treatment of Cancer (Cancer Research Fund, Lung Cancer, and Radiation Oncology Groups). The authors would like to acknowledge the support of the National Cancer Research Institute Radiotherapy Trials Quality Assurance team (Nicki Groom and Elena Wilson); the Manchester Academic Health Science Centre Clinical Trials Unit; Sally Falk and Amy Bossons (CONVERT trial coordinators); David Ryder (Manchester Academic Health Science Centre Trials Co-ordination Unit statistician); the National Institute for Health Research (NIHR) Christie Clinical Research Facility; David Girling, Steve Roberts, Christian Manegold, and Robert Huddart (Independent Data Monitoring Committee Members); and Dirk De Ruysscher and Jason Lester (Independent Members of the Trial Steering Committee). The CONVERT protocol was developed at the ECCO-AACR-EORTC-ESMO ‘Flims’ Workshop on Methods in Clinical Cancer Research. The views expressed are those of the authors and not necessarily those of the NHS, NIHR, or the Department of Health.

Ruysscher and Jason Lester (Independent Members of the Trial Steering Committee). The CONVERT protocol was developed at the ECCO-AACR-EORTC-ESMO ‘Flims’ Workshop on Methods in Clinical Cancer Research. The views expressed are those of the authors and not necessarily those of the NHS, NIHR, or the Department of Health. Contributors CF-F, LA, PL, MS, FBl, RM, NM, and PJW conceived the study and initiated the study design. WA, FBa, ABh, ABe, FC, PF, SH, CLP, MO'B, JP, VS, and JPVM helped with implementation. CF-F is the grant holder. LA provided statistical expertise in clinical trial design. The authors designed the trial, analysed the data, wrote the manuscript (with the first draft written by the first author), made the decision to submit the manuscript for publication, and assured the completeness and accuracy of the data and analysis and of the fidelity of this report to the trial protocol. All authors approved the final manuscript. Declaration of interests CF-F, LA, PL, and FBl report grants from Cancer Research UK during the conduct of this study. The other authors declare no completing interests.

Valle JW, Wasan H, Lopes A, et al. Cediranib or placebo in combination with cisplatin and gemcitabine chemotherapy for patients with advanced biliary tract cancer (ABC-03): a randomised phase 2 trial. Lancet Oncol 2015; 16: 967–78—In figure 3 of the Appendix of this Article, parts A and B were mislabelled; part A shows overall survival data and part B shows progression-free survival data. The appendix has been corrected as of Aug 31, 2015.

Introduction Neoadjuvant chemotherapy (NACT)—ie, chemotherapy begun before breast cancer surgery—was introduced in the 1970s,1 aiming to downstage locally advanced (inoperable) disease and make it operable. NACT was subsequently extended to operable (early) breast cancer, mainly to allow breast-conserving surgery, and is now widely used, particularly for large tumours.2, 3, 4 Furthermore, NACT might be somewhat more likely to eradicate micrometastatic disease than might chemotherapy delayed until after surgery. NACT might mitigate the hypothesised stimulatory effect of surgery on occult disease5 and reduce tumour cell shedding during surgery. NACT might also provide useful in-vivo information about the chemosensitivity of the local (and, by implication, disseminated) tumour to different chemotherapy regimens, helping to guide subsequent drug selection.6, 7 Conversely, by delaying surgery, NACT might increase the risk of metastatic spread, particularly for chemoresistant tumours.

e useful in-vivo information about the chemosensitivity of the local (and, by implication, disseminated) tumour to different chemotherapy regimens, helping to guide subsequent drug selection.6, 7 Conversely, by delaying surgery, NACT might increase the risk of metastatic spread, particularly for chemoresistant tumours. Several randomised trials8, 9, 10, 11, 12, 13, 14, 15, 16, 17 have compared NACT with the same chemotherapy given postoperatively. Interpretation of these trials is complicated, however, as the frequency of breast-conserving surgery often differed between groups because of tumour shrinkage after NACT. In certain trials,14, 15 some good responders to NACT did not receive surgery, and high frequencies of local recurrence with NACT in these trials have been attributed to omission of definitive local therapy. Any such differences in the extent of surgery confound comparisons of the efficacy of NACT with that of adjuvant chemotherapy.18, 19 Another complication is that investigations of the influence of tumour characteristics on outcome need to use prerandomisation data, as analyses by postsurgical characteristics would be substantially biased by downstaging.20 To investigate such issues in more detail than was possible in reviews18, 19 of published data, we did a patient-level meta-analysis of the trials that directly compared any NACT regimen with the same regimen begun postoperatively. Research in context Evidence before this study

Several randomised trials8, 9, 10, 11, 12, 13, 14, 15, 16, 17 have compared NACT with the same chemotherapy given postoperatively. Interpretation of these trials is complicated, however, as the frequency of breast-conserving surgery often differed between groups because of tumour shrinkage after NACT. In certain trials,14, 15 some good responders to NACT did not receive surgery, and high frequencies of local recurrence with NACT in these trials have been attributed to omission of definitive local therapy. Any such differences in the extent of surgery confound comparisons of the efficacy of NACT with that of adjuvant chemotherapy.18, 19 Another complication is that investigations of the influence of tumour characteristics on outcome need to use prerandomisation data, as analyses by postsurgical characteristics would be substantially biased by downstaging.20 To investigate such issues in more detail than was possible in reviews18, 19 of published data, we did a patient-level meta-analysis of the trials that directly compared any NACT regimen with the same regimen begun postoperatively. Research in context Evidence before this study The Early Breast Cancer Trialists' Collaborative Group's ongoing extensive searches of bibliographic databases, including MEDLINE, Embase, the Cochrane Library, and meeting abstracts up to March 2017, identified 16 trials that compared neoadjuvant chemotherapy (NACT) with the same chemotherapy postoperatively. Meta-analyses of published reports indicate that NACT reduced the frequency of mastectomy but did not affect mortality. Interpretation is complicated, however, as the use of breast-conserving surgery often differed between groups because of tumour shrinkage by NACT. In certain trials, some patients with a good response did not receive surgery. Hence, women allocated NACT retained more breast tissue than did those allocated adjuvant chemotherapy, and higher local recurrence frequencies in some neoadjuvant trials than in others have been attributed to omission of definitive local therapy.

tain trials, some patients with a good response did not receive surgery. Hence, women allocated NACT retained more breast tissue than did those allocated adjuvant chemotherapy, and higher local recurrence frequencies in some neoadjuvant trials than in others have been attributed to omission of definitive local therapy. Added value of this study We did a meta-analysis of individual patient data from trials that compared NACT with the same chemotherapy given postoperatively. We assessed effects of patient and tumour characteristics on tumour response, extent of local therapy, local and distant recurrence, breast cancer death, and overall mortality. This individual patient data meta-analysis, involving 4756 women in ten trials, found that the frequencies of clinical response and breast-conserving therapy were higher for smaller, higher-grade, and oestrogen receptor-negative and progesterone receptor-negative tumours, and for one trial using anthracycline and taxane chemotherapy. Although responders to NACT had lower distant recurrence and breast cancer mortality than did non-responders, when responders and non-responders were combined, distant recurrence and breast cancer mortality were similar for NACT and adjuvant chemotherapy. Local recurrence was, however, higher with NACT than with adjuvant chemotherapy, which persisted for 10 years after treatment and was not confined to trials in which surgery could be omitted after response to NACT. Implications of all the available evidence

We did a meta-analysis of individual patient data from trials that compared NACT with the same chemotherapy given postoperatively. We assessed effects of patient and tumour characteristics on tumour response, extent of local therapy, local and distant recurrence, breast cancer death, and overall mortality. This individual patient data meta-analysis, involving 4756 women in ten trials, found that the frequencies of clinical response and breast-conserving therapy were higher for smaller, higher-grade, and oestrogen receptor-negative and progesterone receptor-negative tumours, and for one trial using anthracycline and taxane chemotherapy. Although responders to NACT had lower distant recurrence and breast cancer mortality than did non-responders, when responders and non-responders were combined, distant recurrence and breast cancer mortality were similar for NACT and adjuvant chemotherapy. Local recurrence was, however, higher with NACT than with adjuvant chemotherapy, which persisted for 10 years after treatment and was not confined to trials in which surgery could be omitted after response to NACT. Implications of all the available evidence NACT is as effective as adjuvant chemotherapy in reducing the risk of distant recurrence and death from breast cancer. However, NACT is associated with higher local recurrence than adjuvant chemotherapy, which could be at least partly explained by wider use of breast-conserving therapy after NACT than with postoperative chemotherapy. Strategies to mitigate the increased local recurrence after breast-conserving therapy in tumours downsized by NACT should be considered—eg, careful tumour localisation, detailed pathological assessment, and appropriate radiotherapy.

wider use of breast-conserving therapy after NACT than with postoperative chemotherapy. Strategies to mitigate the increased local recurrence after breast-conserving therapy in tumours downsized by NACT should be considered—eg, careful tumour localisation, detailed pathological assessment, and appropriate radiotherapy. Methods Study design and participants We sought data from all randomised trials in early (ie, operable) breast cancer that began before 2005 and compared NACT with the same chemotherapy begun after surgery (ie, standard adjuvant chemotherapy). NACT always started before surgery, although in some trials, some of the chemotherapy in the NACT group was given postoperatively, whereas all chemotherapy in the control group had to be postoperative (so trials such as NSABP B-2721 were ineligible). Trial identification and data checking were as reported previously22, 23 and conformed to the Preferred Reporting Items for Systematic Review and Meta-Analyses (Individual Patient Data).24 For every woman, we requested information from the trial's principal investigator or another appropriate member of their research group about patient and tumour characteristics, treatments, dates of any local recurrence (breast, chest wall, or regional nodes), distant recurrence, contralateral breast or other second primary cancer, and date last known to be alive or date and underlying cause of death. To avoid bias, we sought data for tumour characteristics recorded before randomisation since these characteristics can be altered by neoadjuvant treatment. We also requested tumour response after NACT (assessed mostly by palpation and mammography). To investigate the influence of NACT on extent of surgery, we sought details of surgery planned at randomisation and surgery actually done. When planned surgery was unknown, it was inferred from clinical tumour size. Patient-level data for radiotherapy were unavailable.

T (assessed mostly by palpation and mammography). To investigate the influence of NACT on extent of surgery, we sought details of surgery planned at randomisation and surgery actually done. When planned surgery was unknown, it was inferred from clinical tumour size. Patient-level data for radiotherapy were unavailable. Statistical analysis A detailed description of the statistical methods has been previously published.22 Primary outcomes assessed were tumour response (complete response [no clinical evidence of disease after NACT], partial response [≥50% reduction in initial size], or stable or progressive disease [<50% reduction, no change, or increased tumour size]), extent of local therapy (mastectomy, lumpectomy [either with or without radiotherapy], and radiotherapy alone), local and distant recurrence, breast cancer death (via subtraction of the log-rank statistics of death without recurrence from those of overall survival23), and overall mortality (death from any cause). Analyses were by intention to treat and are of first isolated local recurrence (site not generally available), any distant recurrence (irrespective of previous local or contralateral recurrence), breast cancer mortality, and all-cause mortality. We treated deaths from new cancers of unknown primary sites as breast cancer deaths. When no recurrence was reported before breast cancer death, we assumed distant recurrence to have just preceded it. We took deaths from an unknown cause without recorded recurrence to be non-breast cancer deaths. Comparisons of NACT response and frequency of breast-conserving therapy used regression models,25 accounting for tumour size and trial. We stratified log-rank analyses by trial, follow-up year, age at entry (<35 years, 35–44 years, 45–54 years, 55–69 years, and ≥70 years), and prerandomisation clinical nodal status (N0 or other).

of NACT response and frequency of breast-conserving therapy used regression models,25 accounting for tumour size and trial. We stratified log-rank analyses by trial, follow-up year, age at entry (<35 years, 35–44 years, 45–54 years, 55–69 years, and ≥70 years), and prerandomisation clinical nodal status (N0 or other). In such analyses, if a log-rank statistic (o–e) has variance v, then, defining z=(o–e)/√v and b=(o–e)/v, the event rate ratio (RR; NACT vs control) is estimated as exp(b) with SE=(RR–1)/z. RRs and confidence limits for RR are derived from those for b (by normal approximations). To test for a trend between n strata (eg, of age) in the effects of treatment, we supposed that stratum number s (s=1,2,…,n) has log-rank statistics (o–e) and v (with grand total over all strata O–E and V). We defined m, the mean stratum number, to be the sum, one term per stratum, of sv/V and define T to be the sum, one term per stratum, of (s–m)(o–e).26 The variance of T, var(T), is then the sum, one term per stratum, of (s–m)2/v. The trend test statistic (ie, the change from one stratum to the next in the log of the event RR) is then T/var(T), which has variance 1/var(T). Tests of whether two trends are the same involve subtraction of the corresponding trend test statistics from each other. A χ2 statistic on one degree of freedom (χ21) for testing of whether some quantity Q differs significantly from zero is given by Q2/var(Q). A χ2 test (on n–1 degrees of freedom) for heterogeneity can be obtained by subtracting (O–E)2/V from the sum of the separate values, one per stratum, of (o–e)2/v. For analyses by regression, we estimated RRs by maximum likelihood; tests for trend and heterogeneity were by likelihood ratio.

m zero is given by Q2/var(Q). A χ2 test (on n–1 degrees of freedom) for heterogeneity can be obtained by subtracting (O–E)2/V from the sum of the separate values, one per stratum, of (o–e)2/v. For analyses by regression, we estimated RRs by maximum likelihood; tests for trend and heterogeneity were by likelihood ratio. Associations between baseline variables and outcome used prerandomisation values. Only two trials provided pathological response data, so correlations of characteristics with response to NACT use clinical response data (available for eight of ten trials). Subgroup analyses compare outcomes in trials in which all women allocated NACT were, or were not, scheduled to receive breast surgery and in women whose initially planned local treatment was mastectomy or breast-conserving therapy (lumpectomy with or without radiotherapy or radiotherapy alone). Sensitivity analyses assess the potential effect27 on local recurrence of competing events (distant recurrence and death without recurrence) and of omission of trials with only first recurrences recorded. p values of 0·05 or less are described as significant. Analyses used Stata 13.1 and R 2.13.2. Data sharing Procedures for data access are available online. Role of the funding source The funders of the study had no role in study design, data analysis, data interpretation, or writing of the report. The secretariat had full access to all the data in the study. The writing committee had final responsibility for the decision to submit for publication.

Data sharing Procedures for data access are available online. Role of the funding source The funders of the study had no role in study design, data analysis, data interpretation, or writing of the report. The secretariat had full access to all the data in the study. The writing committee had final responsibility for the decision to submit for publication. Results Individual patient data were available from ten8, 9, 10, 11, 12, 13, 14, 15, 16, 17 of 16 eligible trials identified and from 4756 (91%) of the 5250 women in total (table 1, appendix pp 2, 6, 17). Trial entry year for participants was 1983–2002, median follow-up was 9 years (IQR 5–14), with the last follow-up in 2013, and median age was 49 years (43–57). 1604 deaths occurred, including 248 (15%) without recurrence. Of the 4756 women included in the analysis, 3838 (81%) were in trials of regimens that included an anthracycline, one of which (902 women) also gave a taxane.13 Four trials (918 women) used MMM (mitoxantrone, methotrexate, and mitomycin-C)11, 12 or CMF (cyclophosphamide, methotrexate, and fluorouracil)8, 17 as NACT; in these trials, some chemotherapy in those allocated NACT was given after surgery (table 1, appendix pp 3–4). No patients received trastuzumab.Table 1 Trials of neoadjuvant versus adjuvant chemotherapy that began by 2005

te, and mitomycin-C)11, 12 or CMF (cyclophosphamide, methotrexate, and fluorouracil)8, 17 as NACT; in these trials, some chemotherapy in those allocated NACT was given after surgery (table 1, appendix pp 3–4). No patients received trastuzumab.Table 1 Trials of neoadjuvant versus adjuvant chemotherapy that began by 2005 Trials (n)* Women (n) Deaths (n)† Median years per woman (IQR) Woman-years by years since entry (thousands) <10 10–19 ≥20 Total No anthracycline or taxane8, 11, 12, 17‡ 4 918 315 7·0 (4·2–9·3) 6·0 0·8 0·2 7·0 Anthracycline, no taxane9, 10, 14, 15, 16 5 2936 1163 10·2 (4·9–15·4) 22·1 7·7 <0·1 29·8 Anthracycline and taxane13 1 902 126 7·9 (5·0–10·7) 6·5 0·5 0 7·0 Total 10 4756 1604 8·6 (4·8–13·7) 34·6 9·0 0·2 43·7 * Data are missing for six small trials that randomised about 500 women, so they were not included in this analysis (appendix p 17). † Includes 1356 deaths with recurrence, 72 of unknown cause without recurrence, and 176 of known cause without recurrence. ‡ In these trials, women allocated to the neoadjuvant group completed their chemotherapy after surgery.

Trials (n)* Women (n) Deaths (n)† Median years per woman (IQR) Woman-years by years since entry (thousands) <10 10–19 ≥20 Total No anthracycline or taxane8, 11, 12, 17‡ 4 918 315 7·0 (4·2–9·3) 6·0 0·8 0·2 7·0 Anthracycline, no taxane9, 10, 14, 15, 16 5 2936 1163 10·2 (4·9–15·4) 22·1 7·7 <0·1 29·8 Anthracycline and taxane13 1 902 126 7·9 (5·0–10·7) 6·5 0·5 0 7·0 Total 10 4756 1604 8·6 (4·8–13·7) 34·6 9·0 0·2 43·7 * Data are missing for six small trials that randomised about 500 women, so they were not included in this analysis (appendix p 17). † Includes 1356 deaths with recurrence, 72 of unknown cause without recurrence, and 176 of known cause without recurrence. ‡ In these trials, women allocated to the neoadjuvant group completed their chemotherapy after surgery. Across all trials, NACT was associated with substantial tumour response (table 2), moderately increased use of breast-conserving therapy in the NACT group compared with the adjuvant chemotherapy group (figure 1), and an absolute increase in 15 year local recurrence of 5·5% (95% CI 2·4–8·6; 21·4% for NACT vs 15·9% for adjuvant chemotherapy), corresponding to a RR of 1·37 (95% CI 1·17–1·61; p=0·0001; figure 2A). The incidence of local recurrence was significantly higher with NACT than with adjuvant chemotherapy in years 0–4 (RR 1·35 [95% CI 1·11–1·64]; p=0·003) and 5–9 (1·53 [1·08–2·17]; p=0·02), with few local recurrences after year 10. Sensitivity analyses indicated no substantial influence of competing risks from other breast events on the RRs for local recurrence (appendix p 18).Figure 1 BCT rate ratios

uvant chemotherapy in years 0–4 (RR 1·35 [95% CI 1·11–1·64]; p=0·003) and 5–9 (1·53 [1·08–2·17]; p=0·02), with few local recurrences after year 10. Sensitivity analyses indicated no substantial influence of competing risks from other breast events on the RRs for local recurrence (appendix p 18).Figure 1 BCT rate ratios Numbers with BCT or mastectomy after chemotherapy. Excludes local therapy unknown (67 patients with NACT and 51 with adjuvant chemotherapies). BCT=breast-conserving therapy. ER=oestrogen receptor. PR=progesterone receptor. NACT=neoadjuvant chemotherapy. Figure 2 Effect of neoadjuvant versus adjuvant chemotherapy on recurrence and mortality Local recurrence (A), distant recurrence (B), breast cancer mortality (C), and death from any cause (D). Three trials recorded causes of any deaths but only the first breast cancer event. Hence, for these trials, distant recurrence includes the first distant recurrence as the first event and death from breast cancer. Error bars are 95% CIs. NACT=neoadjuvant chemotherapy. O–E=observed minus expected. RR=rate ratio. V=variance of O–E. Table 2 Local therapy, planned versus done, in women allocated to neoadjuvant chemotherapy, by clinical response

Local recurrence (A), distant recurrence (B), breast cancer mortality (C), and death from any cause (D). Three trials recorded causes of any deaths but only the first breast cancer event. Hence, for these trials, distant recurrence includes the first distant recurrence as the first event and death from breast cancer. Error bars are 95% CIs. NACT=neoadjuvant chemotherapy. O–E=observed minus expected. RR=rate ratio. V=variance of O–E. Table 2 Local therapy, planned versus done, in women allocated to neoadjuvant chemotherapy, by clinical response Therapy done Clinical response Complete* Partial† Stable or progressive disease‡ Unknown Total Planned breast-conserving therapy Breast-conserving 215 (96%) 256 (90%) 119 (77%) 211 (81%) 801 (87%) Mastectomy 10 (4%) 30 (10%) 35 (23%) 48 (19%) 123 (13%) Unknown 0 0 0 2 (NA) 2 (NA) Total response§ 225/665 (34%) 286/665 (43%) 154/665 (23%) 261 (NA) 926 (100%) Planned mastectomy Breast-conserving 75 (60%) 121 (41%) 30 (12%) 26 (36%) 252 (33%) Mastectomy 49 (40%) 175 (59%) 231 (88%) 47 (64%) 502 (67%) Unknown 0 1 (NA) 2 (NA) 11 (NA) 14 (NA) Total response§ 124/684 (18%) 297/684 (43%) 263/684 (38%) 84 (NA) 768 (100%) Unknown planned therapy Breast-conserving 162 (83%) 164 (76%) 97 (56%) 28 (49%) 451 (70%) Mastectomy 33 (17%) 53 (24%) 76 (44%) 29 (51%) 191 (30%) Unknown 2 (NA) 3 (NA) 8 (NA) 38 (NA) 51 (NA) Total response§ 197/598 (33%) 220/598 (37%) 181/598 (30%) 95 (NA) 693 (100%) All women Breast-conserving 452 (83%) 541 (68%) 246 (42%) 265 (68%) 1504 (65%) Mastectomy 92 (17%) 258 (32%) 342 (58%) 124 (32%) 816 (35%) Unknown 2 (NA) 4 (NA) 10 (NA) 51 (NA) 67 (NA) Total response§ 546/1947 (28%) 803/1947 (41%) 598/1947 (31%) 440 (NA) 2387 (100%) Data are n (%) or n/N (%). NA=not applicable.

(100%) All women Breast-conserving 452 (83%) 541 (68%) 246 (42%) 265 (68%) 1504 (65%) Mastectomy 92 (17%) 258 (32%) 342 (58%) 124 (32%) 816 (35%) Unknown 2 (NA) 4 (NA) 10 (NA) 51 (NA) 67 (NA) Total response§ 546/1947 (28%) 803/1947 (41%) 598/1947 (31%) 440 (NA) 2387 (100%) Data are n (%) or n/N (%). NA=not applicable. * No clinical evidence of disease. † ≥50% reduction in tumour size. ‡ <50% reduction or increase in tumour size. § Percentages are of those with a known response. As anticipated,18, 19 the absolute increase in 10-year local recurrence with NACT was largest in the two trials14, 15 in which, after NACT, many women did not have breast surgery (13·3% [95% CI 5·5–21·1]; 33·7% for NACT vs 20·4% for adjuvant chemotherapy; RR 1·62 [95% CI 1·20–2·19], p=0·002; figure 3B). In the other eight trials,8, 9, 10, 11, 12, 13, 16, 17 surgery was scheduled irrespective of response to NACT, and the absolute increase in 10 year local recurrence was 3·2% (95% CI 0·6–5·8; 15·1% vs 11·9%; RR 1·28 [95% CI 1·06–1·55], p=0·01; figure 3A). However, the RRs for local recurrence in these two sets of trials were not significantly different (heterogeneity p=0·19).Figure 3 Time to recurrence and breast cancer mortality

nd the absolute increase in 10 year local recurrence was 3·2% (95% CI 0·6–5·8; 15·1% vs 11·9%; RR 1·28 [95% CI 1·06–1·55], p=0·01; figure 3A). However, the RRs for local recurrence in these two sets of trials were not significantly different (heterogeneity p=0·19).Figure 3 Time to recurrence and breast cancer mortality Local recurrence for surgery commonly used (A) and less commonly used (B), distant recurrence for surgery commonly used (C) and less commonly used (D), and breast cancer mortality for surgery commonly used (E) and less commonly used (F). Heterogeneity by surgery use: local recurrence p=0·19, distant recurrence p=0·29, and breast cancer mortality p=0·24. Error bars are 95% CIs. NACT=neoadjuvant chemotherapy. O–E=observed minus expected. RR=rate ratio. V=variance of O–E. Three trials recorded causes of any deaths but only the first breast cancer event. Hence, for these trials, distant recurrence includes the first distant recurrence as the first event and death from breast cancer. *Includes Institut Bergonié Bordeaux14 (in NACT group, 33% had radiotherapy alone) and Institut Curie S615 (in NACT group, 51% had radiotherapy alone; in adjuvant chemotherapy group, 46% had radiotherapy alone) trials.

ant recurrence includes the first distant recurrence as the first event and death from breast cancer. *Includes Institut Bergonié Bordeaux14 (in NACT group, 33% had radiotherapy alone) and Institut Curie S615 (in NACT group, 51% had radiotherapy alone; in adjuvant chemotherapy group, 46% had radiotherapy alone) trials. Between-trial RRs for local recurrence ranged from 0·67 (95% CI 0·24–1·91) to 4·59 (1·19–17·8), but this apparent heterogeneity was not significant (χ210=11·8; p=0·30; figure 4A). RRs for local recurrence also did not differ significantly between the three classes of chemotherapy used in these trials (figure 4, appendix p 13), between trials in which chemotherapy in the NACT group was or was not completed after local therapy (appendix p 13), or between use or not of tamoxifen (figure 5, appendix p 13).Figure 4 Rate ratios for the effect of neoadjuvant versus adjuvant chemotherapy on recurrence by trial

these trials (figure 4, appendix p 13), between trials in which chemotherapy in the NACT group was or was not completed after local therapy (appendix p 13), or between use or not of tamoxifen (figure 5, appendix p 13).Figure 4 Rate ratios for the effect of neoadjuvant versus adjuvant chemotherapy on recurrence by trial (A) Local recurrence. (B) Distant recurrence. Three trials recorded causes of any deaths but only the first breast cancer event. Hence, for these trials, distant recurrence includes the first distant recurrence as the first event and death from breast cancer. The appendix (pp 3–4) contains a full description of each trial's chemotherapy regimen. A=doxorubicin (adriamycin). BCCA=British Columbia Cancer Agency. BCSG=Breast Cancer Study Group. C=cyclophosphamide. E=epirubicin. ECTO=European Cooperative Trial in Operable Breast Cancer. EORTC=European Organisation for Research and Treatment of Cancer. F=fluorouracil. Fol=folinic acid. IB=Institut Bergonié. M=methotrexate. Mit=mitomycin-C. Mz=mitoxantrone. NCI=National Cancer Institute. NSABP=National Surgical Adjuvant Breast and Bowel Project. O–E=observed minus expected. P=paclitaxel. Tt=thiotepa. Vc=vincristine. Vd=vindesine. *Chemotherapy regimens given preoperatively in those allocated neoadjuvant and postoperatively in those allocated adjuvant chemotherapy. The number of cycles, agents, and drug doses (in mg/m2) per cycle are given. †The Austrian BCSG VII trial8 has two entries to take into account the two postoperative chemotherapies given to both randomised groups (appendix pp 3–4).

hose allocated neoadjuvant and postoperatively in those allocated adjuvant chemotherapy. The number of cycles, agents, and drug doses (in mg/m2) per cycle are given. †The Austrian BCSG VII trial8 has two entries to take into account the two postoperative chemotherapies given to both randomised groups (appendix pp 3–4). Figure 5 Local recurrence rate ratios For lumpectomy versus mastectomy, χ21=3·3; p=0·07. ER=oestrogen receptor. PR=progesterone receptor. *408 women with missing data had planned local therapy imputed (appendix p 9). †Refers to Institut Curie S615 (appendix p 9). We noted no significant differences between NACT and adjuvant treatment in 15 year distant recurrence (38·2% for NACT vs 38·0% for adjuvant chemotherapy; RR 1·02 [95% CI 0·92–1·14]; p=0·66), breast cancer death (34·4% vs 33·7%; 1·06 [0·95–1·18]; p=0·31), or death from any cause (40·9% vs 41·2%; 1·04 [0·94–1·15]; p=0·45; figure 2B, C, D). The RRs for these three outcomes did not differ significantly between any subgroups of trials, including those for which use of surgery was or was not dependent on response to NACT, those using different types of chemotherapy, or those using or not using tamoxifen (figure 3 C–F and figure 4B, appendix pp 7, 13). Three trials8, 9, 16 collected only first recurrence and death rather than all events; however, sensitivity analyses omitting these trials had no material effect on distant recurrence estimates (appendix p 18). Mortality from causes other than breast cancer was no different between the NACT and adjuvant chemotherapy groups (appendix p 7).

collected only first recurrence and death rather than all events; however, sensitivity analyses omitting these trials had no material effect on distant recurrence estimates (appendix p 18). Mortality from causes other than breast cancer was no different between the NACT and adjuvant chemotherapy groups (appendix p 7). Information about clinical tumour response was available for 1947 (82%) of 2387 patients allocated NACT; 546 (28%) of 1947 had a complete response, 803 (41%) of 1947 had a partial response, and 598 (31%) of 1947 had stable or progressive disease (table 2, appendix p 8). The clinical tumour response to NACT affected surgical treatment decisions: more women with a complete response had breast-conserving therapy (452 [83%] of 544) than did those with a partial response (541 [68%] of 799) or no response (246 [42%] of 588). Consequently, we noted an imbalance by treatment group in the extent of surgery: although breast-conserving therapy was initially intended for equal numbers of patients in each group, actual use of breast-preserving therapy (including no surgery) was 1504 (65%) of 2320 in the NACT group versus 1135 (49%) of 2318 in the adjuvant chemotherapy group excluding patients with unknown surgeries (p<0·0001; figure 1, appendix pp 9, 10).

rapy was initially intended for equal numbers of patients in each group, actual use of breast-preserving therapy (including no surgery) was 1504 (65%) of 2320 in the NACT group versus 1135 (49%) of 2318 in the adjuvant chemotherapy group excluding patients with unknown surgeries (p<0·0001; figure 1, appendix pp 9, 10). Figure 6 shows proportions of women with complete clinical response according to patient and tumour characteristics. Complete response decreased with increasing clinical tumour size (trend p<0·0001) and was higher with oestrogen receptor (ER)-negative biopsies than with ER-positive biopsies (p<0·0001) and with poorly differentiated tumours than with well or moderately differentiated tumours (trend p=0·001, even after allowance for high-grade tumours tending to be ER negative), and was higher in the one trial13 that combined anthracycline and taxane therapy than in the other trials (p<0·0001). Age, nodal status, and planned local therapy did not affect response.Figure 6 Clinical complete response rate ratios Three trials are excluded, as individual responses are not available; 440 women have missing clinical response data. CIs are group specific.25 Rate ratios are scaled such that, within each category, their inverse variance-weighted sum is 1—ie, ratios are with respect to the mean CR. The appendix (p 6) contains data available for each trial. CR=complete response. ER=oestrogen receptor. PR=progesterone receptor.

clinical response data. CIs are group specific.25 Rate ratios are scaled such that, within each category, their inverse variance-weighted sum is 1—ie, ratios are with respect to the mean CR. The appendix (p 6) contains data available for each trial. CR=complete response. ER=oestrogen receptor. PR=progesterone receptor. The proportion of women having breast-conserving therapy in various different subgroups in the NACT and adjuvant chemotherapy groups are shown in figure 1. The strongest predictors of the effect of NACT on breast conservation frequency were tumour size, planned local therapy, and type of chemotherapy (all p<0·0001). The effect of NACT on surgery de-escalation was most apparent among women with large (20–49 mm or ≥50 mm) tumours; we noted little effect of NACT on breast conservation frequency in women with small (<20 mm) tumours. As expected, women with mastectomy originally planned were more likely to have lesser surgery than were those with breast-conserving surgery originally planned. NACT with an anthracycline and taxane combination was also associated with substantially more surgery de-escalation than was NACT with other regimens. In women with node-positive disease, the rate ratio for BCT was higher than for those with node-negative disease (p=0·01). Despite the high frequency of clinical response in patients with ER-negative and poorly differentiated tumours, ER status and tumour grade were not associated with frequency of breast-conserving therapy, although after accounting for grade, ER-negative women did appear to have higher breast-conserving frequencies. Age was not associated with the freqeuncy of breast-conserving therapy.

-negative and poorly differentiated tumours, ER status and tumour grade were not associated with frequency of breast-conserving therapy, although after accounting for grade, ER-negative women did appear to have higher breast-conserving frequencies. Age was not associated with the freqeuncy of breast-conserving therapy. If the increased local recurrence in the NACT groups (figure 3) is due to de-escalation of local therapy from mastectomy to breast-conserving therapy, the RR for the effect on local recurrence of allocation to NACT should be greatest in women for whom mastectomy was originally planned. 252 (33%) of 754 women converted from planned mastectomy to breast-conserving therapy. However, although the RR for local recurrence among all women planned to have a mastectomy was 1·66 (95% CI 1·24–2·21) compared with 1·14 (0·86–1·52) for women with lumpectomy planned, the two-tailed test for heterogeneity was not significant (p=0·07; figure 5, appendix pp 11, 12).

tectomy to breast-conserving therapy. However, although the RR for local recurrence among all women planned to have a mastectomy was 1·66 (95% CI 1·24–2·21) compared with 1·14 (0·86–1·52) for women with lumpectomy planned, the two-tailed test for heterogeneity was not significant (p=0·07; figure 5, appendix pp 11, 12). Heterogeneity between the RRs for local recurrence was lower across all other tumour characteristics than it was for RRs in the subgroup by planned local therapy (figure 5); although the p value for the trend in RR with biopsy grade was 0·05, this p value could have been a chance finding given that it was the most extreme from many subgroup analyses. Despite surgery de-escalation being more common in larger tumours than in smaller tumours, and in the trial combining anthracycline and taxane13 than in trials of other regimens, the proportional increases in local recurrence did not vary significantly by tumour size or chemotherapy regimen (figure 5). RRs also did not differ by age, nodal status, ER or progesterone receptor status, or period of follow-up. Patient-level data for radiotherapy were not available, and trial-level data for radiotherapy intent and practice were incomplete (appendix p 2), so the effect of radiotherapy on local recurrence cannot be studied. Radiotherapy was scheduled for most women who had breast-conserving surgery and actual use of radiotherapy was more frequent in the NACT than in the adjuvant therapy groups (appendix p 2). The RRs for distant recurrence and breast cancer mortality did not vary by any tumour factor measured, type of chemotherapy, timing of chemotherapy use in the NACT group, type of planned local therapy, or period of follow-up (appendix p 13).

more frequent in the NACT than in the adjuvant therapy groups (appendix p 2). The RRs for distant recurrence and breast cancer mortality did not vary by any tumour factor measured, type of chemotherapy, timing of chemotherapy use in the NACT group, type of planned local therapy, or period of follow-up (appendix p 13). As expected, distant recurrence and breast cancer mortality were substantially lower in complete responders than in non-responders (appendix p 14). However, women who had a complete clinical response after NACT had a frequency of local recurrence similar to that of partial responders or non-responders. Ordering of trials by the percentage of patients with a complete clinical response to NACT did not reveal any significant trend of improved recurrence or breast cancer mortality RRs in trials with a higher frequency of response (appendix p 16). No patterns emerged between trials in complete response when considering the year that the trial started or the frequency of breast-conserving therapy within a trial (appendix p 15). Discussion In early breast cancer, high frequencies of complete or partial clinical response can be achieved with NACT, which can lead to a higher frequency of breast-conserving therapy than with adjuvant chemotherapy. However, we found NACT to be associated with a higher frequency of local recurrence than was the same chemotherapy started after surgery. Reassuringly, the increase in local recurrence was not associated with any significant increase in distant recurrence or breast cancer mortality.

apy than with adjuvant chemotherapy. However, we found NACT to be associated with a higher frequency of local recurrence than was the same chemotherapy started after surgery. Reassuringly, the increase in local recurrence was not associated with any significant increase in distant recurrence or breast cancer mortality. More than two thirds of patients receiving NACT responded, with more than a quarter achieving complete clinical response, despite some trials using old chemotherapy regimens and four administering some of the chemotherapy postoperatively. The one regimen that included both anthracycline and taxane had the highest frequency of complete response. Within trials, response was more common in women with small, ER-negative and progesterone receptor-negative, or high-grade tumours, as measured before randomisation, but was little affected by age, nodal status, or planned local therapy.8, 9, 10, 11, 12, 13, 14, 15, 16, 17 As expected, use of NACT was associated with an increase in the use of breast-conserving therapy.

th small, ER-negative and progesterone receptor-negative, or high-grade tumours, as measured before randomisation, but was little affected by age, nodal status, or planned local therapy.8, 9, 10, 11, 12, 13, 14, 15, 16, 17 As expected, use of NACT was associated with an increase in the use of breast-conserving therapy. An increase in the use of breast-conserving therapy in women who responded well to NACT and who would otherwise have had mastectomy is a likely explanation for the increase in local recurrence in patients allocated NACT. As anticipated,18, 19 the absolute increase in local recurrence was greatest in the two trials14, 15 in which surgery could be avoided completely in the event of a complete clinical response to NACT. This apparent heterogeneity of effect was, however, not significant, and NACT appeared to also have resulted in some increase in local recurrence in the aggregated results from the eight other trials. Hence, the increased local recurrence with NACT is not wholly explained by omission of surgery. Other unexamined factors might also have contributed to the increased local recurrence with NACT. For example, after NACT, tumour localisation can be difficult28 and response patterns can be heterogeneous,29 making surgery technically more difficult than without use of NACT. Differing use of radiotherapy or axillary surgery in the NACT group might also have contributed to the higher local recurrence, although patient-level information about this factor was not available. Trial reports indicate that radiotherapy was scheduled for most women who had breast-conserving surgery and that actual use of radiotherapy was, if anything, more frequent in the NACT than in the adjuvant therapy groups. Thus, lesser use of radiotherapy after NACT than that without NACT is unlikely to explain this increase in local recurrence. Indeed, even with radiotherapy, local failure is higher after breast-conserving surgery than after mastectomy without radiotherapy.30

ore frequent in the NACT than in the adjuvant therapy groups. Thus, lesser use of radiotherapy after NACT than that without NACT is unlikely to explain this increase in local recurrence. Indeed, even with radiotherapy, local failure is higher after breast-conserving surgery than after mastectomy without radiotherapy.30 Our finding of an overall increase in local recurrence in the trials using optimal local treatment is at odds with a meta-analysis18 based on published data rather than individual patient data, but this discrepancy could be because the meta-analysis included comparisons that were confounded by differing background systemic therapy.

f an overall increase in local recurrence in the trials using optimal local treatment is at odds with a meta-analysis18 based on published data rather than individual patient data, but this discrepancy could be because the meta-analysis included comparisons that were confounded by differing background systemic therapy. Tumour response is predictive of lower distant recurrence and death than an absence of tumour response. Compared with all women randomly allocated to adjuvant chemotherapy, outcomes were better for those with a complete clinical response after NACT than for those with a partial response and far better than for those with little or no response to NACT. However, even in trials with high frequencies of complete response, NACT was not significantly better than adjuvant chemotherapy with respect to distant recurrence or breast cancer mortality. This finding could be because tumour characteristics that are associated with higher response—such as smaller tumour size—are also associated with lower distant recurrence and are balanced between the NACT and adjuvant groups by randomisation. In each trial, both groups eventually receive the same chemotherapy, so any differences between trials in the efficacy of chemotherapy regimens will apply to both groups. The CTNeoBC study6 reported similar findings in that efficacy as assessed by high pathological complete response at the trial level did not correlate well with long-term efficacy.

ntually receive the same chemotherapy, so any differences between trials in the efficacy of chemotherapy regimens will apply to both groups. The CTNeoBC study6 reported similar findings in that efficacy as assessed by high pathological complete response at the trial level did not correlate well with long-term efficacy. A limitation of our meta-analysis is that we have not been able to assess reliably whether presurgical systemic therapy is more effective at eradicating micrometastatic disease than the same chemotherapy administered after recovery from surgery because of the confounding effect of differences in the extent of surgery between women allocated NACT and those allocated adjuvant chemotherapy. A large trial with the same surgery and radiotherapy in the NACT and adjuvant chemotherapy groups could assess this question. At present, we cannot exclude the possibility that NACT does moderately reduce distant recurrence compared with the same chemotherapy given postoperatively, but that this benefit was obscured by an increase in local recurrence due to less extensive surgery after NACT than in patients who did not receive NACT. Trials of radiotherapy after surgery indicate that substantially decreasing local recurrence does also decrease breast cancer mortality, with about one breast cancer death prevented for every four local recurrences prevented.30 The small, non-significant excess of breast cancer mortality in patients allocated NACT is consistent with this risk ratio, so could therefore be due to the increase in local recurrence, but it could equally well be a chance finding.

about one breast cancer death prevented for every four local recurrences prevented.30 The small, non-significant excess of breast cancer mortality in patients allocated NACT is consistent with this risk ratio, so could therefore be due to the increase in local recurrence, but it could equally well be a chance finding. The main aim of NACT in contemporary practice is to reduce the extent of breast surgery, thereby making breast conservation feasible in women who would otherwise need mastectomy. In the time since the trials in this meta-analysis were done, pathology reporting, surgery, and radiotherapy have improved, and more effective systemic neoadjuvant regimens have been introduced than were available when these trials took place. These changes should increase the likelihood of successful downstaging to allow conservative surgery in current and future practice. But, although improvements in treatment mean local recurrence risk should be lower than in these trials, our findings indicate that tumours downsized by NACT might continue to be associated with higher local recurrence risk after breast-conserving surgery than might tumours of the same dimensions in women who have not received NACT. Strategies to mitigate the increased local recurrence after breast-conserving therapy in tumours downsized by NACT should be considered—for example, careful tumour localisation, detailed pathological assessment, and appropriate radiotherapy.31 Prospective randomised trials would also help to establish the optimal clinical management in this context.

eased local recurrence after breast-conserving therapy in tumours downsized by NACT should be considered—for example, careful tumour localisation, detailed pathological assessment, and appropriate radiotherapy.31 Prospective randomised trials would also help to establish the optimal clinical management in this context. NACT allows more breast-conserving therapy than does adjuvant chemotherapy and provides information about an individual patient's response to a particular chemotherapy regimen. However, it appears to be no better than postoperative adjuvant treatment at reducing breast cancer mortality and, perhaps as a consequence of a reduction of the extent of surgery, NACT is associated with moderately increased local recurrence risk, which persists for at least 10 years. Correspondence to: EBCTCG Secretariat, Medical Research Council Population Health Research Unit, Nuffield Department of Population Health, Oxford OX3 7LF, UK bc.overview@ndph.ox.ac.uk Supplementary Material Supplementary appendix Acknowledgments This study is supported by core funding to the Population Health Research Unit and Clinical Trial Service Unit, Nuffield Department of Population Health, University of Oxford, from Cancer Research UK, the British Heart Foundation, and the UK Medical Research Council, as well as by the UK Department of Health grant RRX108 and Cancer Research UK grant C8225/A21133. The chief acknowledgement is to the women in these trials and the trial personnel.

ent of Population Health, University of Oxford, from Cancer Research UK, the British Heart Foundation, and the UK Medical Research Council, as well as by the UK Department of Health grant RRX108 and Cancer Research UK grant C8225/A21133. The chief acknowledgement is to the women in these trials and the trial personnel. Contributors The EBCTCG secretariat (C Boddington, R Bradley, J Burrett, M Clarke, C Davies, L Davies, D Dodwell, F Duane, V Evans, L Gettins, J Godwin, R Gray, S James, H Liu, Z Liu, E MacKinnon, G Mannu, P McGale, T McHugh, P Morris, H Pan, R Peto, S Read, C Taylor, Y Wang, and Z Wang) identified trials, obtained datasets, and had full access to them. P McGale, R Gray, and R Peto designed and carried out the analyses with computing assistance from Y Wang and Z Wang. P McGale, D Dodwell, R Gray, G von Minckwitz, R Peto, and C Taylor drafted the report with advice from S Anderson, J Bergh, L Gianni, C Lohrisch, and K I Pritchard. All writing committee members contributed to revising the report. Interim analyses were discussed by the steering committee and trialists who supplied data for the analysis.

l, R Gray, G von Minckwitz, R Peto, and C Taylor drafted the report with advice from S Anderson, J Bergh, L Gianni, C Lohrisch, and K I Pritchard. All writing committee members contributed to revising the report. Interim analyses were discussed by the steering committee and trialists who supplied data for the analysis. Writing committee P McGale, D Dodwell, R Gray, G Mannu, R Peto, C Taylor, Y Wang, Z Wang, M Clarke, C Davies, R Bradley, J Braybrooke, H Pan (Medical Research Council Population Health Research Unit, Nuffield Department of Population Health, University of Oxford, Oxford, UK); C Lohrisch (Breast Cancer Outcomes Unit, British Columbia Cancer Agency, Vancouver, BC, Canada); L Gianni (Istituto Nazionale Tumori, Milan, Italy); S Anderson (Department of Biostatistics, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA, USA); K I Pritchard (Sunnybrook Odette Cancer Centre, Toronto, ON, Canada; J Bergh (Karolinska Institutet and University Hospital, Stockholm, Sweden); G von Minckwitz (German Breast Group, Neu-Isenburg, Germany).

on (Department of Biostatistics, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA, USA); K I Pritchard (Sunnybrook Odette Cancer Centre, Toronto, ON, Canada; J Bergh (Karolinska Institutet and University Hospital, Stockholm, Sweden); G von Minckwitz (German Breast Group, Neu-Isenburg, Germany). Groups (and key trialists) contributing data Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan, Italy (P Valagussa, L Gianni, G Bonadonna); European Organisation for Research on the Treatment of Cancer (J Bogaerts, J van der Hage, C J H van de Velde); Austrian Breast and Colorectal Cancer Study Group, Vienna, Austria (L Solkner, M Gnant, R Jakesz); Center for Cancer Research of the National Cancer Institute, Bethesda, MD, USA. (S Steinberg, D Danforth, J Zujewski); Royal Marsden Hospital, Sutton, London, UK (T Powles, S Ashley, H Ford, M Makris); St George's Hospital, London, UK (J-C Gazet, C Coombes, R Sutcliffe); National Surgical Adjuvant Breast & Bowel Project, Pittsburgh, PA, USA (S Anderson, J Costantino, J Bryant, N Wolmark, E Mamounas, B Fisher); Fondation Bergonié—Comprehensive Cancer Centre of South West France, Bordeaux, France (L Mauriac, S Mathoulin); Cancer Control Agency of British Columbia, BC, Canada (J Ragaz, I Olivotto); Institut Curie, Paris, France (B Asselain, P Broet, S Scholl).

Anderson, J Costantino, J Bryant, N Wolmark, E Mamounas, B Fisher); Fondation Bergonié—Comprehensive Cancer Centre of South West France, Bordeaux, France (L Mauriac, S Mathoulin); Cancer Control Agency of British Columbia, BC, Canada (J Ragaz, I Olivotto); Institut Curie, Paris, France (B Asselain, P Broet, S Scholl). EBCTCG steering committee J Bergh, K I Pritchard (co-chairs), K Albain, S Anderson, R Arriagada, W Barlow, J Bartlett, E Bergsten-Nordström, J Bliss, F Boccardo, R Bradley*, E Brain, J Braybrooke*, D Cameron, M Clarke*, A Coates, R Coleman, C Correa, J Costantino, J Cuzick, N Davidson, C Davies*, L Davies*, A Di Leo, D Dodwell*, M Dowsett, F Duane*, M Ewertz, J Forbes, R Gelber, M Gnant, A Goldhirsch, P Goodwin, R Gray*, D Hayes, C Hill, J Ingle, R Jagsi, W Janni, Z Liu*, S Loibl, E MacKinnon*, G Mannu*, M Martín, P McGale*, H Mukai, L Norton, Y Ohashi, S Paik, H Pan*, R Peto*, M Piccart, L Pierce, P Poortmans, V Raina, P Ravdin, M Regan, J Robertson, E Rutgers, D Slamon, J Sparano, S Swain, C Taylor*, A Tutt, G Viale, G von Minckwitz, X Wang, T Whelan, N Wilcken, E Winer, N Wolmark, W Wood, M Zambetti. *EBCTCG Secretariat, Population Health Research Unit. Declaration of interests The writing committee and EBCTCG secretariat declare no competing interests.

events). The non-significant difference in 5-year overall survival of 5% found in PORTEC-3 was smaller than the study was powered to detect, and overall survival and failure-free survival probabilities were higher than expected from previous studies. Long-term outcomes will be analysed, especially for overall survival. The costs of chemoradiotherapy in terms of toxicity and treatment duration should be weighed against the benefits, and this cost–benefit tradeoff could be seen differently from the patient or physician perspective. In a patient preference study done by the ANZGOG group among PORTEC-3 participants,37 more than 50% of patients rated 5% survival improvement sufficient to make chemotherapy worthwhile. Although the trial results are in the range of this benefit for failure-free survival, the overall survival difference was not significant, thus individual patient counselling remains essential. Translational studies of molecular risk factors and tumour characteristics with the tumour samples of the PORTEC-3 participants might identify those who could most benefit from chemotherapy or targeted agents and individualise treatment of women with high-risk endometrial cancer.38

Research in context Evidence before this study We searched PubMed for clinical studies published in English between Jan 1, 1980, and Dec 31, 2006, with the terms “endometrial cancer” AND “radiation therapy” AND “chemotherapy” AND “survival” OR “failure free survival”. We identified six relevant publications. Three randomised controlled trials compared chemotherapy with radiotherapy. The GOG-122 trial compared whole abdominal radiotherapy with doxorubicin–cisplatin chemotherapy in patients with stage III or IV endometrial cancer. In adjusted analysis, improved progression-free survival and overall survival were reported for patients treated with chemotherapy, but with high proportions of patients with toxicity and similar event rates in both groups. Two trials (one in Italy and one in Japan) compared pelvic radiotherapy with three cycles or five cycles of cyclophosphamide–doxorubicin–cisplatin chemotherapy in stage I–III disease and neither showed any difference in overall survival or relapse-free survival. In the Italian trial, most patients had stage III disease, and chemotherapy delayed distant metastases and radiotherapy delayed pelvic recurrence, but without differences in overall survival or progression-free survival. Because of increased pelvic relapse with the use of adjuvant chemotherapy alone, the combination of chemotherapy and radiotherapy merited exploration. The phase 2 RTOG-9708 trial, which assessed toxicity on the chemoradiotherapy schedule on which PORTEC-3 is based, reported 4-year overall survival of 85% and disease-free survival of 81%. Since the start of recruitment to the PORTEC-3 trial, the results of three randomised trials comparing chemoradiotherapy with radiotherapy alone have been published. A small Finnish trial compared radiotherapy alone with chemotherapy plus radiotherapy given in two courses of 28 Gy each; no difference in overall survival or recurrence was reported.

t to the PORTEC-3 trial, the results of three randomised trials comparing chemoradiotherapy with radiotherapy alone have been published. A small Finnish trial compared radiotherapy alone with chemotherapy plus radiotherapy given in two courses of 28 Gy each; no difference in overall survival or recurrence was reported. The NSGO-EC-9501/EORTC-55991 trial was published in a pooled analysis with the unfinished ManGO Iliade phase 3 trial, and showed significantly improved progression-free survival and a trend for improved overall survival with the addition of four cycles of platinum-based chemotherapy given sequentially before or after pelvic radiotherapy. Two more trials (GOG-249 and GOG-258) have not yet been fully published, but have been presented as abstracts at conferences. The results of the GOG-249 trial, which compared pelvic radiotherapy with a combination of three cycles of carboplatin-paclitaxel chemotherapy and vaginal brachytherapy in stage I–II patients with high (intermediate) risk factors reported overlapping progression-free survival and overall survival curves, and significantly more pelvic and para-aortic recurrences in the chemotherapy group. The GOG-258 trial compared chemoradiotherapy (the same schedule as used in the PORTEC-3 trial) with six cycles of carboplatin–paclitaxel chemotherapy alone. No differences in overall or recurrence-free survival were reported, but significantly more vaginal and pelvic or para-aortic recurrences were reported in the chemotherapy group. Added value of this study

The NSGO-EC-9501/EORTC-55991 trial was published in a pooled analysis with the unfinished ManGO Iliade phase 3 trial, and showed significantly improved progression-free survival and a trend for improved overall survival with the addition of four cycles of platinum-based chemotherapy given sequentially before or after pelvic radiotherapy. Two more trials (GOG-249 and GOG-258) have not yet been fully published, but have been presented as abstracts at conferences. The results of the GOG-249 trial, which compared pelvic radiotherapy with a combination of three cycles of carboplatin-paclitaxel chemotherapy and vaginal brachytherapy in stage I–II patients with high (intermediate) risk factors reported overlapping progression-free survival and overall survival curves, and significantly more pelvic and para-aortic recurrences in the chemotherapy group. The GOG-258 trial compared chemoradiotherapy (the same schedule as used in the PORTEC-3 trial) with six cycles of carboplatin–paclitaxel chemotherapy alone. No differences in overall or recurrence-free survival were reported, but significantly more vaginal and pelvic or para-aortic recurrences were reported in the chemotherapy group. Added value of this study We report the overall survival and failure-free survival of patients with high-risk endometrial cancer treated in the international PORTEC-3 trial. Patients were randomly assigned to receive pelvic radiotherapy alone or radiotherapy combined with concurrent (two cycles of cisplatin) and adjuvant (four cycles of carboplatin–paclitaxel) chemotherapy. The addition of chemotherapy to adjuvant radiotherapy significantly improved failure-free survival compared with radiotherapy alone, but not overall survival. Vaginal and pelvic control was high with radiotherapy in both groups. The treatment duration was longer in the chemoradiotherapy group than in the radiotherapy group, and significantly higher rates of adverse events were reported in the chemoradiotherapy group during, and in the first year after, treatment.

l survival. Vaginal and pelvic control was high with radiotherapy in both groups. The treatment duration was longer in the chemoradiotherapy group than in the radiotherapy group, and significantly higher rates of adverse events were reported in the chemoradiotherapy group during, and in the first year after, treatment. Implications of all the available evidence Combined adjuvant chemotherapy and radiotherapy cannot be recommended as a new standard of care for patients with stage I–II endometrial cancer because no survival differences were found and pelvic control was high with radiotherapy alone. Patients with stage III cancer had the greatest benefit with chemoradiotherapy because of their higher risk of disease recurrence; for these patients, combined treatment should be considered to maximise failure-free survival. Nevertheless, the benefits and risks should be discussed for each individual patient.

y alone. Patients with stage III cancer had the greatest benefit with chemoradiotherapy because of their higher risk of disease recurrence; for these patients, combined treatment should be considered to maximise failure-free survival. Nevertheless, the benefits and risks should be discussed for each individual patient. Introduction The majority of women with endometrial cancer present with early-stage disease and have a favourable prognosis. About 15% of women with endometrial cancer are diagnosed with high-risk disease, which comprises endometrioid endometrial cancer stage I, grade 3 with deep invasion or with lymph-vascular space invasion (LVSI), stage II or III endometrioid endometrial cancer, or non-endometrioid (serous or clear cell) histology. Women with high-risk endometrial cancer are at increased risk of distant metastases and cancer-related death.1, 2, 3, 4 Serous and clear cell cancers have a higher risk of aggressive spread and a worse prognosis; however, in the early stages they have similar outcomes to grade 3 endometrioid endometrial cancer.5 Pelvic external beam radiotherapy has been the standard adjuvant treatment for women with high-risk endometrial cancer for many decades, although there is a paucity of evidence on improvement of survival. Randomised trials6, 7 have compared adjuvant chemotherapy with external beam radiotherapy. Radiotherapy was shown to delay pelvic recurrence and chemotherapy was shown to delay distant metastases, but no differences in survival were found.

any decades, although there is a paucity of evidence on improvement of survival. Randomised trials6, 7 have compared adjuvant chemotherapy with external beam radiotherapy. Radiotherapy was shown to delay pelvic recurrence and chemotherapy was shown to delay distant metastases, but no differences in survival were found. Because increased incidence of pelvic relapse has been reported with chemotherapy alone, the combination of external beam radiotherapy with chemotherapy has been explored. In a phase 2 trial (RTOG 9708)8 among women with high-risk endometrial cancer, the combination of external beam radiotherapy with two concurrent cycles of cisplatin, followed by four adjuvant cycles of cisplatin and paclitaxel, was tested, resulting in 4-year overall survival of 85% and disease-free survival of 81%. Because the combination of radiotherapy and chemotherapy (chemoradiotherapy) seemed more effective than either treatment alone, and because data for toxicity and quality of life were lacking, the randomised PORTEC-3 trial was initiated to evaluate the benefit of chemoradiotherapy versus radiotherapy alone for women with high-risk endometrial cancer in terms of overall survival and failure-free survival improvement, as well as toxicity and effects on health-related quality of life. Analysis of 2-year toxicity and health-related quality of life in the PORTEC-3 trial showed significantly higher rates of adverse events and reduced health-related quality of life during and after chemoradiotherapy treatment, with rapid recovery thereafter.9

ll as toxicity and effects on health-related quality of life. Analysis of 2-year toxicity and health-related quality of life in the PORTEC-3 trial showed significantly higher rates of adverse events and reduced health-related quality of life during and after chemoradiotherapy treatment, with rapid recovery thereafter.9 Here, we present the final analysis of the primary survival endpoints of the PORTEC-3 trial. Methods Study design and participants PORTEC-3 was an open-label, randomised, phase 3 trial at 103 centres (oncology centres, university hospitals, regional hospitals, or radiation oncology centres with referrals from regional hospitals) in six clinical trial groups collaborating in the Gynaecological Cancer Intergroup. Participating groups were the National Cancer Research Institute (NCRI; UK), Australia and New Zealand Gynaecologic Oncology Group (ANZGOG; Australia and New Zealand), Mario Negri Gynaecologic Oncology Group (MaNGO; Italy), Canadian Cancer Trials Group (CCTG; Canada), and Fedegyn (France).

ecological Cancer Intergroup. Participating groups were the National Cancer Research Institute (NCRI; UK), Australia and New Zealand Gynaecologic Oncology Group (ANZGOG; Australia and New Zealand), Mario Negri Gynaecologic Oncology Group (MaNGO; Italy), Canadian Cancer Trials Group (CCTG; Canada), and Fedegyn (France). Patients were eligible if they had endometrial cancer with either International Federation of Gynecology and Obstetrics (FIGO) 2009 stage 1A endometrioid endometrial cancer grade 3 with documented LVSI; stage IB endometrioid endometrial cancer grade 3; stage II endometrioid endometrial cancer; stage IIIA, IIIB (parametrial invasion), or IIIC endometrioid endometrial cancer; or serous or clear-cell histology endometrial cancer with stages IA (with invasion), IB, II, or III. Eligibility also included WHO performance score 0–2; adequate bone marrow function (white blood cells ≥3·0 × 109/L, platelets ≥100 × 109/L), liver function (bilirubin ≤1·5 × upper normal limit [UNL], aspartate aminotransferase and alanine aminotransferase ≤2·5 × UNL), kidney function (creatinine clearance >60 mL per min calculated according to Cockroft and Gault10 or >50 mL per min EDTA clearance), and aged 18 years or older (without an upper age limit, because elderly women might benefit from the study treatment if deemed fit enough to undergo chemotherapy). Exclusion criteria were uterine (carcino)sarcoma; malignancy in the 10 years before diagnosis of endometrial cancer; previous pelvic radiotherapy, hormonal therapy, or chemotherapy; bulky cervical involvement with radical hysterectomy; inflammatory bowel disease; residual macroscopic tumour; impaired renal or cardiac function; grade 2 or worse neuropathy; grade 3 or worse hearing impairment; or congenital hearing disorder.

al cancer; previous pelvic radiotherapy, hormonal therapy, or chemotherapy; bulky cervical involvement with radical hysterectomy; inflammatory bowel disease; residual macroscopic tumour; impaired renal or cardiac function; grade 2 or worse neuropathy; grade 3 or worse hearing impairment; or congenital hearing disorder. Surgery comprised total abdominal or laparoscopic hysterectomy with bilateral salpingo-oophorectomy. Lymphadenectomy, whether systemic or sampling, was left to the discretion of participating centres, while lymph node debulking and para-aortic lymph-node sampling were recommended in cases of macroscopic positive pelvic nodes or para-aortic nodes (or both). Lymphadenectomy was not mandated in view of the lack of improvement in overall or progression-free survival in early-stage disease and its associated toxicity, mainly lymph oedema.11, 12 For high-risk disease, the value of lymphadenectomy to direct adjuvant treatment is debated,13 and the international STATEC trial14 has been initiated to address this issue. For serous or clear-cell carcinoma, full surgical staging (with omentectomy, peritoneal biopsies, and lymph node sampling) was strongly recommended. Central pathology review by the groups' reference gynaecopathologists was required before randomisation to confirm patients' final suitability for study entry. Written informed consent was obtained from all patients. The protocol was approved by the Dutch Cancer Society and by the Ethics Committees of all participating groups. The study protocol is available online.

Surgery comprised total abdominal or laparoscopic hysterectomy with bilateral salpingo-oophorectomy. Lymphadenectomy, whether systemic or sampling, was left to the discretion of participating centres, while lymph node debulking and para-aortic lymph-node sampling were recommended in cases of macroscopic positive pelvic nodes or para-aortic nodes (or both). Lymphadenectomy was not mandated in view of the lack of improvement in overall or progression-free survival in early-stage disease and its associated toxicity, mainly lymph oedema.11, 12 For high-risk disease, the value of lymphadenectomy to direct adjuvant treatment is debated,13 and the international STATEC trial14 has been initiated to address this issue. For serous or clear-cell carcinoma, full surgical staging (with omentectomy, peritoneal biopsies, and lymph node sampling) was strongly recommended. Central pathology review by the groups' reference gynaecopathologists was required before randomisation to confirm patients' final suitability for study entry. Written informed consent was obtained from all patients. The protocol was approved by the Dutch Cancer Society and by the Ethics Committees of all participating groups. The study protocol is available online. Randomisation and masking Patients were randomly allocated (1:1) to chemoradiotherapy or radiotherapy alone. Treatment was allocated with a biased-coin minimisation procedure, ensuring balance overall and within each stratum of the stratification factors (participating centre, lymphadenectomy, stage of cancer, and histological type). Patients were registered and randomised by the participating group's data centres and treatment was assigned via a web-based application. The assigned treatment was generated immediately by the randomisation programme and confirmed by email. Participants and investigators were not masked to treatment allocation.

type). Patients were registered and randomised by the participating group's data centres and treatment was assigned via a web-based application. The assigned treatment was generated immediately by the randomisation programme and confirmed by email. Participants and investigators were not masked to treatment allocation. Procedures Central pathology review was done by reference gynaecopathologists (as appointed by each participating group before the start of the trial) to determine final eligibility. The slides and blocks were sent to each participating group's central review pathologists at one gynaecological pathology review site (in France and Italy), two sites (in the UK and the Netherlands), or five to six sites (in Australia and New Zealand, and Canada), with the result of the review confirming the patient's eligibility for the trial being sent to the local investigators within 1 week. Details of pathology review and inter-observer variation compared with local pathology assessment are reported separately.15 In this analysis, review pathology assessment was used. If any particular details were missing, the original pathology was used for these specific items. LVSI was recorded as present or absent. Extensive LVSI in the parametrial tissues was considered stage IIIB. In case of serosal breach, metastases in the stroma of the fallopian tubes, in the ovaries, or on the peritoneal surface of the tubes or ovaries (or both), the stage was defined as IIIA. After determination of eligibility and patient consent, a tumour sample was centrally stored for future translational research.

IIIB. In case of serosal breach, metastases in the stroma of the fallopian tubes, in the ovaries, or on the peritoneal surface of the tubes or ovaries (or both), the stage was defined as IIIA. After determination of eligibility and patient consent, a tumour sample was centrally stored for future translational research. External beam pelvic radiotherapy was given in both treatment groups to a total dose of 48·6 Gy in 1·8 Gy fractions, 5 days a week. For 11 of the 32 UK sites, a dose of 45 Gy (1·8 Gy fractions) was allowed if specified before initiation of the trial. The clinical target volume included the proximal vagina, parametrial tissues, and internal, external, and common iliac lymph node regions up to the L5–S1 level. The clinical target volume was extended to include the aortic bifurcation in case of iliac lymph node involvement; to include the lower peri-aortic region for common iliac node involvement; and to include the higher para-aortic region in case of para-aortic involvement (with a margin of ≥2 cm above the highest involved lymph node). If complete bilateral lymphadenectomy had been done with at least 12 lymph nodes, it was recommended to have the upper clinical target volume border at the iliac bifurcation. In case of cervical involvement (glandular, stromal, or both), a brachytherapy boost was given to the vaginal vault. Brachytherapy dose was equivalent to 14 Gy in 2 Gy fractions (with recommended scheme of 10 Gy high-dose rate [HDR] in fractions of 5 Gy), specified at 5 mm from the vaginal vault surface. Most patients were treated with a four-field technique; use of intensity-modulated radiotherapy was allowed for centres per approval by their group's principal investigator.

in 2 Gy fractions (with recommended scheme of 10 Gy high-dose rate [HDR] in fractions of 5 Gy), specified at 5 mm from the vaginal vault surface. Most patients were treated with a four-field technique; use of intensity-modulated radiotherapy was allowed for centres per approval by their group's principal investigator. Treatment was recommended to start within 4–6 weeks of surgery, but no later than 8 weeks. Overall radiotherapy treatment time was not to exceed 50 days. Radiotherapy quality assurance was not initially part of the trial, because pelvic radiotherapy was standard practice and used in both groups. However, the Trans-Tasman Radiation Oncology Group (TROG) initiated a bench-marking and quality assurance programme for the ANZGOG group,16 and in 2012, a protocol amendment allowed a short quality-assurance programme to be activated for all other participating sites, with independent review of a single radiotherapy plan for each site. Patients in the chemoradiotherapy group received two cycles of intravenous cisplatin 50 mg/m2 in the first and fourth week of external beam pelvic radiotherapy, followed by four cycles of intravenous carboplatin AUC5 and paclitaxel 175 mg/m2 at 21-day intervals. This schedule was based on the RTOG-9708 trial,8 with substitution of cisplatin by carboplatin in the adjuvant phase to reduce toxicity and in view of the use of carboplatin–paclitaxel chemotherapy in metastatic disease.17

owed by four cycles of intravenous carboplatin AUC5 and paclitaxel 175 mg/m2 at 21-day intervals. This schedule was based on the RTOG-9708 trial,8 with substitution of cisplatin by carboplatin in the adjuvant phase to reduce toxicity and in view of the use of carboplatin–paclitaxel chemotherapy in metastatic disease.17 Adjuvant chemotherapy was to be started within 3 weeks after completion of external beam pelvic radiotherapy, and with a 28-day interval from the second concurrent cycle. Toxicity, however, had to be resolved to better than grade 2 before start of chemotherapy. In the event of toxicities, cisplatin was postponed for 1 week. If recovery required more than 1 week, or in the case of neuropathy of grade 2 or worse, cisplatin was discontinued. Carboplatin was postponed or stopped in case of severe haematological toxicity. Paclitaxel was postponed for grade 2 neuropathy and stopped if recovery exceeded 1 week or grade 3 neuropathy developed. Carboplatin and paclitaxel were delayed for other grade 3–4 toxicities, and discontinued if no recovery or reduction to grade 1 occurred. Details on chemotherapy stopping rules have been described previously.9 At each follow-up, patient history was taken with emphasis on toxicities and symptoms of recurrent disease, and physical and pelvic examination were done. Chest radiography, blood count, and chemistry tests (including Ca-125) were to be obtained annually, up to 5 years. Long-term follow-up (by hospital visit or information from the general practitioner) was required at 7 years and 10 years.

and symptoms of recurrent disease, and physical and pelvic examination were done. Chest radiography, blood count, and chemistry tests (including Ca-125) were to be obtained annually, up to 5 years. Long-term follow-up (by hospital visit or information from the general practitioner) was required at 7 years and 10 years. Outcomes The coprimary endpoints were overall survival and failure-free survival. Overall survival was defined as time from date of randomisation to date of death from any cause. Failure-free survival (defined as any relapse or death related to endometrial cancer or treatment) was defined as time from randomisation to date of first failure-free survival event. Failure-free survival events were evaluated by the central data manager, the chief investigator, and the associated investigator, who were unaware of treatment allocation. Women who were alive at the time of analysis were censored at the date of their last follow-up. Secondary endpoints were vaginal, pelvic, or distant recurrence; treatment-related toxicity; and health-related quality of life (published elsewhere9). Recurrences were analysed according to first site of recurrence. Abdominal recurrences outside the pelvic area (peritoneal carcinomatosis, liver, and para-aortic lymph nodal metastases) were considered distant metastases, with specification of site.

d toxicity; and health-related quality of life (published elsewhere9). Recurrences were analysed according to first site of recurrence. Abdominal recurrences outside the pelvic area (peritoneal carcinomatosis, liver, and para-aortic lymph nodal metastases) were considered distant metastases, with specification of site. Toxicity was assessed and graded with Common Terminology Criteria for Adverse Events (CTCAE) version 3.018 at baseline (after surgery), at completion of radiotherapy, after each chemotherapy cycle, at 6-month intervals from randomisation until 5 years, and at 7 years and 10 years. Grade 2 or worse adverse events were to be reported, regardless of the association with study treatment. For evaluation of mild (grade 1) toxicities, patient-reported health-related quality-of-life symptoms were used because patient reporting of grade 1 toxicities was considered most reliable.19 Serious adverse events had to be reported within 24 h, specifying adverse event grade and whether or not they were associated with study treatment. Statistical analysis The PORTEC-3 trial was powered (80%) to detect a 10% difference in 5-year overall survival (increase from 65% to 75%; hazard ratio [HR] 0·67), with a two-sided α value of 0·05. 198 events were required, with a minimum number of 655 patients. The number of required patients was increased to 670 to ensure 655 eligible and evaluable patients. Power calculation of the coprimary endpoint failure-free survival was based on the same principles as overall survival.

, with a two-sided α value of 0·05. 198 events were required, with a minimum number of 655 patients. The number of required patients was increased to 670 to ensure 655 eligible and evaluable patients. Power calculation of the coprimary endpoint failure-free survival was based on the same principles as overall survival. The first prespecified interim analysis was done after 48 overall survival events (a third of the required events) had occurred in September, 2013, only 3 months before reaching complete accrual. In October, 2016, we decided, with permission from the Data Safety Monitoring Board (DSMB), not to do the prespecified second interim analysis at two-thirds of overall survival events, because this would have no consequences for the trial and would reduce α-spending. To maintain an overall α of 0·05, with a nominal α level for the first interim analysis of 0·0002, the final analysis was done with a nominal α of 0·0498. For analysis of the coprimary endpoints, overall survival and failure-free survival with a correlation between the test-statistics of the coprimary endpoints of 0·7859 (based on 136 overall survival events and 186 failure-free survival events), a nominal α of 0·0309 was used for each of the analyses, resulting in an overall α level of 0·0498.20

ary endpoints, overall survival and failure-free survival with a correlation between the test-statistics of the coprimary endpoints of 0·7859 (based on 136 overall survival events and 186 failure-free survival events), a nominal α of 0·0309 was used for each of the analyses, resulting in an overall α level of 0·0498.20 Because deaths in the PORTEC-3 trial were lower than expected at the time of trial design, the required number of overall survival events was not expected to be reached before late 2018. Recurrence was highest in the first 3 years after treatment, with a sharp decline thereafter, and relapse was rare after 5 years. For these reasons, the DSMB approved the final analysis becoming time-based rather than event-based, with final analysis at a median follow-up of 5 years and 42 months additional follow-up after inclusion of the last patient.

3 years after treatment, with a sharp decline thereafter, and relapse was rare after 5 years. For these reasons, the DSMB approved the final analysis becoming time-based rather than event-based, with final analysis at a median follow-up of 5 years and 42 months additional follow-up after inclusion of the last patient. We did statistical analyses using SPSS version 23.0 and R version 3.2.1. All analyses were done by intention to treat, excluding patients who immediately withdrew informed consent and ineligible patients. Differences in relapse and survival rates between the groups were tested with log-rank test and Cox-regression analysis. The analysis of the primary endpoints was adjusted for the stratification factors (participating group, lymphadenectomy, stage of cancer, and histological type), as the appropriate method when using a stratified minimisation procedure at randomisation.21, 22 For adjusted analysis, stratification factors were included as covariates in the Cox model. For analysis of failure-free survival and recurrence, competing-risk methods were used.23 For failure-free survival, intercurrent death was used as a competing risk. For the first failure analysis of recurrences, all other recurrences and death were used as competing risks. Predictive factors were assessed using Cox regression with treatment-by-covariate interaction including the stratification factors, as well as LVSI and age. The median follow-up was estimated with the reverse Kaplan Meier method. This study is registered with ISRCTN, number ISRCTN14387080 and ClinicalTrials.gov, number NCT00411138.

We did statistical analyses using SPSS version 23.0 and R version 3.2.1. All analyses were done by intention to treat, excluding patients who immediately withdrew informed consent and ineligible patients. Differences in relapse and survival rates between the groups were tested with log-rank test and Cox-regression analysis. The analysis of the primary endpoints was adjusted for the stratification factors (participating group, lymphadenectomy, stage of cancer, and histological type), as the appropriate method when using a stratified minimisation procedure at randomisation.21, 22 For adjusted analysis, stratification factors were included as covariates in the Cox model. For analysis of failure-free survival and recurrence, competing-risk methods were used.23 For failure-free survival, intercurrent death was used as a competing risk. For the first failure analysis of recurrences, all other recurrences and death were used as competing risks. Predictive factors were assessed using Cox regression with treatment-by-covariate interaction including the stratification factors, as well as LVSI and age. The median follow-up was estimated with the reverse Kaplan Meier method. This study is registered with ISRCTN, number ISRCTN14387080 and ClinicalTrials.gov, number NCT00411138. Role of the funding source The funding bodies had no role in study design, data collection, data interpretation, data analysis, or writing of this report. The central data manager (KWV), the chief investigator (CLC), the associated investigators (SMdB, RAN), and the trial statistician (HP) had full access to all the data. The decision to submit for publication was made after discussion within the trial management group and with approval of the DSMB. The corresponding author and chief investigator had full access to all the data and the final responsibility to submit for publication.

ial statistician (HP) had full access to all the data. The decision to submit for publication was made after discussion within the trial management group and with approval of the DSMB. The corresponding author and chief investigator had full access to all the data and the final responsibility to submit for publication. Results Between Nov 23, 2006, and Dec 20, 2013, 686 women were enrolled and randomly assigned to chemo-radiotherapy (n=343) or radiotherapy (n=343). 26 patients were excluded: 13 because of immediate informed consent withdrawal and 13 because they did not fulfil the eligibility criteria (figure 1). 660 patients were included in the primary analysis (chemoradiotherapy, n=330; radiotherapy, n=330). Median follow-up was 60·2 months (IQR 48·1–73·1) overall and was 60·0 months (47·8–73·1) in the chemoradiotherapy group and 60·7 months (48·7–72·9) in the radiotherapy group. There were seven major protocol violations: in the chemoradiotherapy group, five patients refused chemotherapy and received radiotherapy only; in the radiotherapy group, two patients asked to switch to chemoradiotherapy (figure 1).Figure 1 Trial profile Patient characteristics were well balanced between the treatment groups (table 1). The median age was 62 years (IQR 56·2–68·0). Lymphadenectomy, lymph node sampling, or full surgical staging were done in 190 patients (58%) in the chemoradiotherapy group and in 192 patients (58%) in the radiotherapy group.Table 1 Patient, tumour, and treatment characteristics

ed between the treatment groups (table 1). The median age was 62 years (IQR 56·2–68·0). Lymphadenectomy, lymph node sampling, or full surgical staging were done in 190 patients (58%) in the chemoradiotherapy group and in 192 patients (58%) in the radiotherapy group.Table 1 Patient, tumour, and treatment characteristics Chemoradiotherapy (n=330) Radiotherapy alone (n=330) Age at randomisation (years) Median 62·4 (56·5–67·9) 62·0 (55·8–68·2) <60 128 (39%) 140 (42%) 60–69 144 (44%) 128 (39%) ≥70 58 (18%) 62 (19%) Participating groups NCRI (UK) 82 (25%) 95 (29%) DGOG (Netherlands) 72 (22%) 66 (20%) ANZGOG (Australia and New Zealand) 60 (18%) 58 (18%) MaNGO (Italy) 52 (16%) 46 (14%) CCTG (Canada) 36 (11%) 29 (9%) Fedegyn (France) 28 (9%) 36 (11%) FIGO 2009 stage Stage IA 39 (12%) 38 (12%) Stage IB 59 (18%) 59 (18%) Stage II 80 (24%) 90 (27%) Stage III 152 (46%) 143 (43%) Histological grade and type EEC grade 1 68 (21%) 56 (17%) EEC grade 2 59 (18%) 73 (22%) EEC grade 3 90 (27%) 95 (29%) Serous 53 (16%) 52 (16%) Clear cell 29 (9%) 33 (10%) Mixed 17 (5%) 13 (4%) Other 14 (4%) 8 (2%) Myometrial invasion <50% 116 (35%) 123 (37%) ≥50% 212 (65%) 206 (63%) Missing 2 (<1%) 1 (<1%) LVSI Yes 197 (60%) 192 (58%) No 133 (40%) 138 (42%) WHO performance score 0–1 323 (99%) 324 (99%) ≥2 5 (2%) 5 (2%) Missing 2 (<1%) 1 (<1%) Comorbidity Diabetes 45 (14%) 36 (11%) Hypertension 116 (35%) 104 (32%) Cardiovascular 29 (9%) 20 (6%) Type of surgery TAH/BSO 95 (29%) 97 (29%) TAH/BSO + LND/full staging 143 (43%) 131 (40%) TLH/BSO 45 (14%) 41 (12%) TLH/BSO + LND/full staging 47 (14%) 61 (18%) Number of nodes removed TAH/BSO or TLH/BSO 0 (0–0) 0 (0–0) TAH/BSO or TLH/BSO +LND/full staging 15 (9–25) 14 (8–22) Missing 9 16 Radiotherapy EBRT completion 329 (100%) 325 (99%) Dose at prescription point Dose <45 Gy 1 (<1%) 4 (1%) Dose 45·0–50·4 Gy 329 (100%) 322 (98%) Dose >50·4 Gy 0 4 (1%) Vaginal brachytherapy boost 151 (46%) 158 (48%) Chemotherapy completed 1 cycle cisplatin 326 (99%) ·· 2 cycles cisplatin 304 (92%) ·· 1 cycle carboplatin and paclitaxel* 302 (91%) and 302 (91%) ·· 2 cycles carboplatin and paclitaxel* 294 (89%) and 291 (88%) ·· 3 cycles carboplatin and paclitaxel* 279 (85%) and 263 (80%) ·· 4 cycles carboplatin and paclitaxel* 262 (79%) and 233 (71%) ·· Data are median (IQR) or n (%). NCRI=National Cancer Research Institute. DGOG=Dutch Gynaecological Oncology Group. ANZGOG=Australia and New Zealand Gynaecologic Oncology Group. MaNGO=Mario Negri Gynaecologic Oncology Group.

axel* 279 (85%) and 263 (80%) ·· 4 cycles carboplatin and paclitaxel* 262 (79%) and 233 (71%) ·· Data are median (IQR) or n (%). NCRI=National Cancer Research Institute. DGOG=Dutch Gynaecological Oncology Group. ANZGOG=Australia and New Zealand Gynaecologic Oncology Group. MaNGO=Mario Negri Gynaecologic Oncology Group. CCTG=Canadian Cancer Trials Group. FIGO=International Federation of Gynecology and Obstetrics. EEC=endometrioid endometrial cancer. LVSI=lymphovascular space invasion. TAH/BSO=total abdominal hysterectomy with bilateral salpingo-oophorectomy. LND=lymph node dissection. TLH=total laparoscopic hysterectomy. EBRT=external beam radiotherapy. * In some cases, both drugs were not given because of toxicities. Radiotherapy was discontinued in one patient (<1%) in the chemoradiotherapy group because of disease progression and five patients (1·5%) in the radiotherapy group because of toxicity (table 1). 329 (100%) of 330 patients in the chemoradiotherapy group and 322 (98%) of 330 patients in the radiotherapy group received an external beam pelvic radiotherapy dose between 45·0 and 50·4 Gy. Vaginal brachytherapy was given in 309 (47%) patients (151 [46%] chemoradiotherapy patients vs 158 [48%] radiotherapy patients). Apart from the protocol indication for brachytherapy boost (cervical invasion), 28 (4%) patients received a brachytherapy boost for locally perceived reasons such as LVSI, grade 3, or stage III.

inal brachytherapy was given in 309 (47%) patients (151 [46%] chemoradiotherapy patients vs 158 [48%] radiotherapy patients). Apart from the protocol indication for brachytherapy boost (cervical invasion), 28 (4%) patients received a brachytherapy boost for locally perceived reasons such as LVSI, grade 3, or stage III. Both cycles of concurrent cisplatin were completed by 304 (92%) of 330 patients in the chemoradiotherapy group. Adjuvant chemotherapy was started by 304 (92%) patients, while 262 (79%) patients completed all four cycles of carboplatin and 233 (71%) patients completed all four cycles of paclitaxel (table 1). At least one dose reduction of cisplatin (to 40 mg/m2) was recorded for five (2%) patients, of carboplatin (from AUC5 to AUC4) for 36 (11%) patients, and of paclitaxel (from 175 mg/m2 to 135 mg/m2) for 50 (15%) patients. Chemotherapy was discontinued in 61 (18%) patients; in 31 (9%) because of toxicity, patient decision in 20 (6%), disease progression in seven (2%), and for other reasons in three (1%). Evaluation of the TROG quality assurance programme for the ANZGOG group showed that a radiotherapy benchmarking exercise before participation in the trial ensured high conformity and low rates of both minor and major contouring deviations.16 Evaluation of radiotherapy plans from centres in other countries is ongoing and will be reported separately.

assurance programme for the ANZGOG group showed that a radiotherapy benchmarking exercise before participation in the trial ensured high conformity and low rates of both minor and major contouring deviations.16 Evaluation of radiotherapy plans from centres in other countries is ongoing and will be reported separately. At final database lock on May 1, 2017, 136 patients had died (61 in the chemoradiotherapy group and 75 in the radiotherapy group) and 186 patients had a failure-free survival event (83 in the chemoradiotherapy group and 103 in the radiotherapy group). Among the patients assigned to chemoradiotherapy, 50 (82%) had died from endometrial cancer, four (7%) from a second cancer, three (5%) from other intercurrent disease, and two (3%) from treatment for metastatic disease. Among the patients assigned to radiotherapy, 68 (91%) had died from endometrial cancer and five (7%) from a second cancer. For the remaining four patients (two patients treated with chemoradiotherapy and two patients with radiotherapy), the cause of death was uncertain. In one patient in the radiotherapy group, death was due to either disease progression or late treatment complications; in two patients in the chemoradiotherapy group and one in the radiotherapy group, death was due to either intercurrent disease or late treatment-related toxicity. These four deaths were counted as failure-free survival events after discussion with the DSMB.

to either disease progression or late treatment complications; in two patients in the chemoradiotherapy group and one in the radiotherapy group, death was due to either intercurrent disease or late treatment-related toxicity. These four deaths were counted as failure-free survival events after discussion with the DSMB. Estimated overall survival adjusted for stratification factors at 5 years was 81·8% (95% CI 77·5–86·2) for patients in the chemoradiotherapy group versus 76·7% (72·1–81·6) for patients in the radiotherapy group (HR 0·76, 95% CI 0·54–1·06; p=0·109; table 2, figure 2). 5-year failure-free survival was 75·5% (70·3–79·9) in the chemoradiotherapy group versus 68·6% (63·1–73·4) in the radiotherapy group (HR 0·71, 0·53–0·95; p=0·022). Without adjusting for the stratification factors, the HR for overall survival was 0·81 (95% CI 0·58–1·13; p=0·213) and for failure-free survival was 0·76 (0·57–1·02; p=0·067; table 2, figure 2).Table 2 Survival and recurrence outcomes

oup versus 68·6% (63·1–73·4) in the radiotherapy group (HR 0·71, 0·53–0·95; p=0·022). Without adjusting for the stratification factors, the HR for overall survival was 0·81 (95% CI 0·58–1·13; p=0·213) and for failure-free survival was 0·76 (0·57–1·02; p=0·067; table 2, figure 2).Table 2 Survival and recurrence outcomes Events 5-year estimate, % (95% CI) Hazard ratio (95% CI) p value Overall survival* ·· ·· 0·76 (0·54–1·06) 0·109 Failure-free survival* ·· ·· 0·71 (0·53–0·95) 0·022 Overall survival† Chemoradiotherapy 61 81·8% (77·5–86·2) 0·81 (0·58–1·13) 0·213 Radiotherapy 75 76·7% (72·1–81·6) ·· ·· Failure-free survival† Chemoradiotherapy 83 75·5% (70·3–79·9) 0·76 (0·57–1·02) 0·067 Radiotherapy 103 68·6% (63·1–73·4) ·· ·· Vaginal recurrence (first recurrence)† Chemoradiotherapy 1 0·3% (0·0–2·1) 0·99 (0·06–15·90) 0·999 Radiotherapy 1 0·3% (0·0–2·1) ·· ·· Pelvic recurrence (first recurrence)† Chemoradiotherapy 3 1·0% (0·3–2·9) 0·60 (0·14–2·49) 0·473 Radiotherapy 5 1·5% (0·6–3·6) ·· ·· Distant metastases (first recurrence)† Chemoradiotherapy 76 22·4% (18·1–27·4) 0·78 (0·58–1·06) 0·108 Radiotherapy 93 28·3% (23·7–33·7) ·· ·· Vaginal recurrence (total)† Chemoradiotherapy 8 2·1% (1·0–4·4) 0·99 (0·37–2·65) 0·995 Radiotherapy 8 2·1% (1·0–4·4) ·· ·· Pelvic recurrence (total)† Chemoradiotherapy 16 4·9% (3·0–7·9) 0·51 (0·28–0·92) 0·026 Radiotherapy 31 9·2% (6·5–12·9) ·· ·· Distant metastases (total)† Chemoradiotherapy 79 23·1% (18·8–28·3) 0·77 (0·57–1·03) 0·077 Radiotherapy 97 29·7% (24·9–35·1) ·· ·· * Data are chemotherapy versus radiotherapy (Cox-adjusted p value), adjusted for stratification factors: participating groups, type of surgery (abdominal hysterectomy and salpingo-oophorectomy vs abdominal surgery plus lymphadenectomy vs laparoscopic procedure vs laparoscopic procedure plus lymphadenectomy), stage (FIGO 2009 IA vs IB vs II vs III), and histological type (endometrioid carcinoma vs serous or clear cell carcinoma).

pating groups, type of surgery (abdominal hysterectomy and salpingo-oophorectomy vs abdominal surgery plus lymphadenectomy vs laparoscopic procedure vs laparoscopic procedure plus lymphadenectomy), stage (FIGO 2009 IA vs IB vs II vs III), and histological type (endometrioid carcinoma vs serous or clear cell carcinoma). † Log-rank p value, unadjusted for stratification factors. Figure 2 Overall survival and failure-free survival Kaplan-Meier survival curves for overall survival (A) and failure-free survival (B) in all patients, and for overall survival (C) and failure-free survival (D) of patients with stage III endometrial cancer. Plog-rank=unadjusted log-rank p value. PCox adjusted=p value adjusted for stratification factors. HR=hazard ratio.

Kaplan-Meier survival curves for overall survival (A) and failure-free survival (B) in all patients, and for overall survival (C) and failure-free survival (D) of patients with stage III endometrial cancer. Plog-rank=unadjusted log-rank p value. PCox adjusted=p value adjusted for stratification factors. HR=hazard ratio. In subgroup analysis, women with stage III endometrial cancer had significantly lower overall survival and failure-free survival than those with stage I–II disease (Table 3, Table 4). 5-year overall survival for stage III cancer was 78·7% (95% CI 72·2–85·7) in the chemoradiotherapy group versus 69·8% (62·4–78·1) in the radiotherapy group (HR 0·71, 95% CI 0·45–1·11; p=0·13; adjusted p=0·074). 5-year failure-free survival for stage III cancer was 69·3% (95% CI 61·1–76·2) in the chemoradiotherapy group versus 58·0% (49·3–65·7) in the radiotherapy group (HR 0·66, 95% CI 0·45–0·97; p=0·031; adjusted p=0·014; figure 2). 5-year failure-free survival for stage I–II patients was 80·8% (74·1–86·0) in the chemoradiotherapy group versus 76·6% (69·5–82·2) in the radiotherapy group (0·85, 0·54–1·33; p=0·47).Table 3 Multivariable analysis of prognostic factors for overall survival

R 0·66, 95% CI 0·45–0·97; p=0·031; adjusted p=0·014; figure 2). 5-year failure-free survival for stage I–II patients was 80·8% (74·1–86·0) in the chemoradiotherapy group versus 76·6% (69·5–82·2) in the radiotherapy group (0·85, 0·54–1·33; p=0·47).Table 3 Multivariable analysis of prognostic factors for overall survival Patients (n) Events (n) 5-year overall survival (95% CI) Hazard ratio (95% CI) p value Total 660 136 79% (74·8–83·9) ·· ·· Treatment group ·· ·· ·· ·· 0·075 Radiotherapy 330 75 77% (72·1–81·6) ·· ·· Chemoradiotherapy 330 61 82% (77·5–86·2) 0·73 (0·52–1·03) ·· Age (years) ·· ·· ·· ·· <0·0001 <60 268 31 89% (85·0–92·9) ·· ·· 60–69 272 66 75% (69·6–80·6) 2·31 (1·48–3·59) ·· ≥70 120 39 67% (58·7–76·3) 3·29 (1·99–5·44) ·· Stage ·· ·· ·· ·· <0·0001 Stage I and II 365 59 83% (79·1–87·3) ·· ·· Stage III 295 77 74% (69·3–79·7) 2·41 (1·66–3·51) ·· Histology and grade ·· ·· ·· ·· <0·0001 Endometrioid grade 1 and 2 258 36 86% (81·9–90·9) ·· ·· Endometrioid grade 3 213 45 79% (73·0–85·7) 1·76 (1·10–2·81) ·· Serous/clear cell 189 55 71% (65·2–77·4) 2·35 (1·48–3·72) ·· LVSI ·· ·· ·· ·· 0·11 No 271 43 85% (80·5–89·4) ·· ·· Yes 389 93 75% (70·9–79·9) 1·36 (0·93–1·98) ·· Lymphadenectomy ·· ·· ·· ·· 0·33 No 278 61 77% (71·4–82·1) ·· ·· Yes 382 75 81% (77·1–85·2) 0·82 (0·55–1·22) ·· Adjusted for participating groups. LVSI=lymph-vascular space invasion. Table 4 Multivariable analysis of prognostic factors for failure-free survival

Patients (n) Events (n) 5-year overall survival (95% CI) Hazard ratio (95% CI) p value Total 660 136 79% (74·8–83·9) ·· ·· Treatment group ·· ·· ·· ·· 0·075 Radiotherapy 330 75 77% (72·1–81·6) ·· ·· Chemoradiotherapy 330 61 82% (77·5–86·2) 0·73 (0·52–1·03) ·· Age (years) ·· ·· ·· ·· <0·0001 <60 268 31 89% (85·0–92·9) ·· ·· 60–69 272 66 75% (69·6–80·6) 2·31 (1·48–3·59) ·· ≥70 120 39 67% (58·7–76·3) 3·29 (1·99–5·44) ·· Stage ·· ·· ·· ·· <0·0001 Stage I and II 365 59 83% (79·1–87·3) ·· ·· Stage III 295 77 74% (69·3–79·7) 2·41 (1·66–3·51) ·· Histology and grade ·· ·· ·· ·· <0·0001 Endometrioid grade 1 and 2 258 36 86% (81·9–90·9) ·· ·· Endometrioid grade 3 213 45 79% (73·0–85·7) 1·76 (1·10–2·81) ·· Serous/clear cell 189 55 71% (65·2–77·4) 2·35 (1·48–3·72) ·· LVSI ·· ·· ·· ·· 0·11 No 271 43 85% (80·5–89·4) ·· ·· Yes 389 93 75% (70·9–79·9) 1·36 (0·93–1·98) ·· Lymphadenectomy ·· ·· ·· ·· 0·33 No 278 61 77% (71·4–82·1) ·· ·· Yes 382 75 81% (77·1–85·2) 0·82 (0·55–1·22) ·· Adjusted for participating groups. LVSI=lymph-vascular space invasion. Table 4 Multivariable analysis of prognostic factors for failure-free survival Patients (n) Events (n) 5-year failure-free survival (95% CI) Hazard ratio (95% CI) p value Total 660 186 72% (66·7–76·7) ·· ·· Treatment group ·· ·· ·· ·· 0·010 Radiotherapy 330 103 68% (63·1–73·4) ·· ·· Chemoradiotherapy 330 83 75% (70·3–79·9) 0·68 (0·51–0·91) ·· Age (years) ·· ·· ·· ·· <0·0001 <60 268 54 81% (75·3–85·0) ·· ·· 60–69 272 87 67% (60·7–72·4) 1·74 (1·23–2·46) ·· ≥70 120 45 64% (54·4–71·7) 2·14 (1·41–3·25) ·· Stage ·· ·· ·· ·· <0·0001 Stage I and II 365 78 79% (73·9–82·6) ·· ·· Stage III 295 108 64% (58·0–69·2) 2·62 (1·90–3·61) ·· Histology and grade ·· ·· ·· ·· <0·0001 Endometrioid grade 1 and 2 258 58 78% (72·7–83·1) ·· ·· Endometrioid grade 3 213 60 71% (64·5–77·1) 1·56 (1·06–2·30) ·· Serous or clear cell 189 68 64% (56·6–70·4) 2·15 (1·46–3·16) ·· LVSI ·· ·· ·· ·· 0·054 No 271 62 77% (71·4–81·8) ·· ·· Yes 389 124 68% (63·4–72·9) 1·36 (0·99–1·87) ·· Lymphadenectomy ·· ·· ·· ·· 0·41 No 278 81 72% (65·7–76·6) ·· ·· Yes 382 105 72% (67·4–76·7) 0·87 (0·61–1·22) ·· Adjusted for participating groups. LVSI=lymph-vascular space invasion.

% (56·6–70·4) 2·15 (1·46–3·16) ·· LVSI ·· ·· ·· ·· 0·054 No 271 62 77% (71·4–81·8) ·· ·· Yes 389 124 68% (63·4–72·9) 1·36 (0·99–1·87) ·· Lymphadenectomy ·· ·· ·· ·· 0·41 No 278 81 72% (65·7–76·6) ·· ·· Yes 382 105 72% (67·4–76·7) 0·87 (0·61–1·22) ·· Adjusted for participating groups. LVSI=lymph-vascular space invasion. Serous cancers (>25% serous component) had significantly lower overall survival and failure-free survival than the other histological subtypes; failure-free was 58% (95% CI 42–70) with chemoradiotherapy versus 48% (34–61) with radiotherapy (HR 0·63, 95% CI 0·36–1·12; p=0·11). The number of patients and events are, however, small in these subgroups (appendix p 3). Isolated vaginal and pelvic recurrences were rare, with isolated vaginal recurrence diagnosed in one (<1%) patient in the chemoradiotherapy group and in one (<1%) patient in the radiotherapy group (p=0·995), and isolated pelvic recurrence in three (1%) patients in the chemoradiotherapy group versus five (2%) patients in the radiotherapy group (p=0·473). Most recurrences were distant metastases: 76 (22%) patients in the chemoradiotherapy group versus 93 (28%) patients in the radiotherapy group were diagnosed with distant metastases (p=0·108). The 5-year estimate of pelvic recurrence (both isolated and combined pelvic and distant recurrences) was 4·9% (95% CI 3·0–7·9) for the chemoradiotherapy group versus 9·2% (6·5–12·9) for the radiotherapy group (p=0·026; table 2).

(28%) patients in the radiotherapy group were diagnosed with distant metastases (p=0·108). The 5-year estimate of pelvic recurrence (both isolated and combined pelvic and distant recurrences) was 4·9% (95% CI 3·0–7·9) for the chemoradiotherapy group versus 9·2% (6·5–12·9) for the radiotherapy group (p=0·026; table 2). In the multivariable analysis, the following covariates were included together with treatment: stage, histological type and grade, type of surgery, participating groups, LVSI, and age. In the presence of these factors, combined chemotherapy and radiotherapy significantly improved failure-free survival. Most factors, except lymphadenectomy, were significantly correlated with failure-free survival (table 4). In multivariable analysis for failure-free survival, only age group was found to be predictive of treatment effect, with a strong treatment-by-age effect (pinteraction=0·012, figure 3). Women aged 70 years or older had the greatest benefit from chemoradiotherapy.Figure 3 Forest plot of multivariable analysis (treatment by covariate interaction) of overall survival (A) and failure-free survival (B) For the multivariable analysis the stratification factors (participating group, lymphadenectomy, stage of cancer, and histological type), lymphovascular space invasion, and age were used. HR=hazard ratio. LVSI=lymphovascular space invasion. FIGO=International Federation of Gynecology and Obstetrics.

In multivariable analysis for failure-free survival, only age group was found to be predictive of treatment effect, with a strong treatment-by-age effect (pinteraction=0·012, figure 3). Women aged 70 years or older had the greatest benefit from chemoradiotherapy.Figure 3 Forest plot of multivariable analysis (treatment by covariate interaction) of overall survival (A) and failure-free survival (B) For the multivariable analysis the stratification factors (participating group, lymphadenectomy, stage of cancer, and histological type), lymphovascular space invasion, and age were used. HR=hazard ratio. LVSI=lymphovascular space invasion. FIGO=International Federation of Gynecology and Obstetrics. Grade 2 or worse adverse events were reported during treatment in 308 (93%) women in the chemoradiotherapy group versus 144 (43%) in the radiotherapy group, and grade 3 or worse in 198 (60%) versus 41 (12%; p<0·0001; table 5); the majority of grade 3 or worse adverse events were haematological. Table 6 shows an overview of adverse events at 6 months after randomisation, which was about 1 month after completion of treatment in the chemoradiotherapy group. There were no treatment-related deaths. From 12 months onwards, no significant differences between the groups were found in grade 3 or worse adverse events (appendix p 5). The number of patients with grade 2 or worse adverse events was 86 (32%) for chemoradiotherapy versus 64 (24%) for radiotherapy at 3 years (p=0·034), and 57 (40%) versus 38 (28%) at 5 years (p=0·033). The most significant and clinically relevant difference between the arms was found for grade 2 or worse sensory neuropathy, which persisted in 20 (8%) women in the chemoradiotherapy group versus one (1%) women in the radiotherapy group at 3 years and 12 (9%) women versus no women at 5 years (both p<0·0001). An extensive overview of adverse events during follow-up is in the appendix (pp 4–5).Table 5 Adverse events reported during treatment

, which persisted in 20 (8%) women in the chemoradiotherapy group versus one (1%) women in the radiotherapy group at 3 years and 12 (9%) women versus no women at 5 years (both p<0·0001). An extensive overview of adverse events during follow-up is in the appendix (pp 4–5).Table 5 Adverse events reported during treatment Grade 2 Grade 3–4 Chemoradiotherapy Radiotherapy p value* Chemoradiotherapy Radiotherapy p value† Any 110 (33%) 103 (31%) <0·0001 198 (60%) 41 (12%) <0·0001 Any grade 3 NA NA ·· 148 (45%) 41 (12%) ·· Any grade 4 NA NA ·· 50 (15%) 0 ·· Auditory or hearing 14 (4%) 3 (1%) 0·011 1 (<1%) 1 (<1%) 1·00 Allergy 23 (7%) 1 (<1%) <0·0001 5 (2%) 0 0·062 Fatigue 69 (21%) 7 (2%) <0·0001 10 (3%) 0 0·0018 Hypertension 19 (6%) 12 (4%) 0·14 6 (2%) 3 (1%) 0·50 Alopecia 187 (57%) 1 (<1%) <0·0001 NA NA ·· Dermatitis 18 (5%) 5 (2%) 0·013 1 (<1%) 1 (<1%) 1·0 Any gastrointestinal 145 (44%) 79 (24%) <0·0001 47 (14%) 18 (5%) <0·0001 Diarrhoea 104 (32%) 69 (21%) <0·0001 35 (11%) 14 (4%) 0·0027 Nausea 68 (21%) 24 (7%) 0·0010 9 (3%) 2 (1%) 0·06 Vomiting 31 (9%) 9 (3%) <0·0001 5 (2%) 0 0·06 Anorexia 30 (9%) 9 (3%) 0·0033 3 (1%) 4 (1%) 1·00 Constipation 32 (10%) 6 (2%) <0·0001 1 (<1%) 0 1·00 Genito-urinary: frequency or urgency 24 (7%) 10 (3%) 0·020 2 (1%) 2 (1%) 1·00 Any haematological 100 (30%) 19 (6%) <0·0001 149 (45%) 18 (5%) <0·0001 Febrile neutropenia NA NA ·· 9 (3%) 1 (<1%) 0·021 Infection with neutropenia 3 (1%) 0 0·0018 7 (2%) 0 0·015 Infection without neutropenia 21 (6%) 1 (<1%) <0·0001 12 (4%) 1 (<1%) <0·0001 Haemoglobin 105 (32%) 0 <0·0001 27 (8%) 0 <0·0001 Leucocytes 98 (30%) 3 (1%) <0·0001 76 (23%) 1 (<1%) <0·0001 Lymphocytes 48 (15%) 16 (5%) <0·0001 109 (33%) 17 (5%) <0·0001 Neutrophils 62 (19%) 1 (<1%) <0·0001 66 (20%) 1 (<1%) <0·0001 Platelets 22 (7%) 0 <0·0001 18 (5%) 0 <0·0001 Metabolic or laboratory 15 (5%) 1 (<1%) <0·0001 3 (1%) 0 0·25 Any neuropathy 82 (25%) 1 (<1%) <0·0001 23 (7%) 0 <0·0001 Motor 13 (4%) 1 (<1%) <0·0001 4 (1%) 0 0·12 Sensory 79 (24%) 0 <0·0001 22 (7%) 0 <0·0001 Any pain 101 (31%) 23 (7%) <0·0001 31 (9%) 4 (1%) <0·0001 Joint 52 (16%) 2 (1%) <0·0001 10 (3%) 0 0·0018 Muscle 52 (16%) 1 (<1%) <0·0001 9 (3%) 0 0·0037 Pelvic, back, or limb 10 (3%) 4 (1%) <0·0001 11 (3%) 0 <0·0001 Pulmonary: dyspnoea 12 (4%) 2 (1%) <0·0001 5 (2%) 0 0·062 Thrombosis or embolism 2 (1%) 0 0·031 4 (1%) 0 0·12 Data are n (%).

(7%) <0·0001 31 (9%) 4 (1%) <0·0001 Joint 52 (16%) 2 (1%) <0·0001 10 (3%) 0 0·0018 Muscle 52 (16%) 1 (<1%) <0·0001 9 (3%) 0 0·0037 Pelvic, back, or limb 10 (3%) 4 (1%) <0·0001 11 (3%) 0 <0·0001 Pulmonary: dyspnoea 12 (4%) 2 (1%) <0·0001 5 (2%) 0 0·062 Thrombosis or embolism 2 (1%) 0 0·031 4 (1%) 0 0·12 Data are n (%). Adverse events are listed that occurred in at least 5% of patients, or were significantly different between the study groups at any timepoint during treatment, or both. Adverse events were calculated at each timepoint. For each adverse event, the maximum grade per patient was calculated (worst ever by patient). Adverse events were graded according to Common Terminology Criteria for Adverse Events version 3.0. Chemoradiotherapy, n=330; radiotherapy, n=330. NA=not applicable. * Significance level for grade 2, 3, and 4. † Significance level for grade 3 and 4. Table 6 Adverse events reported at 6 months after randomisation

Adverse events are listed that occurred in at least 5% of patients, or were significantly different between the study groups at any timepoint during treatment, or both. Adverse events were calculated at each timepoint. For each adverse event, the maximum grade per patient was calculated (worst ever by patient). Adverse events were graded according to Common Terminology Criteria for Adverse Events version 3.0. Chemoradiotherapy, n=330; radiotherapy, n=330. NA=not applicable. * Significance level for grade 2, 3, and 4. † Significance level for grade 3 and 4. Table 6 Adverse events reported at 6 months after randomisation Grade 2 Grade 3–4 Chemoradiotherapy Radiotherapy p value* Chemoradiotherapy Radiotherapy p value† Any 128 (39%) 96 (29%) <0·0001 54 (16%) 25 (8%) <0·0001 Any grade 3 NA NA ·· 49 (15%) 21 (6%) ·· Any grade 4 NA NA ·· 5 (2%) 4 (1%) ·· Auditory or hearing 8 (2%) 3 (1%) 0·22 0 0 1·00 Allergy 2 (1%) 1 (<1%) 1·00 0 0 1·00 Fatigue 10 (3%) 2 (1%) 0·054 1 (<1%) 1 (<1%) 1·00 Hypertension 15 (5%) 18 (5%) 0·75 5 (2%) 5 (2%) 1·00 Alopecia 64 (19%) 0 <0·0001 NA NA ·· Dermatitis 1 (<1%) 0 1·00 0 0 1·00 Any gastrointestinal 19 (6%) 18 (5%) 0·89 7 (2%) 9 (3%) 0·80 Diarrhoea 8 (2%) 11 (3%) 0·20 0 3 (1%) 0·25 Nausea 7 (2%) 5 (2%) 0·35 5 (2%) 2 (1%) 0·45 Vomiting 7 (2%) 6 (2%) 0·45 3 (1%) 0 0·25 Anorexia 1 (<1%) 4 (1%) 0·50 2 (1%) 2 (1%) 1·00 Constipation 7 (2%) 4 (1%) 0·79 1 (<1%) 2 (1%) 1·00 Genito-urinary: frequency or urgency 5 (2%) 6 (2%) 0·77 1 (<1%) 0 1·00 Any haematological 54 (16%) 27 (8%) <0·0001 24 (7%) 6 (2%) 0·0001 Febrile neutropenia NA NA ·· 0 0 1·00 Infection with neutropenia 0 0 1·00 1 (<1%) 0 1·00 Infection without neutropenia 5 (2%) 1 (<1%) 0·22 0 0 1·00 Haemoglobin 24 (7%) 0 <0·0001 3 (1%) 0 0·25 Leucocytes 20 (6%) 1 (<1%) <0·0001 6 (2%) 1 (<1%) 0·12 Lymphocytes 43 (13%) 26 (8%) 0·0015 17 (5%) 5 (2%) 0·015 Neutrophils 4 (1%) 0 0·0018 6 (2%) 0 0·031 Platelets 5 (2%) 0 0·015 2 (1%) 0 0·50 Metabolic or laboratory 2 (1%) 0 0·12 2 (1%) 0 0·50 Any neuropathy 42 (13%) 1 (<1%) <0·0001 8 (2%) 2 (1%) 0·11 Motor 7 (2%) 1 (<1%) 0·09 3 (1%) 2 (1%) 1·00 Sensory 41 (12%) 0 <0·0001 6 (2%) 0 0·031 Any pain 31 (9%) 32 (10%) 0·54 3 (1%) 7 (2%) 0·34 Joint 8 (2%) 6 (2%) 1·00 0 1 (<1%) 1·00 Muscle 5 (2%) 1 (<1%) 0·22 0 0 1·00 Pelvic, back, or limb 10 (3%) 9 (3%) 0·33 0 3 (1%) 0·25 Pulmonary: dyspnoea 1 (<1%) 0 1·00 1 (<1%) 1 (<1%) 1·00 Thrombosis or embolism 3 (1%) 0 0·37 1 (<1%) 1 (<1%) 1·00 Data are n (%). Adverse events are listed that occurred in at least 5% of patients, or were significantly different between the study groups at any timepoint, or both. Adverse events were calculated at each timepoint. For each adverse event, the maximum grade per patient was calculated (worst ever by patient).

1%) 1·00 Data are n (%). Adverse events are listed that occurred in at least 5% of patients, or were significantly different between the study groups at any timepoint, or both. Adverse events were calculated at each timepoint. For each adverse event, the maximum grade per patient was calculated (worst ever by patient). Adverse events were graded according to Common Terminology Criteria for Adverse Events version 3.0. Chemoradiotherapy, n=329; radiotherapy, n=329. NA=not applicable. * Significance level for grade 2, 3, and 4. † Significance level for grade 3 and 4. Discussion The final results of the PORTEC-3 trial showed that the combination of adjuvant chemotherapy and radiotherapy for high-risk endometrial cancer did not significantly improve overall survival. However, chemoradiotherapy did improve 5-year failure-free survival compared with radiotherapy alone. Patients with stage III disease—who had a higher risk of recurrence than those with stages I–II—had a HR of 0·66 and 11% absolute improvement of failure-free survival with chemo-radiotherapy, which is clinically relevant and exceeds the 10% improvement used when designing the study.

al compared with radiotherapy alone. Patients with stage III disease—who had a higher risk of recurrence than those with stages I–II—had a HR of 0·66 and 11% absolute improvement of failure-free survival with chemo-radiotherapy, which is clinically relevant and exceeds the 10% improvement used when designing the study. The improvement in failure-free survival in the chemoradiotherapy group should be weighed against the severity and duration of toxicity of combined treatment, especially since overall survival was not significantly improved. Although significantly higher incidences of adverse events and reduced health-related quality of life were reported in the chemoradiotherapy group during and after treatment,9 rapid recovery was seen, with no differences in grade 3–4 adverse events from 12 months onwards. Grade 2 sensory neuropathy, however, persisted significantly more often in patients treated with chemoradiotherapy, with 25% of patients reporting “quite a bit” or “very much” tingling or numbness at 2 years, compared with 6% for radiotherapy.9 Sensory neuropathy is associated with lower levels of functioning and quality of life, and more fatigue.24

ever, persisted significantly more often in patients treated with chemoradiotherapy, with 25% of patients reporting “quite a bit” or “very much” tingling or numbness at 2 years, compared with 6% for radiotherapy.9 Sensory neuropathy is associated with lower levels of functioning and quality of life, and more fatigue.24 For decades, standard adjuvant treatment for women with high-risk endometrial cancer has been pelvic external beam radiotherapy. It has been hypothesised that chemotherapy might improve survival by reducing the risk of metastatic disease. Randomised trials comparing adjuvant chemotherapy with external beam radiotherapy failed to show an improvement in progression-free survival or overall survival.6, 7 Retrospective studies reported substantial rates of pelvic recurrence if high-risk patients were treated without radiotherapy, supporting the combined use of pelvic radiotherapy with adjuvant chemotherapy, as first explored in the RTOG 9708 phase 2 trial.8, 25, 26

t in progression-free survival or overall survival.6, 7 Retrospective studies reported substantial rates of pelvic recurrence if high-risk patients were treated without radiotherapy, supporting the combined use of pelvic radiotherapy with adjuvant chemotherapy, as first explored in the RTOG 9708 phase 2 trial.8, 25, 26 Randomised studies have compared radiotherapy with the combination of chemotherapy and radiotherapy in patients with high-risk endometrial cancer. The NSGO-EC-9501/EORTC-55991 trial compared external beam radiotherapy alone with external beam radiotherapy and four cycles of platinum-based chemotherapy, given sequentially before or after external beam radiotherapy. A pooled analysis with the ManGO Iliade 3 trial27 with a total cohort of 534 patients showed, in line with our results, improved progression-free survival (78% vs 69%, p=0·01) and a trend for improved survival (82% vs 75%, p=0·07) with the addition of chemotherapy to radiotherapy alone.

fter external beam radiotherapy. A pooled analysis with the ManGO Iliade 3 trial27 with a total cohort of 534 patients showed, in line with our results, improved progression-free survival (78% vs 69%, p=0·01) and a trend for improved survival (82% vs 75%, p=0·07) with the addition of chemotherapy to radiotherapy alone. The schedule of combined radiotherapy with concurrent and adjuvant chemotherapy used in the PORTEC-3 trial seemed likely to be most effective because both treatments were started early after surgery and thus maximum benefit of the combination could be expected. In RTOG-9708, 4-year overall survival was 85% and disease-free survival was 81%. A retrospective single institution study28 reporting on 40 patients with stage IIIA or IIIC endometrial cancer treated with the same combination of chemoradiotherapy showed 5-year overall survival of 85% and relapse-free survival of 79%. In the chemoradiotherapy group of the PORTEC-3 trial the 5-year overall survival probability was 82% and the failure-free survival probability was 76%, thus confirming these results in a much larger trial. Overall and relapse-free survival in the pooled NSGO-EC-9501/EORTC-55991/Iliade trials27 were also similar at 82% and 78%, with only 20% patients with stage III disease.

e 5-year overall survival probability was 82% and the failure-free survival probability was 76%, thus confirming these results in a much larger trial. Overall and relapse-free survival in the pooled NSGO-EC-9501/EORTC-55991/Iliade trials27 were also similar at 82% and 78%, with only 20% patients with stage III disease. High-risk endometrial cancer is heterogeneous, including various histological types and stages of disease. In the NSGO-EC-9501/EORTC-55991 trial, although progression-free survival was significantly improved for patients with endometrioid endometrial cancer, this improvement was not found for serous and clear cell cancers. A Gynecologic Oncology Group (GOG) study29 explored the associations between histology and outcome in advanced or recurrent endometrial cancer patients in chemotherapy trials in 1203 patients. Although serous and clear cell cancers had a worse prognosis than other histological types, no difference in benefit from chemotherapy was found. In the PORTEC-3 trial, women with serous or clear-cell cancers had at least as much improvement in failure-free survival with the addition of chemotherapy as women with endometriod endometrial cancer did. When comparing serous cancers with other histological types, as expected, worse overall survival and failure-free survival were found for serous cancers; patients with serous cancers had a failure-free survival benefit with chemoradiotherapy, but this benefit was not significant in view of the small numbers of serous cancers and events.

ous cancers with other histological types, as expected, worse overall survival and failure-free survival were found for serous cancers; patients with serous cancers had a failure-free survival benefit with chemoradiotherapy, but this benefit was not significant in view of the small numbers of serous cancers and events. The multivariable analysis indicated that women older than 70 years seemed to have a greater failure-free survival benefit from chemotherapy than younger women. Age is a well-known risk factor for endometrial cancer and a greater benefit of chemotherapy in older women has been reported previously.7, 30 Although selection of fitter older women in this randomised trial might have occurred, physicians should not be reticent to counsel older women about the possible benefits of combined chemotherapy and radiotherapy.

ndometrial cancer and a greater benefit of chemotherapy in older women has been reported previously.7, 30 Although selection of fitter older women in this randomised trial might have occurred, physicians should not be reticent to counsel older women about the possible benefits of combined chemotherapy and radiotherapy. Analysed by stage, patients with stage III endometrial cancer who have the highest frequency of recurrence, also had the greatest absolute benefit from the combined treatment. The smaller failure-free survival improvement for patients with stage I–II disease seems not to outweigh the cost in terms of toxicity and quality-of-life impairment. Pelvic control was high (91%) with radiotherapy alone. This finding is in line with the results of the GOG-249 trial, in which patients with stage I and II endometrial cancer with high-intermediate or high-risk factors were randomly assigned to pelvic radiotherapy alone or to chemotherapy (three cycles of carboplatin and paclitaxel) followed by vaginal brachytherapy. No superiority of three cycles chemotherapy plus vaginal brachytherapy over external bean radiotherapy alone was found, with overlapping progression-free and overall survival curves and significantly more pelvic and para-aortic recurrences in the chemotherapy group.31, 32

paclitaxel) followed by vaginal brachytherapy. No superiority of three cycles chemotherapy plus vaginal brachytherapy over external bean radiotherapy alone was found, with overlapping progression-free and overall survival curves and significantly more pelvic and para-aortic recurrences in the chemotherapy group.31, 32 In 47% of all patients, a vaginal brachytherapy boost was given (46% for chemoradiotherapy vs 48% for radiotherapy); the majority because of cervical involvement and 4% because of other reasons, such as LVSI or grade 3 endometrial cancer. This finding is in line with other studies among patients with stage II–III endometrial cancer.33 Although the addition of a brachytherapy boost might have added to the good local control seen in both groups, we do not expect this would have affected the results, because the proportion of women receiving a brachytherapy boost was equal in the two treatment groups.

ng patients with stage II–III endometrial cancer.33 Although the addition of a brachytherapy boost might have added to the good local control seen in both groups, we do not expect this would have affected the results, because the proportion of women receiving a brachytherapy boost was equal in the two treatment groups. To compare the radiotherapy and chemotherapy schedule as used in the PORTEC-3 trial with chemotherapy alone for advanced stage endometrial cancer, the GOG-258 trial randomly assigned participants to receive chemoradiotherapy or six cycles of carboplatin and paclitaxel.34 Final results are pending, but a presented abstract34 reported no differences in overall or recurrence-free survival, while significantly more vaginal and pelvic or para-aortic recurrences were reported in patients treated with chemotherapy alone. Similar results were also reported in a retrospective multicentre study35 of 265 patients with stage IIIC disease treated with chemotherapy, radiotherapy, or both. Patients treated with chemotherapy alone were two to seven times more likely to develop a vaginal recurrence (35%) than those treated with radiotherapy (18%) or chemoradiotherapy (5%), and twice as likely to develop an isolated pelvic recurrence (18% vs 9% vs 7%). These outcomes confirm the importance of combined radiotherapy and chemotherapy to maximise vaginal and pelvic control and relapse-free survival. Furthermore, acute gastrointestinal and genitourinary toxicity of pelvic radiotherapy will be reduced with the current standard use of intensity-modulated radiotherapy.36

%). These outcomes confirm the importance of combined radiotherapy and chemotherapy to maximise vaginal and pelvic control and relapse-free survival. Furthermore, acute gastrointestinal and genitourinary toxicity of pelvic radiotherapy will be reduced with the current standard use of intensity-modulated radiotherapy.36 PORTEC-3 was a multicentre trial with strong international collaboration among six participating groups and, therefore, highly representative of current practice worldwide. Upfront pathology review was done to include only truly high-risk patients in the trial. Analysis of pathology review in the Netherlands and the UK (48% of PORTEC-3 participants) revealed that 8·3% of patients did not fulfil the eligibility criteria after central pathology review. These patients did not enter the trial.15

ront pathology review was done to include only truly high-risk patients in the trial. Analysis of pathology review in the Netherlands and the UK (48% of PORTEC-3 participants) revealed that 8·3% of patients did not fulfil the eligibility criteria after central pathology review. These patients did not enter the trial.15 A limitation of this trial might be that because of the death and failure-free survival event rates were lower than expected at the time of trial design, the required number of overall survival events was not reached and the final analysis was time-based rather than event-based, with final analysis at a median follow-up of 5 years (42 months after inclusion of the last patient). The number of overall survival events was 136 (69% of the required number of overall survival events), and the number of failure-free survival events was 186 (94% of the required events). The non-significant difference in 5-year overall survival of 5% found in PORTEC-3 was smaller than the study was powered to detect, and overall survival and failure-free survival probabilities were higher than expected from previous studies. Long-term outcomes will be analysed, especially for overall survival.

atient counselling remains essential. Translational studies of molecular risk factors and tumour characteristics with the tumour samples of the PORTEC-3 participants might identify those who could most benefit from chemotherapy or targeted agents and individualise treatment of women with high-risk endometrial cancer.38 In conclusion, although treatment with chemoradiotherapy significantly improved 5-year failure-free survival for patients with high-risk endometrial cancer compared with radiotherapy alone, there was no significant difference in overall survival. For women with stage III endometrial cancer, a significant improvement in failure-free survival was found. For each patient, the cost in terms of increased toxicity and longer treatment duration should be weighed against the benefit in terms of improvement in failure-free survival. Because pelvic control was high with radiotherapy alone, this chemoradiotherapy schedule cannot be recommended as a new standard for patients with stage I–II endometrial cancer. However, in view of the higher risk of recurrence among women with stage III disease, this chemoradiotherapy schedule should be considered to maximise failure-free survival, and benefits and risks should be individually discussed. For the study protocol see http://www.msbi.nl/portec3 This online publication has been corrected. The corrected version first appeared at thelancet.com/oncology on March 28, 2018 Supplementary Material Supplementary appendix

In conclusion, although treatment with chemoradiotherapy significantly improved 5-year failure-free survival for patients with high-risk endometrial cancer compared with radiotherapy alone, there was no significant difference in overall survival. For women with stage III endometrial cancer, a significant improvement in failure-free survival was found. For each patient, the cost in terms of increased toxicity and longer treatment duration should be weighed against the benefit in terms of improvement in failure-free survival. Because pelvic control was high with radiotherapy alone, this chemoradiotherapy schedule cannot be recommended as a new standard for patients with stage I–II endometrial cancer. However, in view of the higher risk of recurrence among women with stage III disease, this chemoradiotherapy schedule should be considered to maximise failure-free survival, and benefits and risks should be individually discussed. For the study protocol see http://www.msbi.nl/portec3 This online publication has been corrected. The corrected version first appeared at thelancet.com/oncology on March 28, 2018 Supplementary Material Supplementary appendix Acknowledgments The PORTEC-3 study was supported by a grant from the Dutch Cancer Society (UL2006–4168/CKTO 2006–04; The Netherlands). The PORTEC 3 study was supported in the UK by CRUK (C7925/A8659). Participation in the PORTEC-3 trial by ANZGOG and TROG were supported by the NHMRC Project Grant 570894 (2008) and by a Cancer Australia Grant (awarded through the 2011 round of the Priority-driven Collaborative Cancer Research Scheme and funded by Cancer Australia). Participation by the Italian MaNGO group was partly supported by a grant from the Italian Medicines Agency AIFA (FARM84BCX2). Canadian participation in the PORTEC-3 trial was supported by the Canadian Cancer Research Institute (#015469 and #021039). We thank all the participating groups, their coordinating teams, principal investigators, staff, clinical research teams, and the women who participated in the trial. A complete list of the groups and participating centres is shown in the appendix (pp 1–2). The PORTEC-3 trial involved a strong international collaboration within the Gynaecological Oncology InterGroup (GCIG) and the support of the GCIG officers and member groups is gratefully acknowledged. We thank the members of the DSMB for their invaluable work and guidance throughout the duration of the trial. Finally, we would like to thank the statistical reviewer for their valuable statistical comments and suggestions. This study was presented in part at the Annual Meeting of the American Society of Clinical Oncology (Chicago, IL, USA; June 2–6, 2017).

eir invaluable work and guidance throughout the duration of the trial. Finally, we would like to thank the statistical reviewer for their valuable statistical comments and suggestions. This study was presented in part at the Annual Meeting of the American Society of Clinical Oncology (Chicago, IL, USA; June 2–6, 2017). Contributors CLC was the chief investigator of the trial. MEP, PBO, IMJ, HCK, HWN, LCHWL, VTHBMS, RAN, HP, and CLC were involved in conception and study design. SMdB, MEP, LM, DK, PB, CH, PBO, JAL, PK, AC, AF, M-HB, IMJ-S, HCK, HWN, GW, SB, SC, DP, CH, LCHWL, VTHBMS, NS, VD, RDA, RAN, AF, KWV, HP, and CLC were involved in the collection and assembly of the data. SMdB, RAN, KWV, HP, and CLC were responsible for data analysis and interpretation. SMdB and CLC were responsible for the preparation and writing of the manuscript. All authors contributed to the manuscript and approved the final manuscript. Declaration of interests We declare no competing interests.

Introduction Cancer is an important disease burden in children as it is not easily prevented, with known causes explaining only a small proportion of cases. In Europe, incidence rates range from 140–170 per million person-years (ie, number of incident cases divided by the number of person-years at risk) in populations younger than 15 years and from 180–240 per million in those aged 15–19 years.1 During the past three decades, incidence increased by about 1% per annum for all cancers combined and this increase affected most major diagnostic groups, including leukaemias, lymphomas, and CNS tumours.2 However, in the past decade, incidence appears to have stabilised overall and for the major diagnostic groups in European populations.3, 4, 5, 6, 7, 8, 9, 10 Research in context Evidence before this study We searched PubMed with no language restrictions to identify studies published since Jan 1, 2004, that investigated cancer incidence trends in children (age 0–14 years) and adolescents (age 15–19 years) in Europe and other high-income countries, using the search terms ‘child* cancer incidence trends’ and ‘adolescen* cancer incidence trends’. Evidence from retrieved studies suggested that after a substantial increase in childhood cancer incidence towards the end of the last century, the rate of increase has declined in the 2000s. However, the limitations of most of these studies, in terms of number of cases or timeframe, might have hampered the full understanding of the dynamisms behind the observed trends in this population. Added value of this study

We searched PubMed with no language restrictions to identify studies published since Jan 1, 2004, that investigated cancer incidence trends in children (age 0–14 years) and adolescents (age 15–19 years) in Europe and other high-income countries, using the search terms ‘child* cancer incidence trends’ and ‘adolescen* cancer incidence trends’. Evidence from retrieved studies suggested that after a substantial increase in childhood cancer incidence towards the end of the last century, the rate of increase has declined in the 2000s. However, the limitations of most of these studies, in terms of number of cases or timeframe, might have hampered the full understanding of the dynamisms behind the observed trends in this population. Added value of this study We used all quality data available in Europe for the full calendar years in the period 1991–2010 to evaluate incidence patterns and trends in children and adolescents. Overall, incidence continued to increase during the study period by 0·5% per year in children and 1·0% per year in adolescents. However, we observed some evidence of a deceleration in increasing cancer trends, modulated by age group, European region, and diagnostic group. Based on the large scale of all available European data, our findings add an authoritative account on the geographical patterns and temporal trends of cancer in children and adolescents in Europe. Implications of all the available evidence

We used all quality data available in Europe for the full calendar years in the period 1991–2010 to evaluate incidence patterns and trends in children and adolescents. Overall, incidence continued to increase during the study period by 0·5% per year in children and 1·0% per year in adolescents. However, we observed some evidence of a deceleration in increasing cancer trends, modulated by age group, European region, and diagnostic group. Based on the large scale of all available European data, our findings add an authoritative account on the geographical patterns and temporal trends of cancer in children and adolescents in Europe. Implications of all the available evidence The increasing trends might have resulted from improvements in cancer diagnosis or registration in this age group, although the influence of other factors cannot be excluded. The paucity of evidence of stabilisation of cancer incidence in Europe and the variability of the observed trends support the need for continued international monitoring and research into causal mechanisms. Leukaemias, lymphomas, and tumours of the CNS represent 70% of all cancers observed in European populations younger than 15 years and half of all cancers in those aged 15–19 years1 and therefore contribute considerably to the overall incidence. Changing incidence could result from numerous factors, including variations in diagnostic capacity, completeness of registration, population composition, and exposure to risk factors. Monitoring these trends is important for planning health-care delivery and for aetiological research.

iderably to the overall incidence. Changing incidence could result from numerous factors, including variations in diagnostic capacity, completeness of registration, population composition, and exposure to risk factors. Monitoring these trends is important for planning health-care delivery and for aetiological research. In this study, we documented and interpreted incidence trends in Europe for cancers diagnosed at ages 0–19 years using quality-assured population-based cancer registries. We focused on the 20-year period 1991–2010, assessing the trends separately in children (aged 0–14 years) and adolescents (aged 15–19 years), for all cancers combined and for the major diagnostic groups. Our results update the Automated Childhood Cancer Information System (ACCIS) studies, extending the previously reported period of observation by over a decade.2, 11 Methods Data acquisition We invited all population-based cancer registries operating in European countries (as defined by the UN Statistics Division12) and Cyprus to participate in this population-based registry study. We requested a listing of individual records of cancer cases, and the population in each calendar year by sex and age from official national sources, accompanied by specific information about the geographical and administrative area covered and registration practices.

articipate in this population-based registry study. We requested a listing of individual records of cancer cases, and the population in each calendar year by sex and age from official national sources, accompanied by specific information about the geographical and administrative area covered and registration practices. Information on cancer cases included coded data on sex, age, date of birth, date of diagnosis, tumour sequence number, primary site, morphology, behaviour, and the most valid basis of diagnosis. Most registries coded tumours according to the International Classification of Diseases for Oncology, Third Edition (ICD-O-313) as required, and International Classification of Diseases for Oncology, Second Edition14 codes were converted to ICD-O-3 codes. Subsequently, neoplasms were classified according to the International Classification of Childhood Cancer, Third Edition (ICCC-3).15 We examined individual records for internal consistency,16 verifying unlikely combinations of site with morphology, age or sex with tumour type, basis of diagnosis with morphology, and rare tumour entities. The proportions of cases microscopically verified, identified from death certificate only, or with unknown basis of diagnosis and unspecified morphology contributed to data evaluation. The ACCIS Scientific Committee assessed the distribution of new cases across years and plausibility of case mix and age distribution. Standardised tables, charts, statistics, lists of selected questioned records, and any relevant information provided by the registry or known from published sources were discussed at meetings of the ACCIS Scientific Committee, who decided on the inclusion of each dataset in the ACCIS database. Only datasets with high-quality data were eligible for inclusion in the analyses.

ed questioned records, and any relevant information provided by the registry or known from published sources were discussed at meetings of the ACCIS Scientific Committee, who decided on the inclusion of each dataset in the ACCIS database. Only datasets with high-quality data were eligible for inclusion in the analyses. The cancers included in the analyses were all malignant tumours diagnosed during the complete calendar years 1991–2010, in people younger than 20 years and resident in the contributing registration areas. Several cancer types were excluded because they were not eligible for registration in all participating registries or during the whole study period: myelodysplasias (ICD-O-3 M-codes starting with 998), pilocytic astrocytoma (ICD-O-3 code M-9421), non-melanoma skin cancer (ICCC-3 subgroup XIe and XIIb with site code C44), and carcinoid tumour of the appendix (ICD-O-3 site code C18.1 and M-8240). The data used in this study were submitted and validated during the years 2015–16. Our study design was reviewed and approved by the International Agency for Research on Cancer (IARC) Ethics Committee on June 17, 2015.

The cancers included in the analyses were all malignant tumours diagnosed during the complete calendar years 1991–2010, in people younger than 20 years and resident in the contributing registration areas. Several cancer types were excluded because they were not eligible for registration in all participating registries or during the whole study period: myelodysplasias (ICD-O-3 M-codes starting with 998), pilocytic astrocytoma (ICD-O-3 code M-9421), non-melanoma skin cancer (ICCC-3 subgroup XIe and XIIb with site code C44), and carcinoid tumour of the appendix (ICD-O-3 site code C18.1 and M-8240). The data used in this study were submitted and validated during the years 2015–16. Our study design was reviewed and approved by the International Agency for Research on Cancer (IARC) Ethics Committee on June 17, 2015. Dataset constitution An eligible registry could become a contributor if it provided quality data for all complete calendar years in the 20-year period 1991–2010. The 53 contributing registries in 19 countries (appendix p 2) included nine paediatric registries. The paediatric registries collected and provided data for those aged 0–14 years, and the other cancer registries and the paediatric registry of Belarus provided data for the full target age range (0–19 years). The French national paediatric registry registered haematological malignancies only.

ne paediatric registries. The paediatric registries collected and provided data for those aged 0–14 years, and the other cancer registries and the paediatric registry of Belarus provided data for the full target age range (0–19 years). The French national paediatric registry registered haematological malignancies only. Some populations in France, Germany, Italy, Spain, Switzerland, and the UK were covered by both paediatric and general cancer registries. To avoid double counting of cases in two registries while using the maximum number of cases for analyses in these populations, we allocated each registry to one or more datasets (appendix p 2). We built three datasets—one with 39 registries contributing to the analyses of all cancers, CNS tumours, and other tumours in ages 0–14 years, a second with 32 registries contributing to the analyses of leukaemia and lymphoma in ages 0–14 years, and a third with 45 registries contributing to all the analyses of those aged 15–19 years. The first and second datasets (ages 0–14 years) differed only in the contribution from France; the analyses of all cancers and CNS tumours used a dataset that included data from the French general cancer registries, while the leukaemia and lymphoma dataset included data from the French national paediatric registry of haematological malignancies. The French national registry increased the person-years by ten times and provided 11 770 more haematological malignancies compared with the combined contribution of the French regional registries.

e leukaemia and lymphoma dataset included data from the French national paediatric registry of haematological malignancies. The French national registry increased the person-years by ten times and provided 11 770 more haematological malignancies compared with the combined contribution of the French regional registries. Subnational numbers of cases and person-years were pooled to produce national cancer incidence and countries were further pooled into four European regions (appendix p 2) according to UN definitions.12 The person-years available in each region are shown in the appendix (p 3). Statistical analysis The number of incident cases in the covered areas of the participating registries during the study period determined our sample size. Incidence was calculated as the number of cases divided by the number of person-years in the categories of geographical area, sex, age, and diagnostic group for the given 20-year period and expressed per million person-years. As the age distribution of population at risk differs between countries and over time, we adjusted the reported incidence for the age range 0–14 years for age via direct standardisation using weights 12, 10, and 9 for the three age groups 0–4 years, 5–9 years, and 10–14 years, respectively.17 We also calculated 95% CIs. As each combination of calendar year, sex, and age group had a positive person-years count, we encountered no missing data.

the age range 0–14 years for age via direct standardisation using weights 12, 10, and 9 for the three age groups 0–4 years, 5–9 years, and 10–14 years, respectively.17 We also calculated 95% CIs. As each combination of calendar year, sex, and age group had a positive person-years count, we encountered no missing data. To graphically portray incidence trends for all cancers, we plotted observed incidence against calendar year for each country, and the rates for the four European regions were smoothed using a locally weighted regression18 of the incidence on year.

the age range 0–14 years for age via direct standardisation using weights 12, 10, and 9 for the three age groups 0–4 years, 5–9 years, and 10–14 years, respectively.17 We also calculated 95% CIs. As each combination of calendar year, sex, and age group had a positive person-years count, we encountered no missing data. To graphically portray incidence trends for all cancers, we plotted observed incidence against calendar year for each country, and the rates for the four European regions were smoothed using a locally weighted regression18 of the incidence on year. To assess the average annual percentage change, we fit the natural logarithm of the incidence with year using generalised linear regression models adjusting for age group and region, as appropriate. We quantified changes in incidence as the average annual percentage change, with corresponding 95% CIs. The null hypothesis corresponded to no change in the annual rate, which was equivalent to 0 lying within the 95% CI for the average annual percentage change. We analysed each cancer category, overall and by European region, using Stata (version 14). We examined incidence trends for changes during the study period using Joinpoint Regression Program (version 4.1.0) applied to the log rates, separately for the age groups 0–14 years and 15–19 years in the total dataset, and in each cancer category, overall and by region. The null hypothesis assumed the annual percentage change was constant throughout the study period. We used the permutation test19 to determine the number of joinpoints, by default set to a maximum of three for Europe and five for analyses adjusted for region. Where joinpoints were detected, we reported the annual percentage change with corresponding 95% CIs for each of the linear segments identified between two significant joinpoints.

tion test19 to determine the number of joinpoints, by default set to a maximum of three for Europe and five for analyses adjusted for region. Where joinpoints were detected, we reported the annual percentage change with corresponding 95% CIs for each of the linear segments identified between two significant joinpoints. Role of the funding source The funders of the study external to the collaborating institutions had no role in study design, data collection, data analysis, data interpretation, or writing of the report. MC, MMF, and ES-F had full access to all the data used in the study. The corresponding author had final responsibility for the decision to submit for publication.

study external to the collaborating institutions had no role in study design, data collection, data analysis, data interpretation, or writing of the report. MC, MMF, and ES-F had full access to all the data used in the study. The corresponding author had final responsibility for the decision to submit for publication. Results During the full calendar years of the period 1991–2010, the 53 registries contributed 1·3 billion person-years and 180 335 unique cases. 15 162 (8·4%) of 180 335 cases were excluded, and the proportion of excluded cases varied slightly between age groups and regions. We considered the quality indicators for the 165 173 included cases to be satisfactory (appendix p 3). In 2010, the available population at risk represented 53·2 million (46·6%) of 114·1 million of the European population aged 0–14 years and 9·8 million (22·6%) of 43·2 million aged 15–19 years, with reference to UN population estimates.20 In western Europe, coverage of the childhood population was almost complete for haematological malignancies (27·9 million [93·3%] of 29·9 million), and 17·5 million (58·5%) of 29·9 million were covered for the other malignancies, while 2·1 million (20·1%) of 10·6 million adolescents were covered (appendix p 3).

ates.20 In western Europe, coverage of the childhood population was almost complete for haematological malignancies (27·9 million [93·3%] of 29·9 million), and 17·5 million (58·5%) of 29·9 million were covered for the other malignancies, while 2·1 million (20·1%) of 10·6 million adolescents were covered (appendix p 3). We included all 118 265 eligible cases in patients aged 0–14 years in our analyses of incidence of all cancers. The average annual age-standardised incidence was 137·5 (95% CI 136·7–138·3) per million person-years and incidence increased significantly by 0·54% (0·44–0·65) per year on average (table 1), with no break in the time trend. There was a large decrease in incidence for the national registry of Belarus, contributing to the regional pattern (figure 1). The large fluctuation of annual incidence in Iceland was due to small population size and did not visibly affect the shape of the regional curve (figure 1). The trend in the east was split into two segments, both with non-significant trends (figure 2).Table 1 Number of new cases, world age-standardised incidence per million person-years, and average annual percentage change by diagnostic category in children aged 0–14 years, by European region, 1991–2010

curve (figure 1). The trend in the east was split into two segments, both with non-significant trends (figure 2).Table 1 Number of new cases, world age-standardised incidence per million person-years, and average annual percentage change by diagnostic category in children aged 0–14 years, by European region, 1991–2010 All cancers Leukaemia Lymphoma Malignant CNS tumours Other cancers Cases Incidence per million person-years (95% CI) Annual change (95% CI) Cases Incidence per million person-years (95% CI) Annual change (95% CI) Cases Incidence per million person-years (95% CI) Annual change (95% CI) Cases Incidence per million person-years (95% CI) Annual change (95% CI) Cases Incidence per million person-years (95% CI) Annual change (95% CI) East 19 274 137·9 (135·9 to 139·9) 0·50% (0·21 to 0·79) 5721 42·6 (41·5 to 43·8) 1·00% (0·44 to 1·55) 2652 17·1 (16·5 to 17·8) −1·32% (−2·08 to −0·56) 3367 23·6 (22·8 to 24·4) 0·34% (−0·32 to 1·01) 7534 54·5 (53·3 to 55·8) 0·72% (0·21 to 1·23) North 34 339 131·5 (130·1 to 132·9) 0·40% (0·16 to 0·64) 12 015 47·0 (46·1 to 47·8) 0·59% (0·19 to 1·00) 3848 13·5 (13·0 to 13·9) 0·74% (0·00 to 1·48) 5996 22·7 (22·1 to 23·2) −0·20% (−0·58 to 0·19) 12 480 48·4 (47·6 to 49·3) 0·41% (−0·00 to 0·83) South 14 450 149·9 (147·4 to 152·3) 0·40% (0·07 to 0·72) 4472 47·3 (45·9 to 48·7) 0·68% (0·17 to 1·18) 2137 20·4 (19·5 to 21·2) 0·29% (−0·54 to 1·11) 2317 23·9 (22·9 to 24·8) −0·44% (−1·24 to 0·37) 5524 58·3 (56·8 to 59·9) 0·57% (0·06 to 1·08) West 50 202 138·2 (136·9 to 139·4) 0·70% (0·52 to 0·88) 26 250 47·8 (47·2 to 48·3) 0·59% (0·37 to 0·82) 9821 15·8 (15·5 to 16·1) 0·44% (0·08 to 0·81) 7733 21·0 (20·5 to 21·4) 1·35% (0·79 to 1·93) 18 168 51·2 (50·5 to 52·0) 0·62% (0·36 to 0·89) Europe 118 265 137·5 (136·7 to 138·3) 0·54% (0·44 to 0·65) 48 458 46·9 (46·5 to 47·3) 0·66% (0·48 to 0·84) 18 458 15·8 (15·6 to 16·0) 0·26% (−0·01 to 0·54) 19 413 22·2 (21·9 to 22·6) 0·49% (0·20 to 0·77) 43 706 51·7 (51·2 to 52·2) 0·56% (0·40 to 0·72) Figure 1 Incidence trends of cancer in children aged 0–14 years in Europe, 1991–2010

37·5 (136·7 to 138·3) 0·54% (0·44 to 0·65) 48 458 46·9 (46·5 to 47·3) 0·66% (0·48 to 0·84) 18 458 15·8 (15·6 to 16·0) 0·26% (−0·01 to 0·54) 19 413 22·2 (21·9 to 22·6) 0·49% (0·20 to 0·77) 43 706 51·7 (51·2 to 52·2) 0·56% (0·40 to 0·72) Figure 1 Incidence trends of cancer in children aged 0–14 years in Europe, 1991–2010 Jagged thin lines indicate annual age-standardised rates in countries and smooth red thick lines indicate modelled incidence trends in regions. Figure 2 Overview of observed incidence trends of cancer in children aged 0–14 years for the entire study period and for any time segments identified in joinpoint analysis, by diagnostic group and region of Europe, 1991–2010 Red indicates an increasing trend, blue indicates a decreasing trend, and white indicates no significant trend. Absence of bars indicates no joinpoint was identified for a given category.

Figure 2 Overview of observed incidence trends of cancer in children aged 0–14 years for the entire study period and for any time segments identified in joinpoint analysis, by diagnostic group and region of Europe, 1991–2010 Red indicates an increasing trend, blue indicates a decreasing trend, and white indicates no significant trend. Absence of bars indicates no joinpoint was identified for a given category. We investigated incidence for four diagnostic groups in patients aged 0–14 years (table 1, figure 2, appendix p 4), and included all eligible cases. The combined age-standardised incidence of leukaemia based on 48 458 cases in patients aged 0–14 years was 46·9 (95% CI 46·5–47·3) per million person-years and increased by 0·66% (0·48–0·84) per year, with no change in slope. We observed an increase in each region, although only during the first decade in the north. The overall age-standardised incidence of lymphoma was 15·8 (15·6–16·0) per million person-years (based on 18 458 cases) and varied modestly by region. Incidence increased in the west and decreased in the east, and we detected no change in the slopes. The overall age-standardised incidence of malignant CNS tumours (19 413 cases) of 22·2 (95% CI 21·9–22·6) per million person-years was observed to rise at 0·49% (0·20–0·77) per annum with no joinpoints; by region, we observed an increasing trend only in the west, as the rates were stable in the three other regions.

es. The overall age-standardised incidence of malignant CNS tumours (19 413 cases) of 22·2 (95% CI 21·9–22·6) per million person-years was observed to rise at 0·49% (0·20–0·77) per annum with no joinpoints; by region, we observed an increasing trend only in the west, as the rates were stable in the three other regions. In patients aged 0–14 years, the combined incidence of all 43 706 other cancers increased in Europe overall and in all regions except the north. In joinpoint analysis, the European trend increased only over the first few years, similar to the trend in the west (figure 2). In adolescents (aged 15–19 years), the combined European incidence was 176·2 (95% CI 174·4–178·0) per million person-years based on all 35 138 eligible cases and incidence increased by 0·96% (95% CI 0·73–1·19) per year in the period 1991–2010 (table 2), although no significant change in trend was seen in the second decade (figure 3). We investigated incidence trends by country and region, and regional trends by diagnostic group (figure 4, appendix p 5). In addition to large variations due to small numbers of cases for some countries, we observed a large range in incidence (between around 140 and 240 per million person-years) for Belarus. The slopes of trends in the regions were uneven, except in the south. In the east, a decrease was noted over the time segment 1999–2008 (figure 3).Table 2 Number of new cases, age-specific incidence per million person-years, and average annual percentage change by diagnostic category in adolescents aged 15–19 years, by European region, 1991–2010

regions were uneven, except in the south. In the east, a decrease was noted over the time segment 1999–2008 (figure 3).Table 2 Number of new cases, age-specific incidence per million person-years, and average annual percentage change by diagnostic category in adolescents aged 15–19 years, by European region, 1991–2010 All cancers Leukaemia Lymphoma Malignant CNS tumours Other cancers Cases Incidence per million person-years (95% CI) Annual change (95% CI) Cases Incidence per million person-years (95% CI) Annual change (95% CI) Cases Incidence per million person-years Annual change (95% CI) Cases Incidence per million person-years (95% CI) Annual change (95% CI) Cases Incidence per million person-years (95% CI) Annual change (95% CI) East 8041 169·6 (165·9 to 173·4) 1·23% (0·62 to 1·84) 1023 21·6 (20·3 to 22·9) 1·79% (0·43 to 3·17) 2096 44·2 (42·3 to 46·1) 0·68% (0·10 to 1·26) 749 15·8 (14·7 to 16·9) −0·09% (−1·23 to 1·07) 4173 88·0 (85·4 to 90·7) 1·65% (0·77 to 2·54) North 15 207 167·2 (164·5 to 169·8) 0·62% (0·28 to 0·96) 2192 24·1 (23·1 to 25·1) 0·08% (−0·62 to 0·78) 3981 43·8 (42·4 to 45·1) 0·82% (0·12 to 1·52) 1358 14·9 (14·1 to 15·7) −0·91% (−2·48 to 0·68) 7676 84·4 (82·5 to 86·3) 0·97% (0·65 to 1·28) South 3941 197·0 (190·8 to 203·1) 1·78% (1·36 to 2·20) 501 25·0 (22·8 to 27·2) 2·03% (0·66 to 3·42) 1244 62·2 (58·7 to 65·6) 2·13% (1·26 to 3·01) 288 14·4 (12·7 to 16·1) 0·94% (−0·71 to 2·62) 1908 95·4 (91·1 to 99·6) 1·61% (0·80 to 2·42) West 7949 193·7 (189·4 to 197·9) 1·08% (0·72 to 1·43) 986 24·0 (22·5 to 25·5) 1·35% (0·33 to 2·38) 2125 51·8 (49·6 to 54·0) 1·59% (0·67 to 2·52) 588 14·3 (13·2 to 15·5) −0·10% (−1·43 to 1·25) 4250 103·5 (100·4 to 106·6) 0·95% (0·51 to 1·39) Europe 35 138 176·2 (174·4 to 178·0) 0·96% (0·73 to 1·19) 4702 23·6 (22·9 to 24·3) 0·93% (0·49 to 1·37) 9446 47·4 (46·4 to 48·3) 1·04% (0·65 to 1·44) 2983 15·0 (14·4 to 15·5) −0·40% (−1·19 to 0·39) 18 007 90·3 (89·0 to 91·6) 1·17% (0·82 to 1·53) Figure 3 Overview of observed incidence trends of cancer in adolescents aged 15–19 years for the entire study period and for any significant time segments identified in joinpoint analysis, by diagnostic group and region of Europe, 1991–2010

to 15·5) −0·40% (−1·19 to 0·39) 18 007 90·3 (89·0 to 91·6) 1·17% (0·82 to 1·53) Figure 3 Overview of observed incidence trends of cancer in adolescents aged 15–19 years for the entire study period and for any significant time segments identified in joinpoint analysis, by diagnostic group and region of Europe, 1991–2010 Red indicates an increasing trend, blue indicates a decreasing trend, and white indicates no significant trend. Absence of bars indicates no joinpoint was identified for a given category. Figure 4 Incidence trends of cancer in adolescents aged 15–19 years in Europe, 1991–2010 Jagged thin lines indicate annual age-specific rates in countries and smooth thick red lines indicate modelled incidence trends in regions.

Red indicates an increasing trend, blue indicates a decreasing trend, and white indicates no significant trend. Absence of bars indicates no joinpoint was identified for a given category. Figure 4 Incidence trends of cancer in adolescents aged 15–19 years in Europe, 1991–2010 Jagged thin lines indicate annual age-specific rates in countries and smooth thick red lines indicate modelled incidence trends in regions. For adolescents, the overall incidence of leukaemia was 23·6 (95% CI 22·9–24·3) per million person-years (based on all 4702 eligible cases) and varied little between regions (table 2). Incidence increased except in the north, where it was fairly stable, resulting in an average annual change of 0·93% (0·49–1·37). No change in the slopes was detected. Lymphoma (9446 cases) occurred more frequently than did leukaemia (4702 cases) or malignant CNS tumours (2983 cases) in adolescents compared with children (table 2), and the overall incidence of 47·4 (46·4–48·3) per million person-years varied moderately by region. An increase in incidence was observed for all European data combined and in all European regions investigated. Changes in this trend were detected in the east and west, where the increase was limited to the second and first half of the study period, respectively (figure 3). The overall age-specific incidence of malignant CNS tumours was 15·0 (14·4–15·5) per million person-years. Incidence was stable overall and in the four defined regions (figure 3), with no change in the slopes.

est, where the increase was limited to the second and first half of the study period, respectively (figure 3). The overall age-specific incidence of malignant CNS tumours was 15·0 (14·4–15·5) per million person-years. Incidence was stable overall and in the four defined regions (figure 3), with no change in the slopes. The overall rate for all other cancers in adolescents (18 007 cases) was 90·3 per million. The European avearge annual change of 1·17% (0·82–1·53) differed between three segments—among them a significant increase was detected during the first decade only. The pattern of changes in the slopes by region was similar to that observed for all cancers (figure 3). Discussion In this population-based registry study, we reported on cancer incidence trends in the European population of children (aged 0–14 years) and adolescents (aged 15–19 years) during the full calendar years 1991–2010, thus extending the ACCIS study by 10–132, 11 years of observation. The overall incidence for children and adolescents increased significantly over this 20-year period.

r incidence trends in the European population of children (aged 0–14 years) and adolescents (aged 15–19 years) during the full calendar years 1991–2010, thus extending the ACCIS study by 10–132, 11 years of observation. The overall incidence for children and adolescents increased significantly over this 20-year period. The increase in incidence was not constant by cancer type, region, or over time, and there was a suggestion of stabilisation in the trend for all cancers in adolescents in the dataset for the whole of Europe (appendix p 6). The most pronounced inter-regional diversities included the decreasing incidence of childhood lymphomas in the east (compared with stable or increasing incidence in the other regions), an increase in the incidence of childhood CNS tumours in the west (compared with little change observed in the other regions), and the stable incidence of leukaemia in adolescents in the north (relative to increasing incidence in the other regions).

t (compared with stable or increasing incidence in the other regions), an increase in the incidence of childhood CNS tumours in the west (compared with little change observed in the other regions), and the stable incidence of leukaemia in adolescents in the north (relative to increasing incidence in the other regions). Although our results are largely confirmatory and incremental from previous findings, a major strength of our study is its large size, which permits detection of a moderate rate of increase of 0·5% per year in the age group 0–14 years. Data from large paediatric cancer registries helped to increase the coverage and creation of a specific dataset for childhood haematological malignancies enlarged the relevant person-years in the west (583·8 million) by 53% compared with data available for other neoplasms (380·5 million). We examined the trends separately in children and adolescents to allow for variations in trends between these populations with a different case mix. We validated the quality and completeness of the data from the included registries during a thorough dataset assessment. We consider our results to provide the best available estimates of incidence trends in the European population of children and adolescents for 1991–2010.

Although our results are largely confirmatory and incremental from previous findings, a major strength of our study is its large size, which permits detection of a moderate rate of increase of 0·5% per year in the age group 0–14 years. Data from large paediatric cancer registries helped to increase the coverage and creation of a specific dataset for childhood haematological malignancies enlarged the relevant person-years in the west (583·8 million) by 53% compared with data available for other neoplasms (380·5 million). We examined the trends separately in children and adolescents to allow for variations in trends between these populations with a different case mix. We validated the quality and completeness of the data from the included registries during a thorough dataset assessment. We consider our results to provide the best available estimates of incidence trends in the European population of children and adolescents for 1991–2010. A limitation of our study is the variable coverage of the regions, which might affect the representativeness of the incidence at the national and European regional levels. The available data do not allow speculation on what the incidence would be if coverage was complete—increase of coverage and quality of registration is required for a full picture of the cancer burden. Meanwhile, by use of all quality information available for the study period, our results maximise the number of person-years of comparable data, and provide evidence of changing trends in European populations.

complete—increase of coverage and quality of registration is required for a full picture of the cancer burden. Meanwhile, by use of all quality information available for the study period, our results maximise the number of person-years of comparable data, and provide evidence of changing trends in European populations. We excluded several cancer types that are usually included in reports of childhood cancer incidence from our analyses. The excluded cancer types are classified by the ICCC-314 and contribute to the total cancer burden, and it is therefore preferable that they are registered. In particular, exclusion of non-malignant CNS tumours might have shaped the observed trends. We plan to examine this assumption in a focused analysis of trends of all CNS tumours, by behaviour and diagnostic subgroup, and by considering the differences and changes in coding and their reportability. Our reported total incidence was 5% to 15% lower (depending on the region) than in another report assessing the period 2001–10.1 This disparity is explained by the exclusion of several groups of neoplasms from our study and the period starting one decade earlier, which would, in the presence of an increasing trend, result in a lower overall rate.

was 5% to 15% lower (depending on the region) than in another report assessing the period 2001–10.1 This disparity is explained by the exclusion of several groups of neoplasms from our study and the period starting one decade earlier, which would, in the presence of an increasing trend, result in a lower overall rate. The overall average annual change of 0·54% in children aged 0–14 years is lower than the 1% reported by an ACCIS study based on diagnoses in the 1970s through to the 1990s2 or for the period 1978–97,21 and might indicate a deceleration in the increase of cancer incidence, although the registry datasets included differ somewhat between the studies. A lower rate of increase could suggest the end of an improvement in reporting, but might also reflect reduced reporting discipline, a change in classification of cancers, or other factors. Cancer-specific studies might provide more precise interpretation of these trends.

ts included differ somewhat between the studies. A lower rate of increase could suggest the end of an improvement in reporting, but might also reflect reduced reporting discipline, a change in classification of cancers, or other factors. Cancer-specific studies might provide more precise interpretation of these trends. Studies in European populations of smaller sizes reported stabilising of cancer incidence in children overall and in the three major diagnostic groups.3, 4, 5, 6, 7, 8, 9, 10 By contrast, data collected in the longstanding registries of the Surveillance, Epidemiology, and End Results (SEER) Program suggest continued minor increases in overall childhood cancer incidence over the most recently assessed period 1995–2014,22 consistent with our study. In adolescents aged 15–19 years, varying incidence trends were observed in different European populations for leukaemia, lymphoma, or CNS tumours.3, 4, 7 As for the SEER data,22 our study found an increase in overall cancer incidence and leukaemia and lymphoma incidence, but no change in trends for malignant CNS tumours. A large US study23 reported stable incidence over the period 2001–09 overall and for the three major diagnostic groups in the population younger than 20 years. We identified changes in overall trends that provide some evidence of a deceleration of time trends during the most recent years of our study, along with varying patterns across age groups, regions, and cancer types, although significant joinpoints should not be taken to imply abrupt changes in underlying trends in risk.19 The variability we found warrants continued monitoring of incidence patterns in large populations over long periods, as trends that are not significant over short intervals might result in significant increases over a longer timeframe, and grouping smaller populations with no trend could yield a significant change in a pooled dataset. These considerations might also explain the discrepancy in the measured rate of change between our study and smaller European datasets.3, 4, 5, 6, 7, 8, 9, 10

ght result in significant increases over a longer timeframe, and grouping smaller populations with no trend could yield a significant change in a pooled dataset. These considerations might also explain the discrepancy in the measured rate of change between our study and smaller European datasets.3, 4, 5, 6, 7, 8, 9, 10 Kroll and colleagues24 have linked the increase in childhood cancer incidence in Great Britain with advances in diagnostic technology and improved cancer registration, because of the concurrence of a step increase in leukaemia and non-CNS solid tumour incidence in 2002 with a registration plan enacted in 2001. Similarly, the rise in the incidence of CNS tumours in 1992 coincided with the introduction of innovative methods of diagnosis. As a result, childhood cancer might have been under-reported in Great Britain during parts of the period assessed in this study. Although we are not aware of similar studies in other countries, analogous effects on childhood cancer incidence cannot be excluded elsewhere in Europe and could have affected our results.

s. As a result, childhood cancer might have been under-reported in Great Britain during parts of the period assessed in this study. Although we are not aware of similar studies in other countries, analogous effects on childhood cancer incidence cannot be excluded elsewhere in Europe and could have affected our results. The gradual convergence of environmental factors, lifestyle, and health services across Europe might have contributed to the similarities in incidence trends across the regions, although such changes will probably have less of an effect on cancer in childhood than in older age. Nevertheless, some opposing trends (notably for lymphoma in the east) suggest caution should be taken in ascribing the slowly increasing trends exclusively to improved detection, at least before further detailed analyses by cancer type, age group, and geographical region are done.

cancer in childhood than in older age. Nevertheless, some opposing trends (notably for lymphoma in the east) suggest caution should be taken in ascribing the slowly increasing trends exclusively to improved detection, at least before further detailed analyses by cancer type, age group, and geographical region are done. Overall incidence in children is weighted towards leukaemia, which, in 75% of cases, is the precursor B-cell lymphoid leukaemia. The age-specific curve for this cancer type currently peaks around age 2–4 years in most high-income countries. With increasing social development over time, this peak shifts towards younger ages and becomes more pronounced.2, 25 This change is unlikely to be driven by improved registration, which would affect all ages in the same way. The deficit of lymphoid leukaemia cases seen in poorer settings can be ascribed in part to underdiagnosis,26, 27 but inequalities across social strata might also operate through relevant exposures, such as parental occupation, diet, or reproduction characteristics.28 Continuous socioeconomic development and increasing awareness of primary care practitioners might have contributed to the observed increase in leukaemia cases, although reasons for the stabilisation of leukaemia incidence in northern Europe are unclear.

h as parental occupation, diet, or reproduction characteristics.28 Continuous socioeconomic development and increasing awareness of primary care practitioners might have contributed to the observed increase in leukaemia cases, although reasons for the stabilisation of leukaemia incidence in northern Europe are unclear. Incidence of lymphoma was relatively high compared with incidence of leukaemia in children in eastern Europe. Considering the likely infectious origin of lymphoma at young ages,29 our observed decreasing incidence could be attributed to elimination of certain viral exposures. Furthermore, diagnosis of lymphoma might be delayed until older ages, especially if the decreasing trend in children is accompanied by an increasing trend in adolescents, as seen in the east over the second decade. Possible selective under-reporting could also be considered, as the incidence of leukaemia and other cancers continues to increase in children in the east. The high incidence in southern Europe is consistent with previous reports of possible environmental exposures.30 A further assessment of lymphoma trends by diagnostic subgroups, narrower age groups, and sex could help provide a more specific explanation of the observed temporal changes.

ncrease in children in the east. The high incidence in southern Europe is consistent with previous reports of possible environmental exposures.30 A further assessment of lymphoma trends by diagnostic subgroups, narrower age groups, and sex could help provide a more specific explanation of the observed temporal changes. The stability of malignant CNS tumour incidence in adolescents is encouraging, although drawing firm conclusions should be postponed until incidence trends are examined using data that include non-malignant tumours in the populations in which these are ascertained completely to exclude the possible effect of changes in classification of tumours. In children, because of the shown persisting increase in incidence in malignant CNS tumours, and the exclusion of non-malignant tumours—specifically pilocytic astrocytoma—from our analyses, a further detailed assessment is warranted, especially considering the poor outcome for such tumours.

n classification of tumours. In children, because of the shown persisting increase in incidence in malignant CNS tumours, and the exclusion of non-malignant tumours—specifically pilocytic astrocytoma—from our analyses, a further detailed assessment is warranted, especially considering the poor outcome for such tumours. The trends for all other cancers combined show that the overall increase in incidence is not explained solely by changes in the three major groups of childhood and adolescent cancer. This fact is especially relevant in adolescents, as many cases are germ cell tumours and carcinomas.1 The candidate explanatory groups for the increase in at least a part of the period are thyroid carcinoma,31 testicular tumours,32 and melanoma.33 In particular, the patterns for Belarus are probably shaped by the pronounced increase and later waning of thyroid cancer incidence during the study period, reflecting exposure to the radioactive fallout from the Chernobyl accident.34 These observations also merit detailed studies.

testicular tumours,32 and melanoma.33 In particular, the patterns for Belarus are probably shaped by the pronounced increase and later waning of thyroid cancer incidence during the study period, reflecting exposure to the radioactive fallout from the Chernobyl accident.34 These observations also merit detailed studies. In summary, in this study we aimed to provide an overview of cancer incidence trends in children and adolescents in Europe, as an introduction to a detailed investigation of patterns and trends and their possible determinants by diagnostic group and other subcategories in a concerted series of studies. We found a continued increase in cancer incidence in children and adolescents in Europe over the 20-year period 1991–2010. The diversity of trends across the regions and tumour groups warrants further monitoring and a more detailed assessment of the high-quality data collected by cancer registries, by specific tumour types and population subgroups. Whether the observed increase in cancer incidence in children and adolescents is due to enhanced discovery of cases or changing risk factors, the known frequency of cancer in young people should be considered in cancer control programmes. Without an accurate quantification of the possible causes of this increase, whether real or artifactual, it is prudent to continue aetiological research into the causes of cancer in children and adolescents and to explore possible preventive measures. Supplementary Material Supplementary appendix

In summary, in this study we aimed to provide an overview of cancer incidence trends in children and adolescents in Europe, as an introduction to a detailed investigation of patterns and trends and their possible determinants by diagnostic group and other subcategories in a concerted series of studies. We found a continued increase in cancer incidence in children and adolescents in Europe over the 20-year period 1991–2010. The diversity of trends across the regions and tumour groups warrants further monitoring and a more detailed assessment of the high-quality data collected by cancer registries, by specific tumour types and population subgroups. Whether the observed increase in cancer incidence in children and adolescents is due to enhanced discovery of cases or changing risk factors, the known frequency of cancer in young people should be considered in cancer control programmes. Without an accurate quantification of the possible causes of this increase, whether real or artifactual, it is prudent to continue aetiological research into the causes of cancer in children and adolescents and to explore possible preventive measures. Supplementary Material Supplementary appendix Acknowledgments We gratefully acknowledge the cofunders of this study. Data acquisition was partly supported by the Federal Ministry of Health of the Federal German Government and the development of infrastructure relevant to this study was enabled through the EU's Seventh Framework Programme (FP7/2007–2013) under grant agreement LSSH–CT–2008–21 9453 (EUROCOURSE). The required funding was completed by the International Agency for Research on Cancer.

of Health of the Federal German Government and the development of infrastructure relevant to this study was enabled through the EU's Seventh Framework Programme (FP7/2007–2013) under grant agreement LSSH–CT–2008–21 9453 (EUROCOURSE). The required funding was completed by the International Agency for Research on Cancer. Contributors ES-F, BL, PK, JWC, RP-B, and CAS designed the study, and contributed to data acquisition and evaluation. MC managed the acquired data and prepared evaluation material. ES-F, MMF, BL, MP, IS, FB, JWC, RP-B, and CAS contributed to the detailed plan of data analysis. MF and MC did data analyses and prepared relevant output. ES-F did the literature search and drafted the manuscript, which was reviewed, modified, and approved by all coauthors. Declaration of interests We declare no competing interests.

Introduction Medulloblastoma, the most common malignant childhood brain tumour, is now recognised as an umbrella term for different molecular pathological disease entities. These entities differ in their progenitor cells, characteristic mutations, biological profiles, and clinical behaviour. Currently, WHO classification of CNS tumours recognises four distinct genetically defined entities (WNT, SHH-TP53wild-type, SHH-TP53mut, and non-WNT/non-SHH).1 Non-WNT/non-SHH medulloblastoma encompasses Group3 and Group4, which were defined by epigenetic and mRNA expression signatures2 and are considered provisional variants by the 2016 WHO classification.1 Understanding the molecular pathology and clinical relevance of medulloblastoma subtypes provides substantial opportunities for personalised risk-adapted therapies. Research in context Evidence before this study

Introduction Medulloblastoma, the most common malignant childhood brain tumour, is now recognised as an umbrella term for different molecular pathological disease entities. These entities differ in their progenitor cells, characteristic mutations, biological profiles, and clinical behaviour. Currently, WHO classification of CNS tumours recognises four distinct genetically defined entities (WNT, SHH-TP53wild-type, SHH-TP53mut, and non-WNT/non-SHH).1 Non-WNT/non-SHH medulloblastoma encompasses Group3 and Group4, which were defined by epigenetic and mRNA expression signatures2 and are considered provisional variants by the 2016 WHO classification.1 Understanding the molecular pathology and clinical relevance of medulloblastoma subtypes provides substantial opportunities for personalised risk-adapted therapies. Research in context Evidence before this study International consensus and the 2016 WHO classification recognise the following distinct clinico-molecular disease entities in medulloblastoma: WNT, SHH-TP53wild-type, SHH-TP53mut, and non-WNT/non-SHH (encompassing Group3 and Group4). Standard-risk, non-infant disease (with 75–85% 5-year progression-free survival and affecting 50–60% of patients) represents the largest clinical treatment group of patients. The ongoing pan-European SIOP PNET 5 MB clinical trial defines standard-risk, non-infant disease as the absence of high-risk clinical features such as metastatic disease or subtotal resection, molecular features (MYC or MYCN amplification or TP53 mutation in SHH medulloblastoma), and histological characteristics (large-cell/anaplastic disease). These definitions were established based on previous disease-wide studies. The SIOP PNET 5 MB trial is investigating reduced-intensity therapies for patients classified as standard-risk with expected good prognosis (ie, WNT medulloblastoma), aimed at maintaining overall survival while minimising late toxicities. However, biomarkers that stratify risk within remaining standard-risk patients with non-WNT medulloblastoma have not been identified. Moreover, novel non-WNT/non-SHH medulloblastoma epigenetic subtypes have been recognised; however, these subtypes remain to be validated and implemented clinically. Our own reviews of the literature formed the foundation for the present study; we did not carry out any formal literature searches before the study start date (December, 2015).

NT/non-SHH medulloblastoma epigenetic subtypes have been recognised; however, these subtypes remain to be validated and implemented clinically. Our own reviews of the literature formed the foundation for the present study; we did not carry out any formal literature searches before the study start date (December, 2015). Added value of this study

NT/non-SHH medulloblastoma epigenetic subtypes have been recognised; however, these subtypes remain to be validated and implemented clinically. Our own reviews of the literature formed the foundation for the present study; we did not carry out any formal literature searches before the study start date (December, 2015). Added value of this study To our knowledge, HIT-SIOP PNET 4 is the only completed pan-European clinical trial in patients with standard-risk medulloblastoma. However, to date, systematically collected biological material remaining from this trial was not amenable to contemporary molecular analysis. Application of novel methods to enable assessment of this cohort, and investigation of an independent demographically matched standard-risk medulloblastoma validation cohort, allowed derivation and validation of biomarker-driven, risk-stratification models on the basis of the molecular pathology of standard-risk medulloblastoma, including a novel whole chromosomal cytogenetic aberration signature within standard-risk non-WNT/non-SHH medulloblastoma. These newly described whole chromosomal cytogenetic aberration signatures allowed reallocation of more than 50% of HIT-SIOP PNET 4 patients with standard-risk medulloblastoma into a favourable-risk group, while the remaining patients were classified as high risk. Therefore, findings from this study resolve current patients with standard-risk medulloblastoma into biomarker-defined distinct favourable-risk and high-risk groups, and represent a substantial step in our ability to risk stratify and clinically manage medulloblastoma.

e remaining patients were classified as high risk. Therefore, findings from this study resolve current patients with standard-risk medulloblastoma into biomarker-defined distinct favourable-risk and high-risk groups, and represent a substantial step in our ability to risk stratify and clinically manage medulloblastoma. Implications of all the available evidence The results of this study redefine the concepts of risk stratification in standard-risk medulloblastoma, providing insight into its molecular subtypes, their underpinning biology, and clinical application. Stratification of standard-risk medulloblastoma by use of the biomarkers and validated schemes we describe could allow assignment of 150–200 patients per year in Europe into a favourable-risk group, and such patients could benefit from reduction of treatment intensity. Patients not classified as favourable-risk should be considered high-risk and might benefit from treatment intensification. The molecular risk groups and biomarker schemes presented in this study are amenable to routine diagnostic assessment and provide a foundation for future clinical trials and research investigations.

tients not classified as favourable-risk should be considered high-risk and might benefit from treatment intensification. The molecular risk groups and biomarker schemes presented in this study are amenable to routine diagnostic assessment and provide a foundation for future clinical trials and research investigations. Discovery and validation of clinically meaningful medulloblastoma features in previous clinical trial cohorts have driven advances in the clinical management of the disease. Children younger than 16 years of age at diagnosis with WNT-activated medulloblastomas have consistently achieved favourable outcomes (5-year event-free survival >95%),3, 4 whereas other disease features, including MYC or MYCN amplification, large-cell/anaplastic histology, metastatic disease, or subtotal resection, define high-risk disease (5-year event-free survival <60%).5 These disease features now underpin risk-adapted therapies in ongoing biomarker-driven international prospective clinical studies, such as the SIOP PNET 5 MB (NCT02066220) and SJMB12 (NCT01878617) trials, which aim to improve outcomes through reduced-intensity therapies for favourable-risk patients and randomised assessment of adapted therapies in the remaining patients.

apies in ongoing biomarker-driven international prospective clinical studies, such as the SIOP PNET 5 MB (NCT02066220) and SJMB12 (NCT01878617) trials, which aim to improve outcomes through reduced-intensity therapies for favourable-risk patients and randomised assessment of adapted therapies in the remaining patients. Standard-risk medulloblastoma represents the predominant clinical treatment group (around 60% of patients) and is defined by the absence of clinical, molecular, and histopathological high-risk features. This group encompasses tumours of all variants except high-risk SHH-TP53mut.6, 7 Diagnosis of favourable-risk, WNT disease (around 20% of patients with standard-risk medulloblastoma) provides a clear precedent for therapy de-escalation within clinical trials. By contrast, patients with non-WNT, standard-risk medulloblastoma have heterogeneous outcomes (5-year event-free survival around 75%), and further actionable risk groups are yet to be identified or validated to the point of clinical application. The favourable risk of patients with standard-risk, SHH-TP53wild-type medulloblastoma6, 7 identified in retrospective series requires validation in clinical trials, and reproducible and clinically significant molecular pathological features within non-WNT/non-SHH tumours remain to be defined. Research has found that Group3 and Group4 medulloblastomas represent heterogeneous, biologically overlapping entities—few recurrent mutations have been observed, whole chromosomal cytogenetic aberrations are common,8, 9, 10, 11 and evidence of novel molecular subtypes is emerging.6, 12, 13

tumours remain to be defined. Research has found that Group3 and Group4 medulloblastomas represent heterogeneous, biologically overlapping entities—few recurrent mutations have been observed, whole chromosomal cytogenetic aberrations are common,8, 9, 10, 11 and evidence of novel molecular subtypes is emerging.6, 12, 13 To our knowledge, HIT-SIOP PNET 414 is the first completed, international, prospective clinical trial of non-metastatic childhood medulloblastoma (patients aged 4–21 years at diagnosis) and this cohort of patients represents a unique opportunity to explore the molecular pathology of standard-risk medulloblastoma, its potential for risk stratification, and the development of new therapeutic concepts. Trial participants were postoperatively staged and randomly assigned to treatment with standard or hyperfractionated radiotherapy, followed by chemotherapy with eight cycles of cisplatin, lomustine, and vincristine. No difference in event-free survival was observed between the two treatment groups.14

rapeutic concepts. Trial participants were postoperatively staged and randomly assigned to treatment with standard or hyperfractionated radiotherapy, followed by chemotherapy with eight cycles of cisplatin, lomustine, and vincristine. No difference in event-free survival was observed between the two treatment groups.14 Formalin-fixed, paraffin-embedded (FFPE) tumour material for biological studies was prospectively collected, which enabled confirmation of favourable outcomes in patients with WNT medulloblastoma (defined by immunohistochemistry [IHC]) and identification of chromosome 17 imbalances on a diploid background (17p loss and/or 17q gain, by fluorescence in-situ hybridisation [FISH]) as a marker of poor prognosis.15 However, until now, contemporary molecular characterisation of the HIT-SIOP PNET 4 cohort, and assessment of its clinical relevance, has been restricted by the low quality and quantity of remaining tumour material.

(17p loss and/or 17q gain, by fluorescence in-situ hybridisation [FISH]) as a marker of poor prognosis.15 However, until now, contemporary molecular characterisation of the HIT-SIOP PNET 4 cohort, and assessment of its clinical relevance, has been restricted by the low quality and quantity of remaining tumour material. In this Article, we report comprehensive molecular and pathological characterisation of the HIT-SIOP PNET 4 cohort using novel technologies16, 17 developed and adapted for assessment of the remnant tumour material. This analysis, alongside an independent, demographically matched, standard-risk medulloblastoma validation cohort, enabled the discovery and validation of concerted whole chromosomal aberration signatures with prognostic value for patients with non-WNT/non-SHH medulloblastoma. We describe the development of risk stratification models for standard-risk, non-WNT/non-SHH disease, which might allow reassignment of all patients with standard-risk medulloblastoma into biomarker-defined favourable-risk or high-risk groups.

res with prognostic value for patients with non-WNT/non-SHH medulloblastoma. We describe the development of risk stratification models for standard-risk, non-WNT/non-SHH disease, which might allow reassignment of all patients with standard-risk medulloblastoma into biomarker-defined favourable-risk or high-risk groups. Methods Study design and participants In this retrospective analysis, we assessed remaining tumour samples from patients from the HIT-SIOP PNET 4 trial (NCT01351870).14 Between Jan 1, 2001, and Dec 31, 2006, 338 patients were recruited from 120 different treatment centres in seven European countries (France, Germany, Italy, the Netherlands, Spain, Sweden, and the UK; appendix p 3). The study investigated treatment outcomes in patients aged 4–21 years using either hyperfractionated radiotherapy or standard delivery radiotherapy followed by chemotherapy.1 Standard delivery radiotherapy comprised 23·4 Gy to the craniospinal axis and 54 Gy to the whole posterior fossa, and was given over 42 days in 30 fractions of 1·8 Gy each day for 5 days per week. Hyperfractionated radiotherapy was given in 68 fractions at 1·0 Gy twice per day with an 8 h interval between fractions, given over 48 days. The total craniospinal dose was 36 Gy, and the whole posterior fossa dose was 60 Gy, with a further boost to 68 Gy to the tumour bed. Adjuvant chemotherapy was started 6 weeks after the end of radiotherapy. Eight cycles of cisplatin (70 mg/m2 intravenously) and lomustine (75 mg/m2) on day 1, and vincristine (1·5 mg/m2 intravenously) on days 1, 8, and 15, were given with a 6 week interval between each cycle.14

further boost to 68 Gy to the tumour bed. Adjuvant chemotherapy was started 6 weeks after the end of radiotherapy. Eight cycles of cisplatin (70 mg/m2 intravenously) and lomustine (75 mg/m2) on day 1, and vincristine (1·5 mg/m2 intravenously) on days 1, 8, and 15, were given with a 6 week interval between each cycle.14 Minute remnant material (cytospin-concentrated cellular nuclei preparations) or tumour sections, originally intended for FISH and IHC,15 were available for analysis (samples from 147 patients). We retained tumours from patients with subtotally resected disease18 or categorised as MYCN-amplified to assess their prognostic value in a clinically controlled cohort.6, 11, 15 We excluded MYC-amplified tumours because of their established poor prognosis.5 136 tumour samples met these criteria and underwent molecular investigation. The demographics of the patients who provided these tumour samples (clinical and molecular cohort) and their prognostic features were consistent with the whole trial cohort (table).Table Clinical and molecular characteristics of all cohorts

.5 136 tumour samples met these criteria and underwent molecular investigation. The demographics of the patients who provided these tumour samples (clinical and molecular cohort) and their prognostic features were consistent with the whole trial cohort (table).Table Clinical and molecular characteristics of all cohorts Clinical cohort Clinical and molecular standard-risk cohort All patients in HIT-SIOP PNET 4 (n=338) All subgroups in HIT-SIOP PNET 4 (n=136) Non-WNT/non-SHH in HIT-SIOP PNET 4 (n=91) Non-WNT/non-SHH in validation cohort (Newcastle; n=70) Sex Male 211 (62%) 81 (60%) 61 (67%) 50 (71%) Female 127 (38%) 55 (40%) 30 (33%) 20 (29%) Male:female ratio 1·66:1 1·5:1 2:1 2·5:1 Age at diagnosis (years)* 9·0 (3–20) [7·0–12·0] 9·0 (3–20) [7·0–12·0] 8·0 (4–20) [6·0–10·0] 8·5 (4–18) [8·8–11·4] Treatment Standard radiotherapy 169 (50%) 67 (49%) 43 (47%) 66 (94%) Hyperfractionated radiotherapy 169 (50%) 69 (51%) 48 (53%) 4 (6%) Histology Classic 273 (81%) 111 (82%) 81 (89%) 64 (91%) Desmoplastic/nodular 47 (14%) 25 (18%) 10 (11%) 6 (9%) Large-cell/anaplastic 16 (5%)† 0 0 0 No review 2 (1%) 0 0 0 Resection Gross total resection 286 (90%) 121 (92%) 81 (92%) 54 (80%) Subtotal resection 31 (10%) 10 (8%) 7 (8%) 14 (20%) Follow-up (years) 6·6 (5·6–8·5) 6·7 (5·6–8·4) 6·7 (5·8–8·2) 5·6 (3·1–8·1) Collection era (years) 2001–06 2001–06 2001–06 1990–2014‡ Molecular subgroup WNT ·· 28 (21%) 0 0 SHH ·· 17 (13%) 0 0 Group3 ·· 15 (11%) 15 (16%) 6 (9%) Group4 ·· 76 (56%) 76 (84%) 64 (91%) β-catenin immunohistochemistry Total assessed ·· 121 56 28 Nuclear accumulation ·· 30 (25%) 0 1 (4%) Normal ·· 91 (75%) 56 (100%) 27 (96%) CTNNB1 mutation Total assessed ·· 114 51 56 Mutant ·· 26 (23%) 0 0 Wild-type ·· 88 (77%) 51 (100%) 56 (100%) TP53 mutation in SHH Total assessed ·· 15 0 0 SHH-TP53wild-type ·· 11 (73%) 0 0 SHH-TP53mut ·· 4 (27%) 0 0 MYC amplification Amplified ·· 0 0 0 Not amplified ·· 136 (100%) 91 (100%) 70 (100%) MYCN amplification Amplified ·· 10 (7%) 10 (11%) 6 (9%) Not amplified ·· 126 (93%) 81 (89%) 64 (91%) Chromosome 17 (interphase fluorescence in-situ hybridisation) Total assessed ·· 101 69 17p loss or 17q gain (diploid(cen)) ·· 17 (17%) 15 (22%) NA Others ·· 84 (83%) 54 (78%) NA Data are n (%), median (IQR) or n, unless otherwise indicated. Some percentages do not total 100 because of non-assessable tumours. NA=not analysed.

some 17 (interphase fluorescence in-situ hybridisation) Total assessed ·· 101 69 17p loss or 17q gain (diploid(cen)) ·· 17 (17%) 15 (22%) NA Others ·· 84 (83%) 54 (78%) NA Data are n (%), median (IQR) or n, unless otherwise indicated. Some percentages do not total 100 because of non-assessable tumours. NA=not analysed. * Data are median (range) [IQR]. † The trial was amended in 2003 to exclude cases with large-cell/anaplastic histology. ‡ Median year of diagnosis 2006. We validated and extended our findings in a second independent, demographically matched, retrospective cohort of patients with non-WNT/non-SHH standard-risk medulloblastoma (n=70) collected at UK Children's Cancer and Leukaemia Group and European Society for Paediatric Oncology (SIOPE) associated treatment centres between 1990 and 2014. Patients in this cohort received equivalent therapies (maximal surgical resection [all patients], adjuvant craniospinal radiotherapy [all patients; standard radiotherapy in variable doses—low dose: 24–27 Gy, 39 patients; high dose: 35–39 Gy, 27 patients; hyperfractionated radiotherapy variable doses: 32·4 Gy craniospinal radiotherapy plus 23·4 Gy boost, one patient; 60 Gy hyperfractionated accelerated radiotherapy, one patient; 31/59 Gy, one patient; and 39/54, one patient], and chemotherapy [65 (93%) of 70 patients]).

Gy, 39 patients; high dose: 35–39 Gy, 27 patients; hyperfractionated radiotherapy variable doses: 32·4 Gy craniospinal radiotherapy plus 23·4 Gy boost, one patient; 60 Gy hyperfractionated accelerated radiotherapy, one patient; 31/59 Gy, one patient; and 39/54, one patient], and chemotherapy [65 (93%) of 70 patients]). Written informed consent for tumour collection for biological studies was obtained from patients or their parents. Tumour investigations were done with approval from Newcastle and North Tyneside Research Ethics Committee (study reference 07/Q0905/71)—all tumour material was collected in accordance with this approval. Procedures Because only material of mostly low quantity and quality was available, the HIT-SIOP PNET 4 samples were unsuitable for subgroup assessment using conventional approaches (DNA methylation array19 or mRNA expression analysis by Nanostring20); therefore, we analysed all samples using a mass spectrometry-minimal methylation classifier (MS-MIMIC) assay to assess their molecular subgroup.16 For the validation cohort, samples were of sufficient quality and quantity to do Illumina 450k DNA methylation microarray (62 DNA samples were from frozen material and eight were from FFPE tissue) and consensus methylation subgroup was assigned as described previously.6

IMIC) assay to assess their molecular subgroup.16 For the validation cohort, samples were of sufficient quality and quantity to do Illumina 450k DNA methylation microarray (62 DNA samples were from frozen material and eight were from FFPE tissue) and consensus methylation subgroup was assigned as described previously.6 We assessed amplification of MYC and MYCN oncogenes by interphase FISH15 and estimated gene copy numbers from molecular inversion probe and DNA methylation arrays,19 as previously described. We analysed mutations in exons 4–9 of TP53 and exon 3 of CTNNB1 with Sanger sequencing as previously described.21 We assessed mutations in APC using a customised next-generation DNA sequencing panel (Illumina; San Diego, CA, USA) in samples with CTNNB1wild-type WNT medulloblastoma. We used a molecular inversion probe array (335 000 inversion probes; version 2.0; Affymetrix; Santa Clara, CA, USA) to identify aberrant changes in genomic copy number in samples from the HIT-SIOP PNET 4 trial.17 Raw molecular inversion probe data were analysed using Nexus Copy Number 7.0 Discovery Edition (BioDiscovery; El Segundo, CA, USA). We used SNP-FASST2 segmentation algorithm to make copy number and loss of heterozygosity estimations. We used GISTIC (Genomic Identification of Significant Targets in Cancer, v 1.0) to identify focal chromosomal aberrations (appendix pp 10–12).22 We analysed the validation cohort samples on the Illumina 450k DNA methylation microarray (Illumina; San Diego, CA, USA), and estimated chromosomal and focal copy number changes by use of the R package conumee v 1.13.0, as previously described.6 We defined a whole chromosomal aberration group of patients by hierarchical clustering of recurrent (ie, >15%) aberrations.

Illumina 450k DNA methylation microarray (Illumina; San Diego, CA, USA), and estimated chromosomal and focal copy number changes by use of the R package conumee v 1.13.0, as previously described.6 We defined a whole chromosomal aberration group of patients by hierarchical clustering of recurrent (ie, >15%) aberrations. Event-free survival was defined as the time from surgery to first event (progression or relapse), or date of last follow-up. Patients whose follow-up time exceeded 10 years were right-censored at 10 years. Clinical follow-up data were collected according to the HIT-SIOP PNET 4 trial protocol.14 For the validation cohort, clinical data were collected in the same format from individual treatment centres. Statistical analysis Using hierarchical clustering, we clustered samples classified as non-WNT/non-SHH medulloblastoma subtype by their recurrent whole chromosomal aberration (ie, incidence >15%; appendix p 2). After molecular subgrouping, we observed similar cytogenetic changes and event-free survival between the non-WNT/non-SHH medulloblastomas subclassified as Group3 and Group4 (appendix p 9). Because of these results and the emerging evidence of their shared biology,6, 12 we considered these groups together in subsequent event-free survival analyses. To test the null hypothesis that event-free survival was not associated with clinical, molecular, or pathological variables in patients with Group3 or Group4 medulloblastoma, we constructed Kaplan-Meier curves and compared patient groups with log-rank tests.

Statistical analysis Using hierarchical clustering, we clustered samples classified as non-WNT/non-SHH medulloblastoma subtype by their recurrent whole chromosomal aberration (ie, incidence >15%; appendix p 2). After molecular subgrouping, we observed similar cytogenetic changes and event-free survival between the non-WNT/non-SHH medulloblastomas subclassified as Group3 and Group4 (appendix p 9). Because of these results and the emerging evidence of their shared biology,6, 12 we considered these groups together in subsequent event-free survival analyses. To test the null hypothesis that event-free survival was not associated with clinical, molecular, or pathological variables in patients with Group3 or Group4 medulloblastoma, we constructed Kaplan-Meier curves and compared patient groups with log-rank tests. Using Cox modelling, we tested the prognostic value of clinical markers (gender, radiotherapy type [hyperfractionated vs standard], resection outcome [subtotal vs fully-resected disease], MYCN amplification [yes vs no], histology type [desmoplastic/nodular vs classic histology]), cytogenetic markers (recurrent whole chromosomal aberration [presence vs absence]), and cumulative numbers of total whole chromosomal aberrations (gains vs losses). We verified the proportionality assumption for Cox modelling using scaled Schoenfeld residuals. We derived pragmatic assignments of patient risk by combining whole chromosomal aberrations that were significantly different in univariate testing to define risk groups and assessed their predictive value by calculating total area under the curve (AUC), sensitivity, and specificity at 5 years since diagnosis (appendix p 2). We clustered the tumour samples from the validation cohort by the recurrent whole chromosomal aberrations previously identified in the HIT-SIOP PNET 4 trial standard-risk, non-WNT/non-SHH medulloblastoma cohort and validated the derived risk stratification schemes.

specificity at 5 years since diagnosis (appendix p 2). We clustered the tumour samples from the validation cohort by the recurrent whole chromosomal aberrations previously identified in the HIT-SIOP PNET 4 trial standard-risk, non-WNT/non-SHH medulloblastoma cohort and validated the derived risk stratification schemes. Finally, to better understand the nature of the identified risk groups, we classified the validation cohort according to the recently published refinements of epigenetically defined substructures within non-WNT/non-SHH medulloblastoma.6, 12 Validation cohort samples were assigned to subgroup variants according to these published studies and visualised using t-distributed stochastic neighbour embedding (appendix pp 2–3). We set the significance threshold at p<0·05 for all statistical tests in this study, and two-tailed p values are reported. We assessed significance of association using Fisher's exact test, and visualised the strength of associations using χ2 test residuals. Further detailed methods are provided in the appendix pp 1–3. Statistical and bioinformatic analyses were done with R (version 3.4.2). Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Further detailed methods are provided in the appendix pp 1–3. Statistical and bioinformatic analyses were done with R (version 3.4.2). Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results We successfully assessed methylation subgroup (appendix p 4), genome-wide copy number aberrations (appendix p 4), and mutational features in 136 tumour samples from the HIT-SIOP PNET 4 cohort, recruited from 2001 to 2006 and representing 136 (40%) of 338 patients in the trial.14 Cohort clinical and molecular characteristics are summarised in figure 1 and the table.Figure 1 Clinical and disease-associated molecular features of the HIT-SIOP PNET 4 cohort All 147 patient samples available from the HIT-SIOP PNET 4 cohort with subgroup information are shown, including 11 samples without data on chromosomal aberrations. Black indicates positivity for an assessed feature (for sex, black indicates male and white indicates female). Grey indicates missing data. Red indicates chromosomal losses and blue indicates chromosomal gains. NA=not assessed. Residual scores from χ2 tests of association are shown (darker shades of grey indicate stronger enrichment) alongside p values from Fisher's exact tests.

black indicates male and white indicates female). Grey indicates missing data. Red indicates chromosomal losses and blue indicates chromosomal gains. NA=not assessed. Residual scores from χ2 tests of association are shown (darker shades of grey indicate stronger enrichment) alongside p values from Fisher's exact tests. Our integrative analysis found the expected distributions of clinical, pathological, and molecular features within WHO-defined medulloblastoma entities and their provisional sub-variants1 in the HIT-SIOP PNET 4 cohort (CTNNB1 mutation and chromosome 6 monosomy in WNT medulloblastoma; desmoplastic/nodular pathology and TP53 mutation in SHH medulloblastoma; i17q in non-WNT/non-SHH-medulloblastoma; and SNCAIP duplication and MYCN amplification in Group4; figure 1). Patients with WNT medulloblastoma (28 [21%] of 136) and patients with Group4 disease (76 [56%] of 136) were enriched in HIT-SIOP PNET 4 standard-risk medulloblastoma compared with retrospective disease-wide series,8, 9, 10, 11 as anticipated following exclusion of children younger than 4 years, adults older than 21 years, and patients with high-risk or metastatic disease from this cohort. The prognostic relevance and demographic distribution of key clinical features across the study cohorts were compatible with our previous reports of the entire HIT-SIOP PNET 4 trial (table).14, 15

Our integrative analysis found the expected distributions of clinical, pathological, and molecular features within WHO-defined medulloblastoma entities and their provisional sub-variants1 in the HIT-SIOP PNET 4 cohort (CTNNB1 mutation and chromosome 6 monosomy in WNT medulloblastoma; desmoplastic/nodular pathology and TP53 mutation in SHH medulloblastoma; i17q in non-WNT/non-SHH-medulloblastoma; and SNCAIP duplication and MYCN amplification in Group4; figure 1). Patients with WNT medulloblastoma (28 [21%] of 136) and patients with Group4 disease (76 [56%] of 136) were enriched in HIT-SIOP PNET 4 standard-risk medulloblastoma compared with retrospective disease-wide series,8, 9, 10, 11 as anticipated following exclusion of children younger than 4 years, adults older than 21 years, and patients with high-risk or metastatic disease from this cohort. The prognostic relevance and demographic distribution of key clinical features across the study cohorts were compatible with our previous reports of the entire HIT-SIOP PNET 4 trial (table).14, 15 With a median follow-up of 6·7 years (IQR 5·6–8·4) in the HIT-SIOP PNET 4 cohort, 5-year event-free survival was equivalent between patients who received standard radiotherapy and those who received hyperfractionated radiotherapy (hazard ratio [HR] 0·81, 95% CI 0·36–1·82; p=0·61; figure 2), while patients who had a subtotal resection had a poorer event-free survival at 5 years than those who underwent gross total resection (HR 3·18, 1·08–9·37; p=0·036; figure 2). We found no differences in terms of 5-year event-free survival between the four methylation subgroups (figure 2; WNT 5-year event-free survival 88·5%, 95% CI 77·0–100; Group4 5-year event-free survival 81·6%, 73·3–90·8; Group3 5-year event-free survival 80·0, 62·1–100; SHH 5-year event-free survival 75·3%, 56·9–99·6; WNT vs Group4 HR 0·61, 95% CI 0·18–2·12, p=0·44; SHH vs Group4 1·27, 0·42–3·86, p=0·68; Group3 vs Group4 1·13, 0·32–3·94, p=0·85). Group3 and Group4 had very similar event-free survival curves (figure 2). By contrast, we found a significant association between the presence of whole chromosomal aberrations and favourable event-free survival outcomes compared with the absence of whole chromosomal aberrations (HR 4·05, 95% CI 1·79–9·13; p=0·00077; figure 2).Figure 2 Event-free survival in the HIT-SIOP PNET 4 cohort by clinical and disease-associated molecular features

n between the presence of whole chromosomal aberrations and favourable event-free survival outcomes compared with the absence of whole chromosomal aberrations (HR 4·05, 95% CI 1·79–9·13; p=0·00077; figure 2).Figure 2 Event-free survival in the HIT-SIOP PNET 4 cohort by clinical and disease-associated molecular features Patients (n=136) were grouped as (A) treated with standard radiotherapy vs hyperfractionated radiotherapy, (B) those who had a gross total resection vs subtotal resection, (C) classified as per the four consensus medulloblastoma molecular subgroups, and (D) those with or without whole chromosomal aberrations. Event-free survival for patients with non-WNT/non-SHH disease (n=91) grouped as (E) patients with MYCN amplified vs non-amplified tumours, (F) patients with medulloblastomas presenting an i17q or not, and (G) patients with medulloblastomas with or without whole chromosomal aberration. HR=hazard ratio.

omal aberrations. Event-free survival for patients with non-WNT/non-SHH disease (n=91) grouped as (E) patients with MYCN amplified vs non-amplified tumours, (F) patients with medulloblastomas presenting an i17q or not, and (G) patients with medulloblastomas with or without whole chromosomal aberration. HR=hazard ratio. We investigated the clinical, molecular, and event-free survival characteristics of WHO-defined medulloblastoma molecular entities in the HIT-SIOP PNET 4 cohort.1 25 (89%) of 28 WNT tumours showed the characteristic chromosome 6 monosomy and few other aberrations (appendix pp 5–6). We identified CTNNB1 mutations in 26 (93%) of 28 WNT tumours (appendix pp 5–6). Both CTNNB1 wild-type tumours showed a copy neutral loss of heterozygosity within chromosome 5q (APC) and we identified APC frameshift deletions (E1309fs ΔAAAAG and Q1062fs ΔACAAA). Outcomes within the WNT subgroup were age-dependent. We observed a 5-year event-free survival of 100% in patients younger than 16 years at diagnosis, and all WNT relapses (three [11%] of 28 WNT tumours) occurred in patients aged 16–20 years (p=0·00050; appendix pp 5–6).

ft deletions (E1309fs ΔAAAAG and Q1062fs ΔACAAA). Outcomes within the WNT subgroup were age-dependent. We observed a 5-year event-free survival of 100% in patients younger than 16 years at diagnosis, and all WNT relapses (three [11%] of 28 WNT tumours) occurred in patients aged 16–20 years (p=0·00050; appendix pp 5–6). Tumours classified as SHH in the HIT-SIOP PNET 4 cohort (17 [13%] of 136 patients) also had few whole chromosomal aberrations (appendix pp 7–8). Chromosome 17p loss (TP53) and TP53 mutations were associated with each other (p=0·0090; appendix pp 7–8) and with worse event-free survival. All four (100%) of four events (relapses) were in patients with TP53 mutation or chromosome 17p loss (p=0·0036). We did not observe MYCN amplifications in tumours classified as SHH medulloblastoma, including TP53mut tumours. A previously reported SHH disease risk model (of chromosome 14 loss and GLI2 amplification)11 showed significantly worse event-free survival for patients in this cohort (p=0·00067; appendix pp 7–8).

=0·0036). We did not observe MYCN amplifications in tumours classified as SHH medulloblastoma, including TP53mut tumours. A previously reported SHH disease risk model (of chromosome 14 loss and GLI2 amplification)11 showed significantly worse event-free survival for patients in this cohort (p=0·00067; appendix pp 7–8). The 91 (67%) non-WNT/non-SHH tumours in the HIT-SIOP PNET 4 cohort of 136 were characterised by a higher incidence of whole chromosomal aberrations (eg, chromosome 7 gain, and chromosome 8 and 11 loss; mean 5·3 [SD 4·81] whole chromosomal aberrations per case for non-WNT/non-SHH vs 1·82 [SD 2·56] for WNT and 1·76 [SD 2·00] for SHH; Figure 1, Figure 3, appendix p 9), but isolated chromosome arm alterations were rare, with the exception of i17q (56 [62%] of 91 non-WNT/non-SHH medulloblastomas). However, 16 (18%) of 91 cases had no whole chromosomal aberrations (Figure 1, Figure 2). As expected, we observed structural cytogenetic (eg, i17q) and focal aberrations (including MYCN amplifications, OTX2, CCND2, and 18q12 [TPTE2] gains or amplifications, SNCAIP duplications, and 13q11–12 [SETBP1] loss; figure 1; appendix pp 11–12). Moreover, previously reported prognostic factors (MYCN amplification, i17q alterations, and subtotal resection)11, 18 were not associated with worse event-free survival (figure 2; appendix p 15), while the observed cohort-wide prognostic significance of whole chromosomal aberrations was maintained in this subgroup (figure 2).Figure 3 Identification of two cytogenetically distinct subgroups within non-WNT/non-SHH standard-risk medulloblastoma

t associated with worse event-free survival (figure 2; appendix p 15), while the observed cohort-wide prognostic significance of whole chromosomal aberrations was maintained in this subgroup (figure 2).Figure 3 Identification of two cytogenetically distinct subgroups within non-WNT/non-SHH standard-risk medulloblastoma All 91 patient samples with non-WNT/non-SHH standard-risk medulloblastoma available from HIT-SIOP PNET 4 cohort are shown. (A) The frequency of p, q, and whole chromosome gains and losses for all autosomal chromosomes. (B) Unsupervised hierarchical clustering of chromosomal features. Grey indicates missing data. Residuals from χ2 indicate where whole chromosomal aberration cytogenetic group enrichment has occurred (darker shades of grey indicate stronger relationships), alongside p values from Fisher's exact tests. Total numbers of whole chromosomal losses (red), gains (blue), and changes (black) are shown. Increasing colour intensity indicates a larger number of changes. Chromosomal changes with incidence >15% are shown. We defined whole chromosomal aberration cytogenetic groups by hierarchical clustering. Green represents Group4 medulloblastoma and yellow represents Group3 medulloblastoma. (C) Correlation plot for recurrent (>15% incidence) cytogenetic changes. Circle area is proportional to the strength of correlation, with positive correlations shown in red and negative correlations shown in blue.

archical clustering. Green represents Group4 medulloblastoma and yellow represents Group3 medulloblastoma. (C) Correlation plot for recurrent (>15% incidence) cytogenetic changes. Circle area is proportional to the strength of correlation, with positive correlations shown in red and negative correlations shown in blue. We next investigated whether the observed molecular heterogeneity within the 91 non-WNT/non-SHH medulloblastoma tumours could inform its biological basis and clinical behaviour. Through unsupervised hierarchical cluster analysis of recurrent whole chromosomal aberrations, we identified two clinically and biologically distinct subgroups of tumours (figure 3). The first cytogenetic group was strongly associated with a pattern of i17q in isolation, diploid karyotypes, few recurrent whole chromosomal aberrations, and more relapses (p=0·00084). The second cytogenetic group was characterised by a spectrum of multiple recurrent and co-incident whole chromosomal aberrations (figure 3) and aneuploidy (p<0·0001; appendix p 13), and was associated with fewer relapses.

n, diploid karyotypes, few recurrent whole chromosomal aberrations, and more relapses (p=0·00084). The second cytogenetic group was characterised by a spectrum of multiple recurrent and co-incident whole chromosomal aberrations (figure 3) and aneuploidy (p<0·0001; appendix p 13), and was associated with fewer relapses. Whole chromosomal aberrations within non-WNT/non-SHH medulloblastoma samples were associated with improved 5-year event-free survival (figure 4). 55 (60%) of 91 non-WNT/non-SHH tumours had multiple recurrent and co-incident whole chromosomal aberrations and showed favourable outcomes compared with those without whole chromosomal aberrations (HR 0·16, 95% CI 0·05–0·50; p=0·0015; figure 4). The total number of whole chromosomal aberrations in a given tumour was prognostic for event-free survival. When different whole chromosomal aberration numbers were assessed, time-dependent AUC analysis identified 0 versus 1 or more recurrent whole chromosomal losses as the best discriminator of outcome (figure 4; appendix p 14). However, event-free survival was not only dependent on the total numbers of whole chromosomal aberrations. Analysis of the prognostic effect of specific whole chromosomal aberrations in individual chromosomes showed that chromosome 7 gain (HR 0·15, 95% CI 0·04–0·51, p=0·0025), chromosome 8 loss (HR calculation not possible because of group with no events; p=0·0014 for log-rank test), and chromosome 11 loss (HR 0·10, 95% CI 0·01–0·79, p=0·029) represented the most significant specific whole chromosomal aberrations (appendix pp 15–16).Figure 4 Whole chromosomal aberration-derived risk stratification schemes for non-WNT/non-SHH medulloblastomas

use of group with no events; p=0·0014 for log-rank test), and chromosome 11 loss (HR 0·10, 95% CI 0·01–0·79, p=0·029) represented the most significant specific whole chromosomal aberrations (appendix pp 15–16).Figure 4 Whole chromosomal aberration-derived risk stratification schemes for non-WNT/non-SHH medulloblastomas All 91 available samples from patients in the HIT-SIOP PNET 4 cohort with non-WNT/non-SHH standard-risk medulloblastoma are shown. Event-free survival per (A) whole chromosomal aberration cytogenetic subgroup and (B) recurrent whole chromosomal losses (0 vs 1 or more changes). (C) Proposed optimally performing risk stratification model, with the two identified risk groups. (D) Incidence and distribution of prognostically relevant chromosomal changes. For molecular subgroup, green indicates Group4 and yellow indicates Group3. For risk group, blue indicates high-risk and red indicates low-risk. Black and white indicate presence or absence of a feature, respectively. (E) Event-free survival by the scheme shown in part C. HR=hazard ratio. *HR estimates for favourable-risk vs high-risk were not possible due to the group with no events. p value reported from log-rank test.

es high-risk and red indicates low-risk. Black and white indicate presence or absence of a feature, respectively. (E) Event-free survival by the scheme shown in part C. HR=hazard ratio. *HR estimates for favourable-risk vs high-risk were not possible due to the group with no events. p value reported from log-rank test. Through assessment of event-free survival models for non-WNT/non-SHH medulloblastoma samples within the HIT-SIOP PNET 4 cohort, we identified at least two of the following—chromosome 7 gain, chromosome 8 loss, and chromosome 11 loss—as the optimally performing risk stratification scheme (appendix pp 14–15), outperforming other cytogenetic schemes and trial-based models, such as the SIOP PNET 5 MB trial model, in this patient group (figure 4; appendix p 14). This model, based on combinations of chromosome 7 gain, chromosome 8 loss, and chromosome 11 loss, stratified 38 (42%) of 91 non-WNT/non-SHH medulloblastomas as being favourable risk, with a 5-year event-free survival of 100%, (vs 68%, 95% CI 56·5–81·7 for high-risk tumours; p=0·00014 for log-rank test; figure 4). Further analysis of the high-risk patient group (53 [58%] of 91 patients), showed that 5-year event-free survival was equivalent between patients treated with hyperfractionated therapy or standard radiotherapy (HR 0·52, 95% CI 0·2–1·4, p=0·20 for Wald test; p=0·19 for log-rank test), consistent with findings from the overall HIT-SIOP PNET 4 trial cohort (data not shown).14

%] of 91 patients), showed that 5-year event-free survival was equivalent between patients treated with hyperfractionated therapy or standard radiotherapy (HR 0·52, 95% CI 0·2–1·4, p=0·20 for Wald test; p=0·19 for log-rank test), consistent with findings from the overall HIT-SIOP PNET 4 trial cohort (data not shown).14 We tested the reproducibility of our findings in an independent cohort of 70 non-WNT/non-SHH medulloblastomas, which matched the clinical and demographic characteristics of our HIT-SIOP PNET 4 standard-risk medulloblastoma cohort, collected from 1990 to 2014 (table, figure 5). The median event-free survival for these patients was 5·6 years (IQR 3·1–8·1).Figure 5 Validation of the whole chromosomal aberration-derived subgroups and risk stratification schemes

demographic characteristics of our HIT-SIOP PNET 4 standard-risk medulloblastoma cohort, collected from 1990 to 2014 (table, figure 5). The median event-free survival for these patients was 5·6 years (IQR 3·1–8·1).Figure 5 Validation of the whole chromosomal aberration-derived subgroups and risk stratification schemes All samples from the independent cohort of non-WNT/non-SHH-medulloblastoma (n=70) are shown in A and B. (A) Unsupervised clustering of chromosomal features by relevant chromosomal aberration cytogenetic subgroups. Residuals from χ2 tests indicate where whole chromosomal cytogenetic group enrichment has occurred. Darker shades of grey indicate stronger relationships. p values are from Fisher's exact tests. Total numbers of whole chromosomal losses (red), gains (blue), and changes (black) are shown. Increasing colour intensity indicates a higher number of changes. (B) Relationship of whole chromosomal aberration-defined risk groups to novel Group3 and Group4 disease subtypes. The standard-risk medulloblastoma validation cohort is indicated by filled and open circles according to risk, with relationship to the Schwalbe and colleagues6 and Northcott and colleagues12 cohorts shown by t-distributed stochastic neighbour embedding plots. (C) Event-free survival by whole chromosomal aberration-defined risk scheme. (D) Pooled analysis of event-free survival in the molecularly characterised HIT-SIOP PNET 4 cohort and validation cohort, stratified by derived whole chromosomal aberration-defined risk scheme. (E) Patterns of prognostically important cytogenetic changes in the combined cohort. The validation cohort is labelled black. Risk stratification is labelled red (favourable-risk) and blue (high-risk). HR=hazard ratio. TSNE=t-distributed stochastic neighbour embedding.

whole chromosomal aberration-defined risk scheme. (E) Patterns of prognostically important cytogenetic changes in the combined cohort. The validation cohort is labelled black. Risk stratification is labelled red (favourable-risk) and blue (high-risk). HR=hazard ratio. TSNE=t-distributed stochastic neighbour embedding. The characteristics, incidence, and associated event-free survival outcomes of the identified whole chromosomal aberration-defined subgroups were recapitulated (figure 5; appendix p 17). Our proposed whole chromosomal aberration signature-based model represented the best performing risk stratification scheme (figure 5; appendix p 18). The favourable-risk whole chromosomal aberration signature, defined by chromosome 7 gain, chromosome 8 loss, and chromosome 11 loss, was observed within multiple novel methylation subgroups, and was significantly associated with MBGroup4-LowRisk6 and Group3 and Group4 subtypes VI and VII12 (p<0·0001, appendix p 18). By contrast, the high-risk group was significantly associated with MBGroup4-HighRisk6 and subtype VIII12 (p<0·0001, figure 5; appendix p 18). When we considered our event-free survival models in Group4 patients alone, we found equivalent prognostic relationships in both the HIT-SIOP PNET 4 and validation cohorts (appendix pp 14, 19–20).

igh-risk group was significantly associated with MBGroup4-HighRisk6 and subtype VIII12 (p<0·0001, figure 5; appendix p 18). When we considered our event-free survival models in Group4 patients alone, we found equivalent prognostic relationships in both the HIT-SIOP PNET 4 and validation cohorts (appendix pp 14, 19–20). A pooled analysis applied the validated whole chromosomal aberration signature-based risk-stratification model to the merged non-WNT/non-SHH medulloblastomas from the HIT-SIOP PNET 4, and validation cohorts (n=161) and classified 58 (36%) non-WNT/non-SHH medulloblastomas as favourable-risk and 103 (64%) as high-risk; 5-year event-free survival was 98·3% (95% CI 94·9–100) in the favourable-risk group vs 64% (52·9–73·2) in the high-risk group (HR 25·09, 95% CI 3·44–183·20; p=0·0015; figure 5). Together with established favourable-risk WNT medulloblastomas in patients younger than 16 years (20 [15%] of 136 patients; appendix pp 5–6) and TP53wild-type SHH medulloblastomas (11 [8%] of 136 tumours; appendix pp 7–8), these newly identified chromosomal signatures identified 69 (51%) of 134 (two SHH tumours had unknown TP53 status and were therefore excluded from the calculation) molecularly characterised patients with medulloblastoma from the HIT-SIOP PNET 4 cohort with a favourable prognosis (5-year event-free survival of 100%).

), these newly identified chromosomal signatures identified 69 (51%) of 134 (two SHH tumours had unknown TP53 status and were therefore excluded from the calculation) molecularly characterised patients with medulloblastoma from the HIT-SIOP PNET 4 cohort with a favourable prognosis (5-year event-free survival of 100%). Discussion Implementation of enabling technologies (MS-MIMIC and molecular inversion probe assay) allowed us to systematically assess the molecular pathology of the standard-risk medulloblastoma clinical group within the HIT-SIOP PNET 4 patient cohort. To our knowledge, no equivalent multicentre, prospective investigations of standard-risk medulloblastoma have been reported. Although wider, retrospective medulloblastoma datasets are available, these typically lack the full clinical and molecular annotation necessary to define the standard-risk medulloblastoma group. The standard-risk medulloblastoma group displayed distinct demographics versus disease-wide cohorts.8, 9, 10, 11 WNT and Group4 subgroups were enriched within the standard-risk medulloblastoma cohorts because of the absence of clinicomolecular high-risk features in standard-risk disease.

ard-risk medulloblastoma group. The standard-risk medulloblastoma group displayed distinct demographics versus disease-wide cohorts.8, 9, 10, 11 WNT and Group4 subgroups were enriched within the standard-risk medulloblastoma cohorts because of the absence of clinicomolecular high-risk features in standard-risk disease. The favourable prognosis of patients with WNT medulloblastoma was confirmed in patients from the HIT-SIOP PNET 4 cohort who were younger than 16 years at diagnosis. However, patients older than 16 years did not share this good prognosis, consistent with previous reports.15, 23 Together, these data do not support therapy de-escalation in patients with WNT medulloblastoma older than 16 years of age. Patients with SHH medulloblastoma without TP53 mutations (SHH-TP53wild-type) or chromosome 17p loss similarly had a favourable prognosis. These data validate independent previous findings6, 7 and support the eligibility of these patients for de-escalated or targeted therapies (eg, SMO inhibitors).24

an 16 years of age. Patients with SHH medulloblastoma without TP53 mutations (SHH-TP53wild-type) or chromosome 17p loss similarly had a favourable prognosis. These data validate independent previous findings6, 7 and support the eligibility of these patients for de-escalated or targeted therapies (eg, SMO inhibitors).24 Development of biomarker-driven treatment strategies for the large remaining group of patients with non-WNT/non-SHH disease represents the largest ongoing challenge for standard-risk medulloblastoma. In the absence of high-risk features,5 these patients had a 5-year event-free survival of 81% (95% CI 74–90) in the HIT-SIOP PNET 4 trial. As described in this Article, non-WNT/non-SHH medulloblastoma tumours have few recurrent mutations, and structural chromosomal abnormalities are the most common genomic features.8, 9, 10 When comparing Group4 and Group3 tumours, we found around 90% overlap of chromosomal alterations between the two subgroups and equivalent event-free survival. Coupled with evidence supporting their shared underlying biological mechanisms,6, 12 we considered Group3 and Group4 tumours together in our analysis. We identified two biologically and clinically distinct non-WNT/non-SHH medulloblastoma groups. The first group was a cytogenetically quiet, high-risk group associated with diploid genomes, many with i17q as the sole defining genomic feature. These tumours provide a wider biological context for the poor-risk group of patients with non-WNT disease with chromosome 17p or q defects in a diploid background (chr17(im)/diploid(cen)), previously identified by interphase FISH in this cohort.15 The second group was large and defined by multiple, co-occurring whole chromosomal aberrations, common polyploidy, and improved relative outcomes.

f patients with non-WNT disease with chromosome 17p or q defects in a diploid background (chr17(im)/diploid(cen)), previously identified by interphase FISH in this cohort.15 The second group was large and defined by multiple, co-occurring whole chromosomal aberrations, common polyploidy, and improved relative outcomes. In this whole chromosomal aberration group, using multivariable event-free survival analysis and risk modelling, we deduced a whole chromosomal aberration signature (two or more of chromosome 7 gain, chromosome 8 loss, and chromosome 11 loss), which best defined patients with non-WNT/non-SHH medulloblastoma with favourable prognosis. We validated these findings in an independent demographically matched standard-risk medulloblastoma cohort, and they were reproducible when Group4 patients were considered in isolation. This whole chromosomal aberration signature was detected within a number of novel methylation subgroups within non-WNT/non-SHH medulloblastoma, and associated with the low-risk MBGroup4-LowRisk,6 and Group3 and Group4 subtypes VI and VII.12 By contrast, the high-risk isolated i17q diploid group was associated with high-risk MBGroup4-HighRisk6 and subtype VIII.12 These associations suggest common biological phenotypes and evaluation of their relative contributions to risk stratification could be investigated in future clinically controlled studies.

VII.12 By contrast, the high-risk isolated i17q diploid group was associated with high-risk MBGroup4-HighRisk6 and subtype VIII.12 These associations suggest common biological phenotypes and evaluation of their relative contributions to risk stratification could be investigated in future clinically controlled studies. Biologically and clinically significant whole chromosomal phenotypes are a notable feature of childhood malignancies other than medulloblastoma. Characteristic patterns of non-random whole chromosomal aberrations in neuroblastoma (so-called whole-chromosomal changes phenotype; more than two whole chromosomal aberrations)25, 26 and high hyperdiploid acute lymphoblastic leukaemia (so-called high-hyperdiploidy phenotype [HeH]; 51–65 chromosomes)27 define tumour subgroups with favourable prognoses. Additionally, choroid plexus papillomas and adult infratentorial ependymomas (posterior fossa ependymoma type B) are characterised by multiple whole chromosomal abberations.1 Overall, whole chromosomal aberration signatures are associated with a low number of single nucleotide mutations.

oups with favourable prognoses. Additionally, choroid plexus papillomas and adult infratentorial ependymomas (posterior fossa ependymoma type B) are characterised by multiple whole chromosomal abberations.1 Overall, whole chromosomal aberration signatures are associated with a low number of single nucleotide mutations. This common involvement of whole chromosomal aberration signatures provides strong impetus to understand the underlying molecular pathomechanisms, including errors in mitotic control, chromosome segregation, and function of the spindle apparatus. Although beyond the scope of this study, investigation of associated biology (eg, gene-expression profiles, pathway involvements, and driver events) and the involvement of specific chromosomes (ie, chromosomes 7, 8, and 11), is essential to improve understanding and therapeutic targeting. Potential opportunities include agents that target the spindle apparatus or mitotic control. For instance, vincristine (a component of medulloblastoma treatment regimens) directly targets the spindle apparatus, and the excellent whole chromosomal aberration signature-associated outcomes might be explained by high sensitivity to such treatments. Indeed, the association between HeH acute leukaemia and chemosensitivity associated with increased DNA content has been long established.28

ens) directly targets the spindle apparatus, and the excellent whole chromosomal aberration signature-associated outcomes might be explained by high sensitivity to such treatments. Indeed, the association between HeH acute leukaemia and chemosensitivity associated with increased DNA content has been long established.28 This study has some limitations. The developed risk stratification scheme applies only to non-infant, standard-risk medulloblastoma treated with standard multimodal therapies. Children younger than 4 years, patients treated with chemotherapy only, and high-risk patients require independent assessment and development of appropriate risk stratification schemes. However, our biomarker-driven risk stratification schemes for standard-risk medulloblastoma are readily testable in routine molecular diagnostic practice and, following their validation in independent clinically controlled and biomarker-defined cohorts, could form the basis of international clinical trials aimed at improving outcomes.

marker-driven risk stratification schemes for standard-risk medulloblastoma are readily testable in routine molecular diagnostic practice and, following their validation in independent clinically controlled and biomarker-defined cohorts, could form the basis of international clinical trials aimed at improving outcomes. In summary, our molecular pathological characterisation of the HIT-SIOP PNET 4 cohort identified and independently validated a whole chromosomal aberration signature-defined subgroup of non-WNT/non-SHH medulloblastomas associated with good prognosis. Combination of these newly defined subtypes with the favourable-risk WNT and SHH medulloblastomas validated in our study redistributed around 50% of patients with standard-risk medulloblastoma into a favourable-risk group, who could benefit from reduced-intensity therapies aimed at maintaining overall survival while reducing treatment-associated toxicities and late effects. Patients not classified into this favourable-risk group had a 5-year event-free survival of around 60% and should be considered high risk. In the HIT-SIOP PNET 4 cohort, this model compared favourably with published and currently accepted risk stratification schemes (eg, Shih and colleagues11 and SIOP PNET 5 MB;5 appendix p 14) and redefines our understanding of biomarkers and disease risk within the previously clinically defined standard-risk medulloblastoma patient group. Supplementary Material Supplementary appendix

In summary, our molecular pathological characterisation of the HIT-SIOP PNET 4 cohort identified and independently validated a whole chromosomal aberration signature-defined subgroup of non-WNT/non-SHH medulloblastomas associated with good prognosis. Combination of these newly defined subtypes with the favourable-risk WNT and SHH medulloblastomas validated in our study redistributed around 50% of patients with standard-risk medulloblastoma into a favourable-risk group, who could benefit from reduced-intensity therapies aimed at maintaining overall survival while reducing treatment-associated toxicities and late effects. Patients not classified into this favourable-risk group had a 5-year event-free survival of around 60% and should be considered high risk. In the HIT-SIOP PNET 4 cohort, this model compared favourably with published and currently accepted risk stratification schemes (eg, Shih and colleagues11 and SIOP PNET 5 MB;5 appendix p 14) and redefines our understanding of biomarkers and disease risk within the previously clinically defined standard-risk medulloblastoma patient group. Supplementary Material Supplementary appendix Acknowledgments This study was funded by Cancer Research UK (C8464/A13457 and C8464/A23391), the Swedish Childhood Cancer Foundation (PR2010-0077), the French Ministry of Health/French National Cancer Institute (PHRC 2003: AOM 03075/PHRC 2006: AOM 06161), and the German Children's Cancer Foundation (DKS2011.01; DKS2014.17). We thank Daniel Williamson, Rebecca Hill, Janet Lindsey, and Simon Bailey for their participation in editing of the manuscript.

(PR2010-0077), the French Ministry of Health/French National Cancer Institute (PHRC 2003: AOM 03075/PHRC 2006: AOM 06161), and the German Children's Cancer Foundation (DKS2011.01; DKS2014.17). We thank Daniel Williamson, Rebecca Hill, Janet Lindsey, and Simon Bailey for their participation in editing of the manuscript. Contributors TG, ECS, DH, TP, and SCC designed the study and wrote the manuscript. TG, ECS, DH, AS, and AzM did laboratory experiments and analysis. ECS did bioinformatics analysis. TG, ECS, and DH prepared the figures. TP, DF-B, FD, SR, and BL gathered samples and patient data and provided clinical interpretation. TP and DF-B provided central pathological review. All authors contributed to and approved the final manuscript. Declaration of interests FD received personal fees from Bristol-Myers Squibb, Tesaro Oncology, Servier, and Celgen. SR received grant funding from the German Children's Cancer Foundation (Deutsche Kinderkrebsstiftung). All other authors declare no competing interests.

Introduction Prostate cancer is the most common non-cutaneous malignancy among men living in developed countries.1, 2, 3 For patients with National Comprehensive Cancer Network (NCCN) low-risk or intermediate-risk disease,4 several management approaches can be considered, including external-beam radiotherapy (EBRT), brachytherapy, surgery, and—for some—active surveillance. Data from the randomised ProtecT trial, which compared surgery, EBRT, and active monitoring, have given reassurance that cancer outcomes are similar for low-risk and intermediate-risk disease, regardless of the management option used.5 Therefore, side-effects might influence decision making, with gastrointestinal, genitourinary, and sexual side-effects being common concerns.6 Additionally, the tolerability of treatment for a given patient is crucial, with anaesthetic and intra-operative risks balanced against the inconvenience of multiweek courses of EBRT. Research in context Evidence before this study

Introduction Prostate cancer is the most common non-cutaneous malignancy among men living in developed countries.1, 2, 3 For patients with National Comprehensive Cancer Network (NCCN) low-risk or intermediate-risk disease,4 several management approaches can be considered, including external-beam radiotherapy (EBRT), brachytherapy, surgery, and—for some—active surveillance. Data from the randomised ProtecT trial, which compared surgery, EBRT, and active monitoring, have given reassurance that cancer outcomes are similar for low-risk and intermediate-risk disease, regardless of the management option used.5 Therefore, side-effects might influence decision making, with gastrointestinal, genitourinary, and sexual side-effects being common concerns.6 Additionally, the tolerability of treatment for a given patient is crucial, with anaesthetic and intra-operative risks balanced against the inconvenience of multiweek courses of EBRT. Research in context Evidence before this study At the time of initiation of this study on Jan 25, 2012, to our knowledge there were no published randomised controlled trials of ultra-hypofractionated stereotactic body radiotherapy compared with conventional fractionated or moderately hypofractionated radiotherapy for localised prostate cancer. Standard treatment was radiotherapy in 2 Gy per fraction, to a dose of 74 Gy or 78 Gy. A subsequent change of standard-of-care practice to moderate hypofractionation over the course of 2016 was reflected in the control group of this study. We searched PubMed using the terms [“SBRT” OR “Stereotactic Body Radiotherapy”] AND “Prostate” for studies published in English up to March 31, 2019. We searched the reference lists of the papers identified by our search, and supplemented the search with the authors' knowledge of the field. We identified 16 studies reporting acute toxicity outcomes from ultrahypofractionated radiotherapy to the prostate, including a randomised phase 3 study (HYPO-RT-PC). Grade 2 or worse acute toxicity estimates for ultra-hypofractionation were similar to standard fractionation, ranging from 4–24% for gastrointestinal toxicity and 4–40% for genitourinary toxicity.

xicity outcomes from ultrahypofractionated radiotherapy to the prostate, including a randomised phase 3 study (HYPO-RT-PC). Grade 2 or worse acute toxicity estimates for ultra-hypofractionation were similar to standard fractionation, ranging from 4–24% for gastrointestinal toxicity and 4–40% for genitourinary toxicity. Added value of this study To our knowledge, this is the first published phase 3 randomised trial investigating acute toxicity after ultra-hypofractionated stereotactic body radiotherapy, delivered over five fractions, compared with standard fractionation schedules. Overall, this study showed similar acute toxicity for ultra-hypofractionation compared with standard fractionation, with only Common Terminology Criteria for Adverse Events grade 2 or more severe gastrointestinal toxicity being significantly worse. Proportions of patients with acute grade 3 toxicity were low, which adds to the body of evidence for low acute toxicity, as was also reported for seven-fraction hypofractionated radiotherapy in the HYPO-RT-PC trial. Implications of all the available evidence

To our knowledge, this is the first published phase 3 randomised trial investigating acute toxicity after ultra-hypofractionated stereotactic body radiotherapy, delivered over five fractions, compared with standard fractionation schedules. Overall, this study showed similar acute toxicity for ultra-hypofractionation compared with standard fractionation, with only Common Terminology Criteria for Adverse Events grade 2 or more severe gastrointestinal toxicity being significantly worse. Proportions of patients with acute grade 3 toxicity were low, which adds to the body of evidence for low acute toxicity, as was also reported for seven-fraction hypofractionated radiotherapy in the HYPO-RT-PC trial. Implications of all the available evidence Ultra-hypofractionated radiotherapy over five fractions appears to be tolerable in the short term in men with low-risk of intermediate-risk prostate adenocarcinoma. The HYPO-RT-PC trial showed that a schedule of 42·7 Gy delivered every other day over 2·5 weeks (6·1 Gy per fraction) was non-inferior in terms of failure-free survival compared with conventional fractionation of 78 Gy over 8 weeks (2 Gy per fraction), with similar proportions of late toxicity in each group. Late toxicity and efficacy data for the PACE-B trial are awaited and are required before a new standard of care for localised prostate cancer can be recommended.

ure-free survival compared with conventional fractionation of 78 Gy over 8 weeks (2 Gy per fraction), with similar proportions of late toxicity in each group. Late toxicity and efficacy data for the PACE-B trial are awaited and are required before a new standard of care for localised prostate cancer can be recommended. Hypofractionation—increasing the dose per fraction above the conventional 2 Gy, thus reducing the total fractions required—is an appealing approach. The key advantages are twofold. First, the greater fraction size sensitivity of prostate cancer (indicated by a lower α/β ratio7, 8, 9, 10), relative to the relevant late gastrointestinal and genitourinary side-effects, means that the therapeutic ratio might be improved by hypofractionation.11 Second, fewer fractions are needed with hypofractionation, allowing for quicker and more cost-effective EBRT treatment courses.12

ed by a lower α/β ratio7, 8, 9, 10), relative to the relevant late gastrointestinal and genitourinary side-effects, means that the therapeutic ratio might be improved by hypofractionation.11 Second, fewer fractions are needed with hypofractionation, allowing for quicker and more cost-effective EBRT treatment courses.12 Three major non-inferiority phase 3 randomised controlled trials have confirmed the safety and efficacy of moderate hypofractionation (2·5–3·0 Gy per fraction),11, 13, 14 which has gained acceptance as a standard-of-care option.15, 16 Although the proportions of patients with late toxicity were low, some intertrial differences in the proportion of patients who experienced acute toxicity were observed. The CHHiP trial reported significantly higher proportions of patients with peak acute Radiation Therapy Oncology Group (RTOG) grade 2 or worse gastrointestinal toxicity—38% in both hypofractionated groups—compared with conventional fractionation (25%; p<0·0001 for both comparisons).11 Similarly, the PROFIT trial reported a significantly (p=0·003) higher proportion of patients with cumulative acute RTOG grade 2 or worse gastrointestinal toxic effect proportions in the hypofractionated arm (16·7%) versus conventional fractionation (10·5%).14 For both trials, acute grade 2 or worse genitourinary toxic effects were similar between hypofractionated and conventionally fractionated groups. By contrast, the RTOG-0415 hypofractionation trial did not find a significant difference in acute gastrointestinal or genitourinary toxic effects between groups.13 Although more profound hypofractionation beyond 3·0 Gy per fraction would allow further reductions in the overall treatment time, the accelerated schedule might worsen acute toxicity, as seen in the CHHiP trial,11 potentially leading to late effects.17

e gastrointestinal or genitourinary toxic effects between groups.13 Although more profound hypofractionation beyond 3·0 Gy per fraction would allow further reductions in the overall treatment time, the accelerated schedule might worsen acute toxicity, as seen in the CHHiP trial,11 potentially leading to late effects.17 Substantial evidence exists for the efficacy of ultra-hypofractionation, with over 6000 patients treated in prospective studies and excellent 5-year biochemical progression-free survival in a recent meta-analysis (95·3%, 95% CI 91·3–97·5).18 A phase 3 trial (HYPO-RT-PC) reported good biochemical progression-free survival and acceptable proportions of toxic effects for seven-fraction ultra-hypofractionated radiotherapy.19 To our knowledge, phase 3 randomised toxic effect data for five-fraction treatment have not previously been reported. We report the acute toxicity findings (both clinician-reported and patient-reported) from the PACE-B randomised, controlled trial, which compared standard-of-care conventionally fractionated or moderately hypofractionated radiotherapy with five-fraction stereotactic body radiotherapy for low-risk to intermediate-risk localised prostate cancer. Methods Study design and participants PACE-B is an international, phase 3, open-label, randomised, non-inferiority trial at 37 centres (appendix p 7) in three countries (UK, Ireland, and Canada) aiming to assess non-inferiority of stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy for biochemical or clinical failure.

, phase 3, open-label, randomised, non-inferiority trial at 37 centres (appendix p 7) in three countries (UK, Ireland, and Canada) aiming to assess non-inferiority of stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy for biochemical or clinical failure. The PACE study comprises multiple cohorts (PACE-A, PACE-B, and PACE-C) which were independently randomised. This study, PACE-B, recruited only patients suitable for radical radiotherapy, but not willing to have or not suitable for radical prostatectomy. Eligible patients were men aged at least 18 years, with WHO performance status of 0–2,20 life expectancy of at least 5 years, and histologically confirmed prostate adenocarcinoma. All patients had NCCN low-risk or intermediate-risk disease.4 Low-risk patients had cT1c–T2a (TNM 6th edition21), N0-X, M0-X, Gleason score 6 or less, and prostate-specific antigen (PSA) concentration less than 10 ng/mL. Intermediate-risk patients had at least one of the following criteria: T2c, Gleason score 7 (3 + 4 for PACE; Gleason 4 + 3 was excluded), and PSA 10–20 ng/mL. Distant staging was not mandated. A minimum of ten biopsy cores taken within the last 18 months before randomisation were required, except for those progressing on active surveillance, whose last biopsy was suitable for PACE-B and required treatment (eg, PSA or MRI progression). These patients were stratified as intermediate risk. No PSA adjustment was made for 5-α reductase inhibitor use at randomisation. Treating physicians had discretion to exclude patients for comorbid conditions that made radiotherapy inadvisable (eg, inflammatory bowel disease or substantial urinary tract symptoms). Detailed inclusion and exclusion criteria are in the protocol (appendix pp 82–83).

e for 5-α reductase inhibitor use at randomisation. Treating physicians had discretion to exclude patients for comorbid conditions that made radiotherapy inadvisable (eg, inflammatory bowel disease or substantial urinary tract symptoms). Detailed inclusion and exclusion criteria are in the protocol (appendix pp 82–83). The trial was approved by the London Chelsea research ethics committee (reference 11/LO/1915) in the UK and the relevant institutional review boards in Ireland and Canada. PACE-B was conducted in accordance with the principles of Good Clinical Practice. All participants provided voluntary written informed consent. Randomisation and masking Patients were randomly assigned (1:1) to conventionally fractionated or moderately hypofractionated radiotherapy or stereotactic body radiotherapy. Randomisation was done centrally by the Institute of Cancer Research Clinical Trials and Statistics Unit (ICR-CTSU), by telephone (UK and Ireland) or fax (Canada), with allocation by computer generated random permuted blocks (size four and six) and stratification by centre and risk group (low or intermediate). Sequence generation, enrolment, and trial group assignment were done by ICR-CTSU staff who were not involved in the clinical running of the trial or data collection. Participants and researchers were not masked to treatment assignment.

ize four and six) and stratification by centre and risk group (low or intermediate). Sequence generation, enrolment, and trial group assignment were done by ICR-CTSU staff who were not involved in the clinical running of the trial or data collection. Participants and researchers were not masked to treatment assignment. Procedures Before radiotherapy, three or more prostatic fiducial markers were strongly recommended for all participants to permit more accurate image-guided radiotherapy and CT or MRI fusion. Bowel preparation (enema) was suggested, along with moderate bladder filling. The radiotherapy planning CT scan took place at least 7 days after fiducial placement. A radiotherapy planning MRI scan was strongly recommended to be fused to the CT scan (preferably by fiducial match) for improved prostate anatomical definition. The clinical target volume (CTV) was the prostate only (low-risk patients) or prostate and proximal 1 cm of seminal vesicles (intermediate-risk patients). The recommended conventionally fractionated or moderately hypofractionated radiotherapy CTV to planning target volume (PTV) expansion was 5–9 mm isometric, except posteriorly 3–7 mm. The recommended stereotactic body radiotherapy CTV to PTV expansion was 4–5 mm isometric, except posteriorly 3–5 mm. Dose constraints were applied to organs at risk and were amended during the trial. A history of the constraints used with numbers of patients randomised to each iteration is presented in the appendix (pp 4–5). Additional detail on radiotherapy preparation and final dose constraints used from March 24, 2016, are in the protocol (appendix pp 93–100). Androgen deprivation therapy was not permitted.

ial. A history of the constraints used with numbers of patients randomised to each iteration is presented in the appendix (pp 4–5). Additional detail on radiotherapy preparation and final dose constraints used from March 24, 2016, are in the protocol (appendix pp 93–100). Androgen deprivation therapy was not permitted. The conventionally fractionated or moderately hypofractionated radiotherapy PTV dose was 78 Gy in 39 daily fractions or, following an approved protocol amendment (on March 24, 2016), 62 Gy in 20 daily fractions. This change followed the CHHiP trial data supporting moderate hypofractionation,11 but with a higher dose (62 Gy vs 60 Gy) because the PACE-B protocol prohibits androgen deprivation therapy. Data from the PROFIT trial which assessed 60 Gy in 20 fractions, without androgen deprivation therapy, were not available at that time.14 After the protocol amendment, centres were required to choose either 78 Gy in 39 fractions or 62 Gy in 20 fractions as their control treatment for all subsequent patients. The stereotactic body radiotherapy PTV dose was 36·25 Gy in five fractions over 1–2 weeks (ie, daily or alternate days, at centre discretion), with an additional secondary CTV dose target of 40 Gy. The CyberKnife treatment platform (Accuray; Sunnyvale, CA, USA) was initially mandatory for all stereotactic body radiotherapy; however, sponsorship changes prompted a protocol amendment (on Oct 24, 2014) permitting stereotactic body radiotherapy delivery on conventional linear accelerators. Detailed prescription objectives, along with minor variations permitted, are listed in the protocol (appendix pp 93–100).

l stereotactic body radiotherapy; however, sponsorship changes prompted a protocol amendment (on Oct 24, 2014) permitting stereotactic body radiotherapy delivery on conventional linear accelerators. Detailed prescription objectives, along with minor variations permitted, are listed in the protocol (appendix pp 93–100). Treatment was mandated to commence within 12 weeks of randomisation, with 8 weeks or less strongly recommended. Image-guided radiotherapy (preferably fiducial based) was mandated. For stereotactic body radiotherapy, intrafractional motion monitoring was permitted; otherwise, a repeat static image was required for stereotactic body radiotherapy fraction delivery extending beyond 3 min. A radiotherapy quality assurance programme was undertaken for each centre to ensure consistency with the trial protocol and quality of radiotherapy treatments (appendix p 113).

oring was permitted; otherwise, a repeat static image was required for stereotactic body radiotherapy fraction delivery extending beyond 3 min. A radiotherapy quality assurance programme was undertaken for each centre to ensure consistency with the trial protocol and quality of radiotherapy treatments (appendix p 113). Participants were assessed on alternate weeks during conventionally fractionated or moderately hypofractionated radiotherapy and on the final fraction for stereotactic body radiotherapy, and at weeks 2, 4, 8, and 12 after the end of treatment for all patients in both groups. Two clinician-reported outcomes were collected—RTOG (gastrointestinal and genitourinary domains) at baseline and every visit and Common Terminology Criteria for Adverse Events (CTCAE version 4.03) at baseline and follow-up weeks 2, 4, 8, and 12, with additional end-of-treatment assessment for patients in the stereotactic body radiotherapy group. Specific CTCAE items in the gastrointestinal composite are anal pain, colitis, constipation, diarrhoea, diverticulitis, faecal incontinence, fistula, gastrointestinal pain, haemorrhoids, gastrointestinal haemorrhage, proctitis, rectal pain, gastrointestinal unspecified, and rectal prolapse. Specific CTCAE items in the genitourinary composite are bladder spasm, cystitis, haematuria, prostatic obstruction, urinary frequency, urinary incontinence, urinary retention, urinary urgency, and urethral stricture. We used paper questionnaires to collect four patient-reported outcome measures—expanded prostate cancer index composite short form (EPIC-26) and the Vaizey faecal incontinence score, at baseline and weeks 4 and 12; international prostate symptom score (IPSS), at baseline and weeks 2, 4, 8, and 12; and the international index of erectile function 5-question (IIEF-5) score, at baseline and week 12. Subsequent follow-up is ongoing (and will continue until all patients have reached 10 years), with the full schedule, along with criteria for removal of patients from the study, available in the protocol (appendix pp 84–92). Regular toxicity and patient-reported outcome assessment occurs during follow up. Study recruitment is complete.

-up is ongoing (and will continue until all patients have reached 10 years), with the full schedule, along with criteria for removal of patients from the study, available in the protocol (appendix pp 84–92). Regular toxicity and patient-reported outcome assessment occurs during follow up. Study recruitment is complete. The EPIC tool gives a measure of patient-reported quality of life in genitourinary, gastrointestinal, sexual, and general domains. The Vaizey questionnaire measures patient quality of life relating to faecal incontinence and the IPSS records patient experience of various facets of urinary function. For each scale, the baseline, worst, worst above baseline, and week 12 (residual) toxic effects were of interest, with exact definitions detailed in the statistical analysis plan (appendix pp 145–47). Outcomes The primary endpoint of PACE-B is freedom from biochemical or clinical failure, the data for which is not yet mature. This acute toxicity report is a prespecified subanalysis of the PACE-B trial. A statistical analysis plan for this substudy (appendix pp 129–158) was prospectively written, with worst grade 2 or worse RTOG toxic effects score, up to week 12 follow-up after radiotherapy finished, for both gastrointestinal and genitourinary systems, as coprimary sub-study endpoints.

ubanalysis of the PACE-B trial. A statistical analysis plan for this substudy (appendix pp 129–158) was prospectively written, with worst grade 2 or worse RTOG toxic effects score, up to week 12 follow-up after radiotherapy finished, for both gastrointestinal and genitourinary systems, as coprimary sub-study endpoints. Separately for gastrointestinal and genitourinary systems, the numerator was patients with recorded RTOG grade 2 or worse toxic effects at any point after baseline up to week 12 after radiotherapy. The denominator was all patients with at least one RTOG score completed after baseline up to week 12 after radiotherapy. Patients were recorded as missing if no such score was returned. This endpoint was pragmatically chosen, as only RTOG assessments were done for patients in the conventionally fractionated or moderately hypofractionated group during radiotherapy. PACE-B secondary endpoints were acute toxicity (CTCAE), late toxicity (CTCAE and RTOG), progression-free survival, disease-specific survival, overall survival, distant progression, commencement of hormone therapy, and acute and late patient-reported toxicity (EPIC, IPSS, IIEF-5, and Vaizey scales). All but the physician-reported and patient-reported toxicity outcomes will be reported elsewhere.

and RTOG), progression-free survival, disease-specific survival, overall survival, distant progression, commencement of hormone therapy, and acute and late patient-reported toxicity (EPIC, IPSS, IIEF-5, and Vaizey scales). All but the physician-reported and patient-reported toxicity outcomes will be reported elsewhere. Statistical analysis For this acute toxicity analysis, patients were analysed per protocol, with those receiving one or more fractions of conventionally fractionated or moderately hypofractionated radiotherapy or stereotactic body radiotherapy included. Patients who did not receive radiotherapy were excluded from this analysis. Patients receiving both conventionally fractionated or moderately hypofractionated radiotherapy and stereotactic body radiotherapy fractions were excluded unless the reason was toxicity-related, where analysis was on the first treatment fraction received. The PACE-B trial targeted recruitment of 858 patients to exclude a hazard ratio [HR] of 1·45 in biochemical or clinical failure at 5 years, with consideration given to also excluding a 6% increase in grade 2 gastrointestinal or genitourinary late toxicity at 2 years (appendix pp 107–08). For this acute toxicity substudy, we assumed conventionally fractionated or moderately hypofractionated radiotherapy group acute RTOG grade 2 or worse toxic effect proportions of 25% (gastrointestinal) and 40% (genitourinary), as per the CHHiP trial.11 With two-sided α=0·025 for each endpoint, we estimated that ongoing PACE-B recruitment would provide 83% power to exclude a 10% increase in acute gastrointestinal toxic effects and 84·5% power to exclude an 11% increase in acute genitourinary toxic effects for the stereotactic body radiotherapy group (appendix pp 139–40).

h two-sided α=0·025 for each endpoint, we estimated that ongoing PACE-B recruitment would provide 83% power to exclude a 10% increase in acute gastrointestinal toxic effects and 84·5% power to exclude an 11% increase in acute genitourinary toxic effects for the stereotactic body radiotherapy group (appendix pp 139–40). We used the χ2 test to compare treatment groups for the coprimary endpoints. Secondary endpoints were compared using appropriate statistical tests (appendix pp 145–47). To reduce the effect of multiple comparisons, p<0·001 was considered significant for secondary comparisons. The different durations of radiotherapy (conventionally fractionated or moderately hypofractionated radiotherapy 4 or 7·8 weeks; stereotactic body radiotherapy 1 or 2 weeks) led to differing timepoints of toxicity assessment. RTOG (assessed during radiotherapy) and CTCAE (assessed at end of radiotherapy for stereotactic body radiotherapy) graphical presentation is therefore protrayed as four different groups and also displayed grouped as 1–2 weeks and 4–7·5 weeks (interpolation detailed in appendix p 6). We calculated confidence intervals for the difference in proportions by normal approximation.

t end of radiotherapy for stereotactic body radiotherapy) graphical presentation is therefore protrayed as four different groups and also displayed grouped as 1–2 weeks and 4–7·5 weeks (interpolation detailed in appendix p 6). We calculated confidence intervals for the difference in proportions by normal approximation. We rescaled EPIC-26 scores to a 0–100 point scale, with higher scores representing better quality of life.22 Subdomains were scored if sufficient questions were completed as follows: urinary incontinence (four of four questions), urinary obstructive (four of four questions), bowel (five of six questions), sexual (five of six questions), and hormonal (four of five questions).22 A clinically important point reduction in EPIC-26 subdomain score was as follows: urinary incontinence (8 points), urinary obstruction (6 points), bowel (5 points), sexual (11 points), and hormonal (5 points).23 IPSS severity categories were assessed as none (0 points), mild (1–7 points), moderate (8–19 points), and severe (20–35 points).24

point reduction in EPIC-26 subdomain score was as follows: urinary incontinence (8 points), urinary obstruction (6 points), bowel (5 points), sexual (11 points), and hormonal (5 points).23 IPSS severity categories were assessed as none (0 points), mild (1–7 points), moderate (8–19 points), and severe (20–35 points).24 Exploratory examination of CyberKnife versus standard linear accelerators for patients undergoing stereotactic body radiotherapy was prespecified in the protocol when amendment permitted standard linear accelerators (Aug 5, 2014). The prespecified statistical analysis plan called for a multivariate analysis, which will be published subsequently, but a post-hoc decision due to the paucity of published non-CyberKnife toxicity data was made to analyse the worst RTOG grade 2 or more severe gastrointestinal and genitourinary toxic effects for patients undergoing stereotactic body radiotherapy, split by CyberKnife and non-CyberKnife use, interpreted at a significant p-value of 0·001. Since centre-level effects could influence this non-randomised analysis (eg, variation in toxic effect reporting), we did similar analysis for patients undergoing conventionally fractionated or moderately hypofractionated radiotherapy, separated by whether their centre used CyberKnife or non-CyberKnife for stereotactic body radiotherapy treatments.

ld influence this non-randomised analysis (eg, variation in toxic effect reporting), we did similar analysis for patients undergoing conventionally fractionated or moderately hypofractionated radiotherapy, separated by whether their centre used CyberKnife or non-CyberKnife for stereotactic body radiotherapy treatments. The study was overseen by a trial steering committee and an independent data monitoring committee (IDMC; appendix p 3). Analyses are based on a snapshot of data taken on May 28, 2019, and were done with STATA version 15.1. The IDMC gave approval for release of acute toxicity results before release of primary endpoint (efficacy) results. The PACE study is prospectively registered at ClinicalTrials.gov, NCT01584258. Role of the funding source The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

The study was overseen by a trial steering committee and an independent data monitoring committee (IDMC; appendix p 3). Analyses are based on a snapshot of data taken on May 28, 2019, and were done with STATA version 15.1. The IDMC gave approval for release of acute toxicity results before release of primary endpoint (efficacy) results. The PACE study is prospectively registered at ClinicalTrials.gov, NCT01584258. Role of the funding source The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Between Aug 7, 2012, and Jan 4, 2018, we randomly assigned 874 men to conventionally fractionated or moderately hypofractionated radiotherapy (n=441) or stereotactic body radiotherapy (n=433; figure 1). Median follow-up was 12 weeks (IQR 12–12), matching the time period authorised for data release. 11 patients received non-protocol regimens due to crossing over treatment groups (figure 1; appendix p 8); one cross over was toxicity-related (a patient in the stereotactic body radiotherapy group with grade 3 urinary toxicity), meaning that 432 patients received at least one fraction of conventionally fractionated or moderately hypofractionated radiotherapy and 416 patients received at least one fraction of stereotactic body radiotherapy. Two men received both stereotactic body radiotherapy and conventionally fractionated or moderately hypofractionated radiotherapy treatments—one patient who received two fractions of stereotactic body radiotherapy (14·5 Gy) then developed grade 3 toxicity (urosepsis) and switched to conventionally fractionated or moderately hypofractionated radiotherapy (further 46 Gy in 23 fractions) was not excluded from the toxicity analysis because he had toxic effects after two fractions of sterotactic body radiotherapy and one patient who received a single incomplete fraction of stereotactic body radiotherapy (<7·25 Gy, set-up issues) and switched to conventionally fractionated or moderately hypofractionated radiotherapy (further 55 Gy in 20 fractions) was excluded (figure 1).Figure 1 Trial profile

ns of sterotactic body radiotherapy and one patient who received a single incomplete fraction of stereotactic body radiotherapy (<7·25 Gy, set-up issues) and switched to conventionally fractionated or moderately hypofractionated radiotherapy (further 55 Gy in 20 fractions) was excluded (figure 1).Figure 1 Trial profile Crossovers between treatment groups were analysed per-protocol for this acute toxicity substudy. Dose fractionation regimens administered within each group are shown. *One patient who received two fractions of stereotactic body radiotherapy then developed grade 3 toxicity (urosepsis) and switched to conventionally fractionated or moderately hypofractionated radiotherapy (further 46 Gy in 23 fractions) was not excluded from the toxicity analysis because he had toxic effects after two fractions of sterotactic body radiotherapy. †One patient who received a single incomplete fraction of stereotactic body radiotherapy (<7·25 Gy, set-up issues) and switched to conventionally fractionated or moderately hypofractionated radiotherapy (further 55 Gy in 20 fractions) was excluded.

had toxic effects after two fractions of sterotactic body radiotherapy. †One patient who received a single incomplete fraction of stereotactic body radiotherapy (<7·25 Gy, set-up issues) and switched to conventionally fractionated or moderately hypofractionated radiotherapy (further 55 Gy in 20 fractions) was excluded. Baseline characteristics for each per-protocol treatment group were similar (table 1). Four (21%) of 19 patients on a 5-α reductase inhibitor at baseline had a PSA value of 10–20 ng/mL. Radiotherapy delivery techniques (planning, image-guided radiotherapy, and margins) differed between arms, as expected, although recorded supportive prescribing appeared to be similar (appendix pp 9–11). Despite fiducial recommendations for both groups, more patients in the stereotactic body radiotherapy group received fiducial markers (303 [73%] of 415 patients) than did patients in the conventionally fractionated or moderately hypofractionated radiotherapy group (245 [57%] of 432 patients). RTOG and CTCAE form completion was excellent at all timepoints (appendix p 12). Patient illness caused non-completion of three RTOG forms (two in the conventionally fractionated or moderately hypofractionated radiotherapy group and one in the stereotactic body radiotherapy group) and one CTCAE form in the stereotactic body radiotherapy group. Patient-reported outcome assessment completion varied by scale (appendix pp 13–14). One patient randomly assigned to stereotactic body radiotherapy died because of myocardial infarction before receiving trial treatment and was excluded from per-protocol analyses; no other deaths were reported up to 12 weeks after completion of radiotherapy.Table 1 Baseline characteristics

by scale (appendix pp 13–14). One patient randomly assigned to stereotactic body radiotherapy died because of myocardial infarction before receiving trial treatment and was excluded from per-protocol analyses; no other deaths were reported up to 12 weeks after completion of radiotherapy.Table 1 Baseline characteristics Conventionally fractionated or moderately hypofractionated radiotherapy group (n=432) Stereotactic body radiotherapy group (n=415) Age (years) 69·7 (65·6–73·9) 69·6 (65·3–73·8) Ethnicity Black 25 (6%) 35 (8%) East Asian 3 (1%) 4 (1%) Mixed heritage 2 (<1%) 2 (<1%) South Asian 9 (2%) 19 (5%) White 386 (89%) 352 (85%) Other 7 (2%) 3 (1%) Family history of prostate cancer No 321 (74%) 300 (72%) Yes 85 (20%) 85 (20%) Unknown 26 (6%) 30 (7%) WHO performance status 0 382 (88%) 372 (90%) 1 48 (11%) 43 (10%) 2 2 (<1%) 0 National Comprehensive Cancer Network risk score Low 38 (9%) 30 (7%) Intermediate 394 (91%) 385 (93%) T stage T1c 78 (18%) 76 (18%) T2a 130 (30%) 105 (25%) T2b 57 (13%) 81 (20%) T2c 167 (39%) 153 (37%) Gleason grade 3 + 3 84 (19%) 61 (15%) 3 + 4 348 (81%) 354 (85%) Pre-treatment PSA (ng/mL) Mean 8·7 (3·7) 8·6 (4·0) Median 8·0 (6·3–11·0) 8·0 (5·5–11·0) <10 299 (69%) 283 (68%) 10–20 133 (31%) 132 (32%) Pre-treatment testosterone (nmol/L) <1·7 0 2 (<1%) ≥1·7 391 (91%) 376 (91%) Unknown 41 (9%) 37 (9%) Active surveillance before trial enrolment Yes 160 (37%) 146 (35%) No 258 (60%) 256 (62%) Unknown 14 (3%) 13 (3%) Prostate volume (mL) <40 153 (35%) 160 (39%) 40–<80 200 (46%) 170 (41%) ≥80 16 (4%) 21 (5%) Unknown 63 (15%) 64 (15%) α blockers at randomisation Yes 68 (16%) 67 (16%) No 361 (84%) 344 (83%) Unknown 3 (1%) 4 (1%) Aspirin at randomisation Yes 74 (17%) 63 (15%) No 355 (82%) 347 (84%) Unknown 3 (1%) 5 (1%) Statin at randomisation Yes 153 (35%) 126 (30%) No 275 (64%) 283 (68%) Unknown 4 (1%) 6 (1%) Anticholinergic for bladder symptoms at randomisation Yes 16 (4%) 10 (2%) No 414 (96%) 400 (96%) Unknown 2 (<1%) 5 (1%) 5-α reductase inhibitor at randomisation Yes 9 (2%) 10 (2%) No 416 (96%) 387 (93%) Unknown 7 (2%) 18 (4%) Phosphodiesterase-5 inhibitor at randomisation Yes 12 (3%) 6 (1%) No 412 (95%) 392 (94%) Unknown 8 (2%) 17 (4%) Data are median (IQR), n (%), or mean (SD). PSA=prostate-specific antigen.

96%) Unknown 2 (<1%) 5 (1%) 5-α reductase inhibitor at randomisation Yes 9 (2%) 10 (2%) No 416 (96%) 387 (93%) Unknown 7 (2%) 18 (4%) Phosphodiesterase-5 inhibitor at randomisation Yes 12 (3%) 6 (1%) No 412 (95%) 392 (94%) Unknown 8 (2%) 17 (4%) Data are median (IQR), n (%), or mean (SD). PSA=prostate-specific antigen. Worst RTOG grade 2 or more severe gastrointestinal toxic effects did not differ significantly between conventionally fractionated or moderately hypofractionated radiotherapy (53 [12%] of 432 patients) and stereotactic body radiotherapy (43 [10%] of 415 patients; difference −1·9 percentage points, 95% CI −6·2 to 2·4; p=0·38; table 2).Table 2 Radiation Therapy Oncology Group adverse events Conventionally fractionated or moderately hypofractionated radiotherapy (n=432) Stereotactic body radiotherapy (n=415) Grade 1 Grade 2 Grade 3 Grade 4 Grade 1 Grade 2 Grade 3 Grade 4 Gastrointestinal 264 (61%) 49 (11%) 4 (1%) 0 219 (53%) 42 (10%) 1 (<1%) 0 Genitourinary 254 (59%) 111 (26%) 6 (1%) 1 (<1%) 236 (57%) 86 (21%) 8 (2%) 2 (<1%) Data are n (%). No death due to adverse events were reported. Worst RTOG grade 2 or more severe genitourinary toxic effects also did not differ significantly between conventionally fractionated or moderately hypofractionated radiotherapy (118 [27%] of 432 patients) and stereotactic body radiotherapy (96 [23%] of 415 patients; difference −4·2 percentage points, 95% CI −10·0 to 1·7; p=0·16).

grade 2 or more severe genitourinary toxic effects also did not differ significantly between conventionally fractionated or moderately hypofractionated radiotherapy (118 [27%] of 432 patients) and stereotactic body radiotherapy (96 [23%] of 415 patients; difference −4·2 percentage points, 95% CI −10·0 to 1·7; p=0·16). For RTOG secondary endpoints, we observed no significant differences between conventionally fractionated or moderately hypofractionated radiotherapy and stereotactic body radiotherapy for any comparison for gastrointestinal toxic effects (appendix p 15), including worst RTOG gastrointestinal toxic effects of grade 3 or more severe (four [1%] of 432 patients in the conventionally fractionated or moderately hypofractionated radiotherapy group vs one [<1%] of 415 patients in the stereotactic body radiotherapy group; difference −0·7 percentage points, 95% CI −1·7 to 0·3; p=0·37), nor for genitourinary toxic effects (appendix p 16), including worst RTOG genitourinary toxic effects of grade 3 or more severe (seven [2%] of 432 patients in the conventionally fractionated or moderately hypofractionated radiotherapy group vs ten [2%] of 415 patients in the stereotactic body radiotherapy group; difference 0·8 percentage points, −1·1 to 2·7; p=0·47).

6), including worst RTOG genitourinary toxic effects of grade 3 or more severe (seven [2%] of 432 patients in the conventionally fractionated or moderately hypofractionated radiotherapy group vs ten [2%] of 415 patients in the stereotactic body radiotherapy group; difference 0·8 percentage points, −1·1 to 2·7; p=0·47). We recorded RTOG acute toxicity over time for gastrointestinal and genitourinary toxic effects and observed a similar time course of toxicity peak and recovery between the groups (figure 2). Graphical representation of the four different durations of treatment separately (stereotactic body radiotherapy 1 week and 2 weeks and conventionally fractionated or moderately hypofractionated radiotherapy 4 weeks and 7·8 weeks) is shown in the appendix (p 17). The RTOG baseline, worst, worst (exceeding baseline), and week-12 after radiotherapy gastrointestinal and genitourinary toxic effects are summarised in the appendix (pp 15–16).Figure 2 Acute Radiation Therapy Oncology Group toxicity for gastrointestinal (A) and genitourinary (B) systems

is shown in the appendix (p 17). The RTOG baseline, worst, worst (exceeding baseline), and week-12 after radiotherapy gastrointestinal and genitourinary toxic effects are summarised in the appendix (pp 15–16).Figure 2 Acute Radiation Therapy Oncology Group toxicity for gastrointestinal (A) and genitourinary (B) systems As each group allowed two different treatment durations (CFMHRT 78 Gy in 39 fractions and 62 Gy in 20 fractions; SBRT 36·25 Gy in five fractions over 1 or 2 weeks) it was necessary to interpolate data where assessments did not overlap. Raw data are presented in the appendix (p 17), with all four schedules shown separately. Numbers at risk for each arm are asynchronous because they are shown only at data collection timepoints (which are non-simultaneous relative to the start of radiotherapy). Week 0 is the baseline toxicity score taken before start of radiotherapy. CFMHRT=conventionally fractionated or moderately hypofractionated radiotherapy. SBRT=stereotactic body radiotherapy.

ous because they are shown only at data collection timepoints (which are non-simultaneous relative to the start of radiotherapy). Week 0 is the baseline toxicity score taken before start of radiotherapy. CFMHRT=conventionally fractionated or moderately hypofractionated radiotherapy. SBRT=stereotactic body radiotherapy. A summary table of all common and serious CTCAE adverse events is provided in the appendix (pp 18–19). 17 serious adverse events were reported (five in the conventionally fractionated or moderately hypofractionated radiotherapy group and 12 in the stereotactic body radiotherapy group) up to 12 weeks after radiotherapy, of which 15 (five in the conventionally fractionated or moderately hypofractionated radiotherapy and ten in the stereotactic body radiotherapy group) were related to treatment (appendix p 20). We recorded CTCAE acute toxicity over time for composite gastrointestinal and genitourinary toxic effects, and observed a similar time course of toxicity peak and recovery between stereotactic body radiotherapy and conventionally fractionated or moderately hypofractionated radiotherapy (figure 3). Graphical representation of the four different durations of treatment separately (stereotactic body radiotherapy 1 week and 2 weeks and conventionally fractionated or moderately hypofractionated radiotherapy 4 weeks and 7·8 weeks) is shown in the appendix (p 21). Data for composite gastrointestinal and genitourinary toxic effects, at baseline, worst, worst (exceeding baseline), and week 12 after radiotherapy are summarised in the appendix (pp 22–23), with the results of hypothesis testing. Stereotactic body radiotherapy was statistically significantly worse compared with the conventionally fractionated or moderately hypofractionated radiotherapy for two of the CTCAE secondary endpoints analysed—worst CTCAE grade 2 or more severe gastrointestinal toxic effects (36 [8%] of 430 patients vs 65 [16%] of 415 patients; difference 7·3 percentage points, 95% CI 2·9–11·7; p=0·0011), corroborated by worst CTCAE grade 2 or more severe gastrointestinal toxic effects exceeding baseline (34 [8%] of 427 patients vs 63 [15%] of 413 patients; difference 7·3 percentage points, 95% CI 3·0–11·6; p=0·00095; appendix p 22). Diarrhoea grade 2 and worst proctitis grade 2 occurred more frequently in the stereotactic body radiotherapy group. We found no significant difference in worst CTCAE grade 2 or more severe gastrointestinal toxic effects by week 12.

patients; difference 7·3 percentage points, 95% CI 3·0–11·6; p=0·00095; appendix p 22). Diarrhoea grade 2 and worst proctitis grade 2 occurred more frequently in the stereotactic body radiotherapy group. We found no significant difference in worst CTCAE grade 2 or more severe gastrointestinal toxic effects by week 12. We observed no other significant differences in CTCAE gastrointestinal secondary endpoints for conventionally fractionated or moderately hypofractionated radiotherapy compared with stereotactic body radiotherapy (appendix p 22), including worst CTCAE gastrointestinal grade 3 or more severe toxic effects (three [1%] of 430 patients in the conventionally fractionated or moderately hypofractionated radiotherapy vs three [1%] of 415 patients in the stereotactic body radiotherapy group). We observed no significant differences in CTCAE genitourinary secondary endpoints between the conventionally fractionated or moderately hypofractionated radiotherapy and stereotactic body radiotherapy groups (appendix p 23), including worst CTCAE genitourinary grade 3 or more severe toxic effects (three [1%] of 430 patients vs seven [2%] of 415 patients). Further tables broken down into individual CTCAE toxicity items, separately for gastrointestinal and genitourinary systems, are presented in the appendix (pp 24–39) and show baseline CTCAE toxicity, worst acute CTCAE toxicity, worst (exceeding baseline) acute CTCAE toxicity, and week 12 CTCAE toxicity.Figure 3 Acute CTCAE toxicity for gastrointestinal (A) and genitourinary systems

, separately for gastrointestinal and genitourinary systems, are presented in the appendix (pp 24–39) and show baseline CTCAE toxicity, worst acute CTCAE toxicity, worst (exceeding baseline) acute CTCAE toxicity, and week 12 CTCAE toxicity.Figure 3 Acute CTCAE toxicity for gastrointestinal (A) and genitourinary systems As each group allowed two different treatment durations (CFMHRT 78 Gy in 39 fractions and 62 Gy in 20 fractions; SBRT 36·25 Gy in five fractions over 1 or 2 weeks) it was necessary to interpolate data. Raw data are presented in the appendix (p 21), with all four schedules presented separately. Numbers at risk for each arm are asynchronous because they are shown only at data collection timepoints (which are non-simultaneous relative to the start of radiotherapy). The initial points for CFMHRT are connected by grey dashed lines to emphasise that there were no CTCAE assessments during radiotherapy delivery. Week 0 is the baseline toxicity score taken before start of radiotherapy. CTCAE=Common Terminology Criteria for Adverse Events. CFMHRT=conventionally fractionated or moderately hypofractionated radiotherapy. SBRT=stereotactic body radiotherapy.

mphasise that there were no CTCAE assessments during radiotherapy delivery. Week 0 is the baseline toxicity score taken before start of radiotherapy. CTCAE=Common Terminology Criteria for Adverse Events. CFMHRT=conventionally fractionated or moderately hypofractionated radiotherapy. SBRT=stereotactic body radiotherapy. EPIC-26 mean changes in subdomain scores over time were similar, both for change from baseline (figure 4) and absolute scores (appendix p 40). Comparison over each of the five EPIC-26 subdomains and overall urinary bother for scores at baseline, worst, worst minus baseline, and week 12 after radiotherapy showed no significant differences between the trial groups (appendix p 41). We observed no significant difference between the study groups in the proportion of patients with a clinically significant reduction from baseline for any EPIC-26 subdomain score area, neither assessed at any time (appendix p 42) nor at week-12 only (appendix p 43).Figure 4 Changes from baseline in expanded prostate cancer index composite (26 question) subdomains

the study groups in the proportion of patients with a clinically significant reduction from baseline for any EPIC-26 subdomain score area, neither assessed at any time (appendix p 42) nor at week-12 only (appendix p 43).Figure 4 Changes from baseline in expanded prostate cancer index composite (26 question) subdomains Urinary bother is graphed separately, as it does not form part of the urinary incontinence or obstructive subdomain scores. Error bars show 95% CIs for estimates of mean subdomain scores. The time period between baseline scoring and week 4 after radiotherapy follow-up is variable, since the total time of radiotherapy delivery varied (SBRT in 1 or 2 weeks; CFMHRT in 4 or 7·8 weeks). Week 0 is the baseline score taken before start of radiotherapy. Scores are change from baseline, with 0 representing no change. CFMHRT=conventionally fractionated or moderately hypofractionated radiotherapy. SBRT=stereotactic body radiotherapy. IPSS subscores, total score, and quality of life over time were similar between the study groups, both for change from baseline (appendix p 44) and absolute scores (appendix p 45). We observed no significant differences between treatment groups for median scores of worst IPSS total, week-12 IPSS total, worst IPSS quality of life, or week-12 IPSS quality of life (appendix p 46). IPSS severity categories (none, mild, moderate, or severe) over time were similar between the treatment groups (appendix p 47), with no significant differences in IPSS total score categories at baseline, worst, and week-12 after radiotherapy (appendix p 48).

ity of life, or week-12 IPSS quality of life (appendix p 46). IPSS severity categories (none, mild, moderate, or severe) over time were similar between the treatment groups (appendix p 47), with no significant differences in IPSS total score categories at baseline, worst, and week-12 after radiotherapy (appendix p 48). For IIEF-5, we observed no significant differences between conventionally fractionated or moderately hypofractionated radiotherapy and stereotactic body radiotherapy at baseline or at week 12 after radiotherapy (appendix p 49). Vaizey score changes were similar between treatment groups for both change from baseline and absolute scores (appendix pp 50–51). We observed no significant differences between treatment groups for Vaizey scores at baseline, worst, worst change from baseline, and week 12 after radiotherapy (appendix p 52).

endix p 49). Vaizey score changes were similar between treatment groups for both change from baseline and absolute scores (appendix pp 50–51). We observed no significant differences between treatment groups for Vaizey scores at baseline, worst, worst change from baseline, and week 12 after radiotherapy (appendix p 52). For the stereotactic body radiotherapy group, worst RTOG gastrointestinal grade 2 or more severe (without reference to baseline) toxic effects for non-CyberKnife (27 [11%] of 245 patients) versus CyberKnife (16 [9%] of 170 patients) delivery was not different (difference −1·6 percentage points, 95% CI −7·5 to 4·3; p=0·597), consistent with observations over time (appendix p 53). For patients in the stereotactic body radiotherapy group, worst RTOG grade 2 or more severe genitourinary (without reference to baseline) toxic effects for non-CyberKnife (75 [31%] of 245 patients) versus CyberKnife (21 [12%] of 170 patients) delivery were significantly different (difference −18·3 percentage points, 95% CI −10·7 to −25·9; p<0·0001), consistent with observations over time (appendix p 54). Given the non-randomised nature of comparison of non-CyberKnife with CyberKnife delivery, we examined the conventionally fractionated or moderately hypofractionated radiotherapy toxicity in non-CyberKnife centres compared with CyberKnife centres. For patients in the conventionally fractionated or moderately hypofractionated radiotherapy group, worst RTOG gastrointestinal grade 2 or more severe (without reference to baseline) toxic effects in non-CyberKnife-using centres (25 [10%] of 252 patients) versus CyberKnife centres (28 [16%] of 180 patients) were not different (difference 5·6 percentage points, 95% CI −0·8 to 12·1; p=0·078), consistent with grade 2 and grade 3 observations over time (appendix p 55). For patients in the conventionally fractionated or moderately hypofractionated radiotherapy group, worst RTOG grade 2 or more severe genitourinary (without reference to baseline) toxic effects in non-CyberKnife-using centres (73 [29%] of 252 patients) versus CyberKnife-using centres (45 [25%] of 180 patients) were not significantly different (difference −4·0 percentage points, 95% CI −12·4 to 4·5; p=0·361), contrary to possible graphical interpretation over time (appendix p 56).

ce to baseline) toxic effects in non-CyberKnife-using centres (73 [29%] of 252 patients) versus CyberKnife-using centres (45 [25%] of 180 patients) were not significantly different (difference −4·0 percentage points, 95% CI −12·4 to 4·5; p=0·361), contrary to possible graphical interpretation over time (appendix p 56). Discussion This pre-planned analysis of acute toxicity in the PACE-B trial, occurring up to 12 weeks after radiotherapy delivery completion, does not suggest that patients have greater acute RTOG toxic effects with stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy. Of the secondary endpoints examined, only worst CTCAE grade 2 or more severe composite toxic effects (both with and without reference to baseline) showed significantly higher proportions of patients with toxic effects when treated with stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy. Differences in CTCAE toxicity were resolved by week 12 after completion of radiotherapy. Patient-reported outcomes were similar between the trial groups. Overall, our results do not provide consistent evidence of higher acute toxicity with stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy.

solved by week 12 after completion of radiotherapy. Patient-reported outcomes were similar between the trial groups. Overall, our results do not provide consistent evidence of higher acute toxicity with stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy. It is notable that the control group in our trial (conventionally fractionated or moderately hypofractionated radiotherapy) had lower acute toxicity than in the preceding CHHiP trial,11 with toxicity more comparable to the PROFIT trial (appendix p 57).14 Although image-guided radiotherapy was mandatory in both the PACE and PROFIT14 trials, it was only used in 30% of CHHiP participants, which could have caused this difference. PACE also used smaller margins and benefitted from use of highly conformal techniques, such as volumetric modulated arc therapy. The CHHiP trial used androgen deprivation therapy for most patients, which was not permitted in PACE or PROFIT; however, androgen deprivation therapy is not known to alter acute toxicity. Both PROFIT and CHHiP assessed acute RTOG weekly during radiotherapy versus two-weekly assessment in PACE. Conceivably, the cumulative proportion of higher worst RTOG grade 2 or more severe events in CHHiP and PROFIT versus PACE-B might result from recall selection bias due to more frequent sampling in PROFIT and CHHiP.

oth PROFIT and CHHiP assessed acute RTOG weekly during radiotherapy versus two-weekly assessment in PACE. Conceivably, the cumulative proportion of higher worst RTOG grade 2 or more severe events in CHHiP and PROFIT versus PACE-B might result from recall selection bias due to more frequent sampling in PROFIT and CHHiP. The most similar phase 3 randomised controlled trial to PACE-B is the Scandinavian HYPO-RT-PC trial, which randomly assigned (1:1) intermediate-risk and high-risk patients with prostate cancer to 78 Gy in 39 fractions over 7·8 weeks or 42·7 Gy in seven fractions over 2·5 weeks, without androgen deprivation therapy.19 Important differences between PACE-B and HYPO-RT-PC are as follows: HYPO-RT-PC recruited 11% high-risk patients and 89% intermediate-risk patients (vs 8% low-risk patients and 92% intermediate-risk patients in PACE-B), treated a CTV of prostate only, and mostly (80%) used three-dimensional (3D) conformal radiotherapy. Image-guided radiotherapy (fiducial markers or guidance catheter) and planning MRI were used for all patients in HYPO-RT-PC. The control groups differ between HYPO-RT-PC (all 78 Gy in 39 fractions) and PACE-B (70% receiving 62 Gy in 20 fractions). This difference is important given the higher acute gastrointestinal toxicity observed for moderate hypofractionation in the CHHiP trial.11 HYPO-RT-PC made only a single end-of-treatment toxicity assessment during the acute toxicity window, and reported significantly higher RTOG genitourinary and patient-reported outcome acute toxic effects with ultra-hypofractionation. Comparison of RTOG toxicity for PACE-B with HYPO-RT-PC (estimates approximated from graphs in paper19) produces similar results, although reported grade 3 to grade 4 toxicity for HYPO-RT-PC is higher than most reports of ultra-hypofractionation (appendix p 58). Although measured on different patient-reported outcome scales to HYPO-RT-PC, our results do not suggest a difference in patient-reported outcome acute side-effects.

uces similar results, although reported grade 3 to grade 4 toxicity for HYPO-RT-PC is higher than most reports of ultra-hypofractionation (appendix p 58). Although measured on different patient-reported outcome scales to HYPO-RT-PC, our results do not suggest a difference in patient-reported outcome acute side-effects. We identified no up to date systematic literature review of acute toxicity in this setting. Therefore, we prospectively collated acute toxicity data from smaller studies of stereotactic body radiotherapy in low-risk and intermediate-risk prostate cancer (appendix p 58). The PACE-B outcomes appear to be broadly in line with results anticipated from earlier phase work. For example, a multicentre phase 2 study of 309 men25 recorded cumulative acute toxicity of CTCAE gastrointestinal grade 2 or worse of 12% and CTCAE genitourinary grade 2 or worse of 26%, similar to the 15·7% and 30·8%, respectively, for patients in the stereotactic body radiotherapy group in PACE-B.

arlier phase work. For example, a multicentre phase 2 study of 309 men25 recorded cumulative acute toxicity of CTCAE gastrointestinal grade 2 or worse of 12% and CTCAE genitourinary grade 2 or worse of 26%, similar to the 15·7% and 30·8%, respectively, for patients in the stereotactic body radiotherapy group in PACE-B. Strengths of these data relate predominantly to trial design. This is a large phase 3 randomised, controlled trial, and represents, to our knowledge, the first published phase 3 acute toxicity data on five-fraction stereotactic body radiotherapy compared with standard fractionation. PACE-B reflects real world prostate radiotherapy practice, with multiple centres recruiting in the UK, Canada, and Ireland. This study incorporates modern planning practice, with no patients receiving 3D conformal radiotherapy. The protocol amendment relating to treatment in the control group strengthened the trial by allowing most patients in that group to receive moderate hypofractionation at 62 Gy in 20 fractions, close to the 60 Gy in 20 fractions regimen shown to be effective in CHHiP11 and PROFIT.14 The PACE-B acute toxicity sampling frequency exceeded HYPO-RT-PC (assessed only at end of radiotherapy and 6 months). Combined with the high proportions of assessment forms returned, this is a major strength given the dynamic nature of acute toxicity.

20 fractions regimen shown to be effective in CHHiP11 and PROFIT.14 The PACE-B acute toxicity sampling frequency exceeded HYPO-RT-PC (assessed only at end of radiotherapy and 6 months). Combined with the high proportions of assessment forms returned, this is a major strength given the dynamic nature of acute toxicity. Limitations arise from the external applicability of the patients recruited to PACE-B. These results cannot necessarily be extrapolated to higher-risk patients, nor alternative treatment techniques. Randomised data regarding toxicity after stereotactic body radiotherapy, with concurrent androgen deprivation therapy and a larger target volume, will be acquired by the PACE-C trial. This trial cohort will randomly assign unfavourable intermediate-risk and lower high-risk patients to either stereotactic body radiotherapy or moderately hypofractionated radiotherapy. The absence of treatment blinding is always a limitation for subjective endpoints, such as toxicity. Although blinding has been achieved in previous radiotherapy trials,26, 27 it is not feasible for most studies. We also note the higher fiducial marker use for image-guided radiotherapy in patients undergoing stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy in PACE-B. Mandatory fiducials would have prevented some centres participating, slowing trial recruitment. Furthermore, the multiple radiotherapy schedule durations meant that some undesirable interpolation was needed to present two arm graphs (RTOG and CTCAE). This fact also means that the follow up of 12 weeks after radiotherapy refers to quite different period of time for someone receiving 1 week of stereotactic body radiotherapy (ie, 13 weeks from the start of radiotherapy) versus 7·8 weeks of conventional fractionation (ie, 19·8 weeks after commencing treatment). Future trials should consider a follow-up schedule fixed by radiotherapy start date rather than end date.

eriod of time for someone receiving 1 week of stereotactic body radiotherapy (ie, 13 weeks from the start of radiotherapy) versus 7·8 weeks of conventional fractionation (ie, 19·8 weeks after commencing treatment). Future trials should consider a follow-up schedule fixed by radiotherapy start date rather than end date. Stereotactic body radiotherapy is already the standard of care in some centres and is an option for men with low and favourable intermediate-risk prostate cancer in the NCCN guidelines.28 The HYPO-RT-PC trial suggested similar oncological outcomes with ultra-hypofractionation.19 This result was attenuated by increased acute toxicity in the study, notably higher grade 3 or worse toxic effects than other reports of stereotactic body radiotherapy, which might potentially be driven by the 3D conformal radiotherapy technique predominantly used in the HYPO-RT-PC study. Other earlier phase studies, most of which used the same 36·25 Gy dose as PACE (appendix p 58), suggest good oncological outcomes and low late toxicity with stereotactic body radiotherapy, but the mature results of PACE-B are required before definite oncological outcome statements can be made.

in the HYPO-RT-PC study. Other earlier phase studies, most of which used the same 36·25 Gy dose as PACE (appendix p 58), suggest good oncological outcomes and low late toxicity with stereotactic body radiotherapy, but the mature results of PACE-B are required before definite oncological outcome statements can be made. The method of stereotactic body radiotherapy delivery—for example, CyberKnife versus non-CyberKnife—might influence acute toxicity, a prespecified area of interest after the introduction of conventional linear accelerator stereotactic body radiotherapy. There are many reasons why there might be a systematic difference between CyberKnife and non-CyberKnife stereotactic body radiotherapy outcomes, including variations in dosimetry, image guidance, and treatment times (typically 45 min for CyberKnife and <5 min for conventional linear accelerators). Our post-hoc analysis of the same primary endpoint RTOG metrics shows similar grade 2 or worse gastrointestinal toxic effects, but less grade 2 or worse genitourinary toxic effects with CyberKnife. We compared conventionally fractionated or moderately hypofractionated radiotherapy toxic effects between centres using CyberKnife versus those not using CyberKnife and found no significant difference for either worst RTOG grade 2 or more severe gastrointestinal or genitourinary toxic effects. We caution that this result is hypothesis-generating and intend to explore further in multivariate analyses once digital imaging and communications in medicine data have been centralised for all patients.

ignificant difference for either worst RTOG grade 2 or more severe gastrointestinal or genitourinary toxic effects. We caution that this result is hypothesis-generating and intend to explore further in multivariate analyses once digital imaging and communications in medicine data have been centralised for all patients. To our knowledge, we present the first published prospective phase 3 acute toxicity results for random assignment of patients between five-fraction stereotactic body radiotherapy and either conventional or moderately hypofractionated radiotherapy. Our results do not suggest that patients have greater acute RTOG toxic effects with stereotactic body radiotherapy compared with conventionally fractionated or moderately hypofractionated radiotherapy. The absence of increased toxicity in the stereotactic body radiotherapy group is reassuring given the higher acute toxicity suggested in the only previously published phase 3 ultra-hypofractionation trial,19 especially given the more abbreviated (five-fraction) investigational radiotherapy protocol used in PACE-B. Results regarding late toxicity and biochemical control from PACE-B will be reported in the next 3-4 years.

igher acute toxicity suggested in the only previously published phase 3 ultra-hypofractionation trial,19 especially given the more abbreviated (five-fraction) investigational radiotherapy protocol used in PACE-B. Results regarding late toxicity and biochemical control from PACE-B will be reported in the next 3-4 years. Data sharing The Institute of Cancer Research Clinical Trials and Statistics Unit (ICR-CTSU) supports the wider dissemination of information from the research it conducts and increased cooperation between investigators. Trial data is collected, managed, stored, shared, and archived according to ICR-CTSU Standard Operating Procedures to ensure the enduring quality, integrity, and utility of the data. Formal requests for data sharing are considered in line with ICR-CTSU procedures with due regard given to funder and sponsor guidelines. Requests are via a standard proforma describing the nature of the proposed research and extent of data requirements. Data recipients are required to enter a formal data sharing agreement, which describes the conditions for release and requirements for data transfer, storage, archiving, publication, and intellectual property. Requests are reviewed by the trial management group (TMG) in terms of scientific merit and ethical considerations including patient consent. Data sharing is undertaken if proposed projects have a sound scientific or patient benefit rationale as agreed by the TMG and approved by the independent data monitoring and steering committee as required. Restrictions relating to patient confidentiality and consent will be limited by aggregating and anonymising identifiable patient data. Additionally, all indirect identifiers that might lead to deductive disclosures will be removed in line with Cancer Research UK Data Sharing Guidelines.

steering committee as required. Restrictions relating to patient confidentiality and consent will be limited by aggregating and anonymising identifiable patient data. Additionally, all indirect identifiers that might lead to deductive disclosures will be removed in line with Cancer Research UK Data Sharing Guidelines. Supplementary Material Supplementary appendix

steering committee as required. Restrictions relating to patient confidentiality and consent will be limited by aggregating and anonymising identifiable patient data. Additionally, all indirect identifiers that might lead to deductive disclosures will be removed in line with Cancer Research UK Data Sharing Guidelines. Supplementary Material Supplementary appendix Acknowledgments The PACE-B study is funded by Accuray. This study is endorsed by Cancer Research UK and the Clinical Trials and Statistics Unit at the Institute of Cancer Research (ICR-CTSU) coordinated the trial. The trial funder, Accuray, was also the sponsor of the trial until February, 2014, when sponsorship was transferred to The Royal Marsden NHS Foundation Trust. Accuray had no role in data collection, which was managed by a third party before February, 2014. All data analysis was done by ICR-CTSU. The sponsor (The Royal Marsden NHS Foundation Trust) received funding from Accuray for study management, international study coordination, and analysis. Excess service costs were met by the UK's Comprehensive Local Research Networks. Trial recruitment was facilitated within centres by the National Institute for Health Research (NIHR) Cancer Research Network. Funding for delegated tasks outside the UK was as follows: in Ireland the study was coordinated by the Irish Clinical Oncology Research Group CLG trading as Cancer Trials Ireland and in Canada the study was supported by the Prostate Cure Foundation. The ICR-CTSU receives programme grant funding from Cancer Research UK (C1491/A15955), which supported in part this endorsed study (CRUKE/12/025). This paper represents independent research part funded by the NIHR Biomedical Research Centre at the Royal Marsden NHS Foundation Trust and the Institute of Cancer Research. The views expressed are those of the authors and not necessarily those of the UK National Health Service, the NIHR, or the UK Department of Health. We would like to thank David Dearnaley for his support and advice over the course of this trial. We thank our patients, the investigators, and the research support staff at all participating centres. We also thank the independent data monitoring committee and trial steering committee.

R, or the UK Department of Health. We would like to thank David Dearnaley for his support and advice over the course of this trial. We thank our patients, the investigators, and the research support staff at all participating centres. We also thank the independent data monitoring committee and trial steering committee. Contributors NvA is the chief investigator. DH, CG, EH, and NvA designed the trial. DH, CC, SBu, CG, EH, and NvA developed the protocol. ACT, PO, HvdV, AL, WC, DF, ST, SJ, AM, JS, PC, KK, JF, AC, ISD, DH, AD, and NvA recruited participants. ACT, PO, HvdV, AL, WC, DF, ST, SJ, AM, JS, PC, KK, JF, AC, ISD, DH, SBr, CC, SBu, AD, CG, KM, and NvA collected the data. DHB, ACT, PO, AL, WC, DF, ST, SJ, AM, JS, SBr, CC, SBu, AD, CG, VH, KM, ON, EH, and NvA are members of the PACE Trial Management Group. DHB, CG, VH, and EH did the statistical analyses. DHB, CG, VH, EH, and NvA interpreted the data. ON leads the Physics Quality Assurance Group. SBr, CC, and SBu managed the trial. DHB, ACT, CG, VH, EH, and NvA wrote the manuscript. All authors provided critical academic review of content for the manuscript and gave final approval for submission of the work to the journal.

, CG, VH, EH, and NvA interpreted the data. ON leads the Physics Quality Assurance Group. SBr, CC, and SBu managed the trial. DHB, ACT, CG, VH, EH, and NvA wrote the manuscript. All authors provided critical academic review of content for the manuscript and gave final approval for submission of the work to the journal. Declaration of interests DHB reports a PhD studentship award from Cancer Research UK, during the conduct of the study. ACT reports grants from Accuray, during the conduct of the study, grants and personal fees from Elekta, grants from Merck Sharpe & Dohme, and personal fees from Janssen, Astellas, and Ferring, outside the submitted work. AL reports grants from Prostate Cure Foundation to his institution during the conduct of the study and has a patent issued for a prostate immobilisation device. DF reports personal fees from Janssen, Sanofi, Novartis, and Ipsen, and speaker fees from Bayer, outside the submitted work. SJ reports grants and personal fees from Augmenix, and personal fees from Astellas, Bayer, Janssen, and Movember, outside the submitted work. AM reports a travel grant for conference attendance from Bayer, outside the submitted work. JS reports non-financial support from Bayer and personal fees from Janssen and Astellas, outside the submitted work. DH reports grants from Accuray, during the conduct of the study and personal fees from Accuray, outside the submitted work. SBr reports grants from Accuray, during the conduct of the study. CG reports grants from Accuray during the conduct of the study. VH reports grants from Accuray, during the conduct of the study. KM reports grants from Accuray during the conduct of the study. EH reports grants from Accuray and Cancer Research UK, during the conduct of the study, grants and non-financial support from Merck Sharp & Dohme, Astra Zeneca, and Bayer, and grants from Janssen-Cilag, Kyowa Hakko UK, Alliance Pharma (previously Cambridge Laboratories), and Aventis Pharma Limited (Sanofi), outside the submitted work. NvA reports grants and personal fees from Accuray, during the conduct of the study. All other authors declare no competing interests.

de Boer SM, Powell ME, Mileshkin L, et al. Adjuvant chemoradiotherapy versus radiotherapy alone in women with high-risk endometrial cancer (PORTEC-3): patterns of recurrence and post-hoc survival analysis of a randomised phase 3 trial. Lancet Oncol 2019; 20: 1273–85—In this Article, data (hazard ratios, 95% CIs, and p values) in the following sentence on p 1279 have been corrected: “In women with stage I–II disease, 5-year overall survival was 83·8% (95% CI 78·4–89·5) with chemoradiotherapy versus 82·0% (95% CI 76·5–87·7) with radiotherapy alone (HR 0·84 [95% CI 0·52–1·38]; p=0·50), and 5-year failure-free survival was 81·3% (95% CI 74·7–86·3) with chemoradiotherapy versus 77·3% (95% CI 70·5–82·7) with radiotherapy alone (HR 0·87 [95% CI 0·56–1·36] p=0·54; appendix p 8).” On p 1282, the same data (hazard ratios and 95% CIs) in the following sentence have also been corrected: “For women with stage I–II endometrial cancer, combined adjuvant treatment yielded only a small absolute improvement of 2% (HR 0·84; 95% CI 0·52–1·38) in 5-year overall survival and of 4% (0·87; 0·56–1·36) in failure-free survival.” These corrections have been made to the online version as of Sept 2, 2019, and the printed version is correct.

Introduction Breast cancer is the most common cancer diagnosis worldwide, and the oestrogen receptor is expressed in most tumours. Endocrine therapies targeting the oestrogen receptor are an integral component of treatment for oestrogen receptor-positive breast cancer, but resistance develops in almost all patients with advanced disease. Several resistance mechanisms have been identified, including alteration of the PI3K/AKT pathway. This pathway is altered in more than 50% of oestrogen receptor-positive advanced breast cancers, most frequently through somatic hotspot mutation in exons 9 and 20 of PIK3CA, encoding the p110α isoform of PI3K.1, 2, 3, 4 Less frequently, pathway alteration is induced by loss of function mutation or deletion of the negative regulator PTEN or activating mutations in AKT1. PI3K pathway alteration is associated with endocrine therapy resistance through ligand independent activation of the oestrogen receptor.5, 6 Conversely, preclinical data show compensatory increases in ligand-dependent oestrogen receptor transcription following PI3K pathway inhibition.7, 8, 9 A rationale therefore exists for simultaneously targeting both the oestrogen receptor and PI3K pathways. Research in context Evidence before this study

Introduction Breast cancer is the most common cancer diagnosis worldwide, and the oestrogen receptor is expressed in most tumours. Endocrine therapies targeting the oestrogen receptor are an integral component of treatment for oestrogen receptor-positive breast cancer, but resistance develops in almost all patients with advanced disease. Several resistance mechanisms have been identified, including alteration of the PI3K/AKT pathway. This pathway is altered in more than 50% of oestrogen receptor-positive advanced breast cancers, most frequently through somatic hotspot mutation in exons 9 and 20 of PIK3CA, encoding the p110α isoform of PI3K.1, 2, 3, 4 Less frequently, pathway alteration is induced by loss of function mutation or deletion of the negative regulator PTEN or activating mutations in AKT1. PI3K pathway alteration is associated with endocrine therapy resistance through ligand independent activation of the oestrogen receptor.5, 6 Conversely, preclinical data show compensatory increases in ligand-dependent oestrogen receptor transcription following PI3K pathway inhibition.7, 8, 9 A rationale therefore exists for simultaneously targeting both the oestrogen receptor and PI3K pathways. Research in context Evidence before this study We searched PubMed between Jan 1, 2009, and July 31, 2019, to identify publications directly relevant to the FAKTION clinical setting using the search terms “AKT” or “PI3K” or “mTOR” and “oestrogen receptor” and “breast cancer” and “metastatic” and “inhibitor” or “inhibition”. We also searched PubMed for publications in the same period using the terms “capivasertib” or “AZD5363”. We did not use any language restrictions in our search. We found no reports of randomised trials investigating the inhibition of AKT in combination with endocrine therapies in oestrogen receptor-positive, HER2-negative breast cancer. The only other randomised phase 2 study of AKT inhibition in patients with oestrogen receptor-positive, HER2-negative breast cancer showed no advantage of addition of capivasertib to paclitaxel chemotherapy. Five randomised, placebo-controlled trials tested the addition of PI3K or mTOR inhibitors to endocrine therapies. These studies have shown pan-PI3K and beta-sparing inhibitors to have an unfavourable toxicity profile and low clinical activity, and they are no longer in development for this indication. The alpha-specific PI3K inhibitor, alpelisib has activity in combination with fulvestrant, but only in PIK3CA-mutant tumours and toxicity remains problematic. mTOR inhibition with everolimus has shown activity, again at the cost of substantial toxicity, but the effect is agnostic to perturbation of the PI3K pathway.

n. The alpha-specific PI3K inhibitor, alpelisib has activity in combination with fulvestrant, but only in PIK3CA-mutant tumours and toxicity remains problematic. mTOR inhibition with everolimus has shown activity, again at the cost of substantial toxicity, but the effect is agnostic to perturbation of the PI3K pathway. Added value of this study To our knowledge, this study is the first randomised trial to report on the addition of an AKT inhibitor to endocrine therapy in oestrogen receptor-positive metastatic breast cancer after previous aromatase inhibitor therapy. The results showed an improvement in progression-free survival and response rate with addition of capivasertib to endocrine therapy, suggesting synergy, in contrast to the poor efficacy in combination with chemotherapy. Adverse events were common, but manageable with dose reduction, and did not seem to compromise efficacy. Implications of all the available evidence Several approaches to targeting the PI3K/AKT/mTOR pathway have been shown to be effective in metastatic oestrogen receptor-positive, HER2-negative breast cancer in combination with endocrine therapy. AKT and mTOR inhibition seems to be active in a broader population of patients than PI3K inhibitors. The intermittent scheduling of capivasertib in this study contrasts with the continuous treatment with PI3K and mTOR inhibitors in the majority of previous publications, potentially resulting in improved tolerability. The FAKTION data support further investigation of AKT inhibition with capivasertib in combination with fulvestrant in a phase 3 trial.

apivasertib in this study contrasts with the continuous treatment with PI3K and mTOR inhibitors in the majority of previous publications, potentially resulting in improved tolerability. The FAKTION data support further investigation of AKT inhibition with capivasertib in combination with fulvestrant in a phase 3 trial. Several clinical trials have reported improved progression-free survival with inhibitors of the PI3K pathway in combination with endocrine therapies. Distal inhibition with the mTORC1 inhibitor everolimus improved progression-free survival in combination with the aromatase inhibitor exemestane irrespective of the alteration status of the PI3K pathway, albeit at the cost of additional toxicity.3, 10 By contrast, proximal pathway inhibition with the PI3Kα-subunit specific inhibitor alpelisib significantly enhanced the efficacy of fulvestrant, but only in tumours harbouring PIK3CA hotspot mutations.11 This finding has led to selected approval for alpelisib, in combination with fulvestrant, in this specific subgroup of patients with oestrogen receptor-positive advanced breast cancer. An unmet need therefore remains for patients whose tumours do not carry PIK3CA hotspot mutations.