Browse the corpus

Introduction In Sub-Saharan Africa, babies in the first month of life have the highest risk of death, and the region has made little progress in reducing this high mortality rate.1 Severe anaemia is a major public health problem in sub-Saharan Africa, and children younger than 2 years are the most frequently affected. The prevalence of severe anaemia in children in hospital is 8–29%, with case fatality ranging from 8% to 18%.2 In children with severe uncompensated anaemia, blood transfusion can reduce mortality substantially.3 More than 50% of deaths happen within 4 h of admission, and early intervention and a source of safe blood are key components of the treatment of severe anaemia in childhood.4, 5 Supply of conventional blood for transfusion in sub-Saharan Africa is insufficient, with only an estimated 52% of demand being met and a shortfall of at least 2 million units a year.6, 7, 8 In situations when blood supply is limited, or young children need a substantial number of transfusions, the umbilical cord is a novel and potentially important source of blood.9, 10, 11 Use of cord blood might not only enable an increase in the number of small-volume transfusions but also reduce pressure on stocks of conventional adult-donated blood, thereby augmenting supplies for emergency transfusions for other vulnerable groups. In sub-Saharan Africa, a scarcity of blood for transfusion is implicated in 25% of maternal deaths due to haemorrhage.12

in the number of small-volume transfusions but also reduce pressure on stocks of conventional adult-donated blood, thereby augmenting supplies for emergency transfusions for other vulnerable groups. In sub-Saharan Africa, a scarcity of blood for transfusion is implicated in 25% of maternal deaths due to haemorrhage.12 To test the feasibility of cord blood transfusion, we have established a cord blood donation programme on the labour ward at Coast Provincial General Hospital in Mombasa, Kenya. Previously, we have shown the acceptability to mothers of cord blood donation and transfusion, the feasibility of a two-stage informed consent process for cord blood donation, and the quality of variable volumes of whole cord blood stored in a fixed volume of anticoagulant preservative solution.13, 14 We also reported that, for cord blood obtained by our study team, the frequency of both bacterial contamination and seroreactivity for HIV, hepatitis B and C viruses, and syphilis compare favourably with those for conventional adult blood donated to the regional blood transfusion centre in Mombasa.15 To our knowledge, we report here the first clinical trial of allogeneic cord blood transfusion in children with severe anaemia. The aim of our study was to assess the safety and efficacy of umbilical cord red blood cell transfusion in children with severe anaemia.

to the regional blood transfusion centre in Mombasa.15 To our knowledge, we report here the first clinical trial of allogeneic cord blood transfusion in children with severe anaemia. The aim of our study was to assess the safety and efficacy of umbilical cord red blood cell transfusion in children with severe anaemia. Methods Participants We designed an open-label single-arm study with the aim to produce preliminary data for safety, harm, and haematological efficacy of umbilical cord red blood cell transfusion in children with severe anaemia. We recruited children younger than 12 years who were admitted for paediatric care at Kilifi District Hospital, Kenya. We designed our eligibility criteria to identify children for whom a transfusion would provide clinical benefit, based on WHO clinical guidelines, but exclude those who were critically ill.16 Children were eligible for inclusion in the study if they had severe anaemia (haemoglobin ≤100 g/L in babies aged 3 months or younger, or ≤40 g/L in children older than 3 months) and the attending clinician requested a blood transfusion. We excluded children with any of these clinical features of critical illness: coma (Blantyre coma scale ≤2); prostration; shock; deep (acidotic) breathing; and hyperbilirubinaemia requiring exchange transfusion. Furthermore, we judged children ineligible for the study if they had received a previous cord blood transfusion as part of this trial or were already enrolled in another intervention trial. We only enrolled a child into the study if sufficient cord blood was available.

yperbilirubinaemia requiring exchange transfusion. Furthermore, we judged children ineligible for the study if they had received a previous cord blood transfusion as part of this trial or were already enrolled in another intervention trial. We only enrolled a child into the study if sufficient cord blood was available. All caregivers of participating children gave written informed consent. The Kenyan national ethics committee and the research ethics committee of the Liverpool School of Tropical Medicine (UK) reviewed and approved the study protocol. Procedures We obtained cord blood from placentas donated at Coast Provincial General Hospital in Mombasa. We screened all donations for HIV, hepatitis B and C viruses, and syphilis, as described previously.15 All samples were quarantined until they were screened for bacterial contamination, which we did by incubating a 4 mL sample of cord blood in 40 mL of brain-heart infusion at 37°C for 48 h, as described previously.15, 17 Screening was by microscopic examination of a Gram-stained smear. We transported units of screened cord blood by road to Kilifi (a distance of 50 km) at 2–6°C. We stored samples vertically in racks at 2–6°C to enable sedimentation of red blood cells.

brain-heart infusion at 37°C for 48 h, as described previously.15, 17 Screening was by microscopic examination of a Gram-stained smear. We transported units of screened cord blood by road to Kilifi (a distance of 50 km) at 2–6°C. We stored samples vertically in racks at 2–6°C to enable sedimentation of red blood cells. We used an electronic database to record the volume, haemoglobin concentration, and blood group of cord blood units. As soon as a blood transfusion was requested for an eligible child, we referred to this database to ascertain whether sufficient cord blood was available. The hospital clinical laboratory at Kilifi District Hospital used standard methods for blood grouping and cross-matching. At least 2·2 g/kg of haemoglobin was required for transfusion from a maximum of two group-identical or group-compatible cord blood units. Thus, we selected cord blood units based on estimated haemoglobin content, rather than volume. Furthermore, no child received a transfusion with more than 3·5 mL/kg of the preservative citrate phosphate dextrose adenine (CPDA-1).

or transfusion from a maximum of two group-identical or group-compatible cord blood units. Thus, we selected cord blood units based on estimated haemoglobin content, rather than volume. Furthermore, no child received a transfusion with more than 3·5 mL/kg of the preservative citrate phosphate dextrose adenine (CPDA-1). Research staff from the Kenya Medical Research Institute (KEMRI)/Wellcome Trust research programme provided 24 h clinical cover at Kilifi District Hospital on the paediatric wards and paediatric high-dependency unit. At admission, all children underwent structured clinical assessment, including anthropometric measurements and standard laboratory investigations, such as estimation of haemoglobin concentration (Beckman Coulter, Villepinte, France), a blood film examination for malaria, and blood culture. We did haemoglobin electrophoresis retrospectively to detect haemoglobin S in children older than 3 months. Full investigation of the cause of severe anaemia was not part of the study protocol.

n of haemoglobin concentration (Beckman Coulter, Villepinte, France), a blood film examination for malaria, and blood culture. We did haemoglobin electrophoresis retrospectively to detect haemoglobin S in children older than 3 months. Full investigation of the cause of severe anaemia was not part of the study protocol. Before the cord blood transfusion, and for a period of 24 h afterwards, we admitted children to the paediatric high-dependency unit. In children with severe acute malnutrition (defined as a weight-for-height Z score less than −3 in children older than 3 months), we transfused umbilical cord red blood cells over a period of 3 h with a maximum permitted volume of 10 mL/kg, and we administered 1 mg/kg of furosemide intravenously at the start of the transfusion, according to clinical guidelines.16 For children without severe acute malnutrition, we transfused umbilical cord red blood cells over a period of 4 h with a maximum permitted volume of 20 mL/kg, and we did not administer furosemide.

nd we administered 1 mg/kg of furosemide intravenously at the start of the transfusion, according to clinical guidelines.16 For children without severe acute malnutrition, we transfused umbilical cord red blood cells over a period of 4 h with a maximum permitted volume of 20 mL/kg, and we did not administer furosemide. For the first 2 h of the cord blood transfusion, we did continuous physiological monitoring. We recorded temperature, pulse rate, respiration rate, oxygen saturation, and blood pressure before transfusion, 15 min after the start, and every 30 min thereafter. After 2 h, we obtained a blood sample to estimate serum potassium (iLyte ion selective electrode analyser; Instrumentation Laboratory, USA) and calcium (Selectra E; Vital Scientific, Netherlands). We took a blood sample for haemoglobin estimation 24 h after the start of the cord blood transfusion, unless a haemoglobin measurement was requested for clinical management of the child before this time, in which case we used this result. We obtained a further blood sample for haemoglobin estimation from children who returned for follow-up after 28 days.

in estimation 24 h after the start of the cord blood transfusion, unless a haemoglobin measurement was requested for clinical management of the child before this time, in which case we used this result. We obtained a further blood sample for haemoglobin estimation from children who returned for follow-up after 28 days. When a child was discharged from hospital, we gave their caregiver the cost of the fare home and the return fare back to the hospital, and we invited the caregiver to bring the child to the hospital 28 days after the cord blood transfusion. We encouraged them to come back to the hospital before then if they had any concerns about their child. We recorded details of the location of the child's home. Children who returned to hospital at 28 days had a structured clinical assessment. For those who did not attend hospital, a fieldworker followed up at home, confirming whether the child was alive and well by either direct observation or discussion with an adult family member; caregivers were also encouraged to bring their child to the hospital for a full review.

had a structured clinical assessment. For those who did not attend hospital, a fieldworker followed up at home, confirming whether the child was alive and well by either direct observation or discussion with an adult family member; caregivers were also encouraged to bring their child to the hospital for a full review. Detection of adverse reactions was a two-stage process comprising rigorous surveillance of adverse events (monitoring of harm) and an independent expert judgment about the relation of the adverse event to the umbilical cord red blood cell transfusion (assessment of imputability). To capture adverse events, a clinician reviewed every child and did a study-specific structured clinical assessment 2 h after the end of the transfusion, 24 h after the end of the transfusion, and at discharge from hospital. For the remainder of the child's time in hospital, monitoring of harm was by review of the daily clinical record kept by the attending clinicians. We defined serious adverse reactions as any serious adverse event (ie, any untoward medical occurrence that is fatal, life-threatening, disabling, prolongs admission, or results in admission)18 that was judged probably or certainly related to the transfusion. We defined adverse reactions as any adverse event (ie, any untoward medical occurrence)18 judged probably or certainly related to the transfusion.

dical occurrence that is fatal, life-threatening, disabling, prolongs admission, or results in admission)18 that was judged probably or certainly related to the transfusion. We defined adverse reactions as any adverse event (ie, any untoward medical occurrence)18 judged probably or certainly related to the transfusion. The principal investigator (OWH) and an independent local safety monitor (a skilled consultant paediatrician) reviewed all serious adverse events and prepared a case summary, which was sent to the safety review committee. This committee consisted of three paediatricians who were independent of the study, with extensive experience of the clinical care of children in sub-Saharan Africa. The safety review committee and the local safety monitor reached consensus about the probability that a serious adverse event was caused by the transfusion of umbilical cord red blood cells and assigned the event an imputability score based on an established four-point scale, ranging from unlikely (0) to certain (3).19 A study clinician (FH) and the principal investigator (OWH) reviewed all other (non-serious) adverse events, which were described according to an established adverse reaction nomenclature.20 They used the same four-point imputability scale19 to score the probability of a causative relation of an adverse event with umbilical cord red blood cell transfusion. A summary of these adverse events was reviewed by a safety review committee and the local safety monitor.

n established adverse reaction nomenclature.20 They used the same four-point imputability scale19 to score the probability of a causative relation of an adverse event with umbilical cord red blood cell transfusion. A summary of these adverse events was reviewed by a safety review committee and the local safety monitor. The trial was to be stopped in the event of a suspected unexpected serious adverse reaction and not recommenced until a full review had been undertaken by the safety review committee and their recommendations seen and approved by both research ethics committees. Moreover, in the event of a serious adverse event, the safety review committee advised whether they felt that the trial should continue with no change to the protocol, continue with a change to the protocol, or be stopped. Outcomes The primary outcome measure was the frequency and nature of adverse reactions occurring during or within at least 28 days of the umbilical cord red blood cell transfusion. The secondary outcome measure was the median change from pretransfusion levels in haemoglobin concentrations 24 h and 28 days after cord blood transfusion.

mary outcome measure was the frequency and nature of adverse reactions occurring during or within at least 28 days of the umbilical cord red blood cell transfusion. The secondary outcome measure was the median change from pretransfusion levels in haemoglobin concentrations 24 h and 28 days after cord blood transfusion. In the event of an adverse reaction after a cord blood transfusion (which comprised a maximum of two blood units), imputability could not be assigned to one of the two units. Therefore, the denominator for the primary outcome was the number of children receiving a transfusion. Children who received a subsequent conventional blood transfusion during the follow-up period were included in the analysis of the primary outcome, because these transfusions could themselves be evidence of harm related to a cord blood transfusion. However, children receiving a conventional blood transfusion were not included in the analysis of haemoglobin change at 28 days, because subsequent transfusions would have confounded the effect of the cord blood transfusion.

se these transfusions could themselves be evidence of harm related to a cord blood transfusion. However, children receiving a conventional blood transfusion were not included in the analysis of haemoglobin change at 28 days, because subsequent transfusions would have confounded the effect of the cord blood transfusion. Statistical analysis We estimated from previous data that 100 children fulfilling the eligibility criteria for the trial would be admitted to Kilifi District Hospital during a period of 1 year and that cord blood would be available and consent for a transfusion given for 40–80% of these children. Thus, during 1 year of study, 40–80 children might be recruited to the trial. We intended to run the trial for 1 year; therefore, we set these numbers as a minimum and maximum sample size. The appendix (p 1) shows estimates for the frequency of adverse reactions at these minimum and maximum sample sizes. We expressed binary data as a percentage with 95% CIs where appropriate. When event frequencies were zero, we calculated a one-sided 97·5% CI with a lower limit of zero. We summarised continuous data with medians and range or IQR. We compared noted differences in continuous data with non-parametric statistics (Wilcoxon rank-sum test). This trial is registered on ISRCTN.com, number ISRCTN66687527. Role of the 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 data in the study and had final responsibility for the decision to submit for publication.

This trial is registered on ISRCTN.com, number ISRCTN66687527. Role of the 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 data in the study and had final responsibility for the decision to submit for publication. Results Between June 26, 2007, and May 20, 2008, 413 children admitted to Kilifi District Hospital, Kenya, needed a blood transfusion; of these, 87 were eligible for our trial (figure). An umbilical cord red blood cell donation of either sufficient haemoglobin content or the correct blood group was unavailable for 24 children, and the caregiver declined consent for six children. Thus, 57 children were recruited to the study. Two participants were withdrawn before umbilical cord red blood cell transfusion. In one case, the laboratory made an error during compatibility testing and no further cord blood was available. In the second case, clinical review soon after recruitment showed deep breathing, which was an exclusion criterion.

the study. Two participants were withdrawn before umbilical cord red blood cell transfusion. In one case, the laboratory made an error during compatibility testing and no further cord blood was available. In the second case, clinical review soon after recruitment showed deep breathing, which was an exclusion criterion. 55 children received umbilical cord red blood cells from 74 cord blood donations. Of these, 24 children were aged 3 months or younger and 31 were older than 3 months (table 1); the median age of children in the study was 12 months (range 2 days to 5 years 8 months). Children weighed between 1·1 kg and 14·5 kg (median, 5·3 kg). The median weight-for-height Z score was 1·9 (range −4·4 to −0·9), and seven children had severe acute malnutrition. All children with severe acute malnutrition received 10 mL/kg of umbilical cord red blood cells; for those without severe acute malnutrition, the median volume transfused was 14 mL/kg (range 10–20).

kg). The median weight-for-height Z score was 1·9 (range −4·4 to −0·9), and seven children had severe acute malnutrition. All children with severe acute malnutrition received 10 mL/kg of umbilical cord red blood cells; for those without severe acute malnutrition, the median volume transfused was 14 mL/kg (range 10–20). In children aged 3 months or younger, pretransfusion haemoglobin was a median of 87 g/L (range 55–100, IQR 78–92). In children older than 3 months, median pretransfusion haemoglobin was 32 g/L (range 19–40, IQR 27–38). In all children, within a median of 24 h (IQR 17–24) of the cord blood transfusion, haemoglobin had risen by a median of 26 g/L (IQR 21–31; table 2). 33 children did not receive a further transfusion; after median follow-up of 29 days (IQR 28–35), the median rise in haemoglobin was 50 g/L (IQR 10–68). In children aged 3 months or younger, haemoglobin rose by a median of 5 g/L (IQR 2–12) at a median of 29 days (28–36) of follow-up, compared with a median increase of 61 g/L (53–82) in children older than 3 months (median follow-up 30 days [28–35]; p<0·0001).

–35), the median rise in haemoglobin was 50 g/L (IQR 10–68). In children aged 3 months or younger, haemoglobin rose by a median of 5 g/L (IQR 2–12) at a median of 29 days (28–36) of follow-up, compared with a median increase of 61 g/L (53–82) in children older than 3 months (median follow-up 30 days [28–35]; p<0·0001). In the seven children with severe acute malnutrition (all older than 3 months), who received a maximum of 10 mL/kg umbilical cord red blood cells, the median rise in haemoglobin 24 h after transfusion was 21 g/L (IQR 20–29). In 24 children older than 3 months without severe acute malnutrition, who received between 10 mL/kg and 16 mL/kg umbilical cord red blood cells (median 13 mL/kg), the median rise in haemoglobin 24 h after transfusion was 26 g/L (IQR 22–31; p=0·15). Five of seven children with severe acute malnutrition had haemoglobin measured at 28-day follow-up; the median rise in haemoglobin in this subgroup was 81 g/L (IQR 78–82), compared with 59 g/L (IQR 53–68) in 15 of 24 children without severe acute malnutrition for whom haemoglobin concentration was measured at follow-up.

ive of seven children with severe acute malnutrition had haemoglobin measured at 28-day follow-up; the median rise in haemoglobin in this subgroup was 81 g/L (IQR 78–82), compared with 59 g/L (IQR 53–68) in 15 of 24 children without severe acute malnutrition for whom haemoglobin concentration was measured at follow-up. Of the 55 children who received an umbilical cord red blood cell transfusion, ten had a serious adverse event (one event per child) and 43 children had 94 adverse events (table 3; appendix p 2) The most frequent adverse events were anaemia (n=14), weight loss (n=12), and vomiting (n=10). In no case was transfusion of umbilical cord red blood cells judged probably or certainly implicated and, thus, the frequency of serious adverse reactions and adverse reactions was 0% (one-sided 97·5% CI 0–6·5). Of the ten serious adverse events recorded, four were new signs of critical illness (deep breathing or prostration) noted during the pretransfusion assessment, before transfusion of umbilical cord red blood cells. Therefore, the cord blood transfusion was excluded as a potential cause of these serious adverse events. Of the remaining six serious adverse events, one child died, three events were judged life-threatening, and two resulted in admission after discharge (table 4).

before transfusion of umbilical cord red blood cells. Therefore, the cord blood transfusion was excluded as a potential cause of these serious adverse events. Of the remaining six serious adverse events, one child died, three events were judged life-threatening, and two resulted in admission after discharge (table 4). Discussion Our findings show that, in a population of children from Kenya who were admitted to hospital with severe anaemia, transfusion of sedimented red blood cells from umbilical cord donations was safe and efficacious. No serious adverse events or adverse events were certainly or probably attributable to cord blood transfusion. Furthermore, the haemoglobin concentration after transfusion rose by 26 g/L after 24 h and by 50 g/L at about 28 days. Our findings accord with previous scant data for allogeneic cord-blood transfusion.9

ious. No serious adverse events or adverse events were certainly or probably attributable to cord blood transfusion. Furthermore, the haemoglobin concentration after transfusion rose by 26 g/L after 24 h and by 50 g/L at about 28 days. Our findings accord with previous scant data for allogeneic cord-blood transfusion.9 Although we excluded children with signs of critical illness at the time of study recruitment, adverse events were recorded in many participants, which were unrelated to the cord blood transfusion. In four children, signs of critical illness were detected at the clinical assessment undertaken just before cord blood transfusion. To withdraw these critically ill children from the study at that stage, and to secure and cross-match adult-donated blood, would have introduced an unacceptable delay in their management. This difficulty highlights the challenge of undertaking studies focusing on safety and harm in children admitted to hospital in sub-Saharan Africa. Robust monitoring frameworks are needed to identify potential associations between the effects of an intervention and other confounding factors. A weakness of our study is that, for children who did not attend the hospital for follow-up at 28 days, no structured clinical assessment was done. However, all children were followed up in the community by a non-clinical fieldworker, and the death of one participant was identified in this way.

her confounding factors. A weakness of our study is that, for children who did not attend the hospital for follow-up at 28 days, no structured clinical assessment was done. However, all children were followed up in the community by a non-clinical fieldworker, and the death of one participant was identified in this way. The rise in haemoglobin recorded 24 h after cord blood transfusion (median 26 g/L) accords with estimates based on the haemoglobin content of transfused blood and the circulating volume of children: for a child with a circulating volume of 80 mL/kg, transfusion of 2·2 g/kg of haemoglobin might be expected to raise the haemoglobin concentration by 28 g/L. However, although cord blood units were selected for transfusion based on an estimation of the unit haemoglobin content, we cannot ascertain from these data how much haemoglobin was actually issued and transfused. The significant rise in haemoglobin 28 days after transfusion in children older than 3 months, compared with infants aged 3 months or younger, accords with previous data from Kilifi and other sites in east Africa.3, 4, 21, 22 However, increases in haemoglobin over a similar period have also been seen in children with severe anaemia who do not receive a transfusion,3, 21, 22 which highlights the importance of other strategies to manage severe anaemia—eg, treatment of infection, use of anthelmintics and haematinics, and diet. The relative importance of these interventions will depend on the cause of anaemia, which we did not investigate here.

aemia who do not receive a transfusion,3, 21, 22 which highlights the importance of other strategies to manage severe anaemia—eg, treatment of infection, use of anthelmintics and haematinics, and diet. The relative importance of these interventions will depend on the cause of anaemia, which we did not investigate here. Infants younger than 3 months in our study were likely to have very different reasons for their anaemia compared with the older children—eg, many infants were presumed to have anaemia of prematurity. Several of these children needed further blood transfusions and, in those who did not, the effect of one umbilical cord red blood cell transfusion at 28 days was much more modest (5 g/L). However, the number of young infants who were eligible for a cord blood transfusion is noteworthy. This group of patients has a high burden of mortality in sub-Saharan Africa and potentially might benefit substantially from more evidence about the role of transfusion in prevention of high death rates.1 These young children might benefit in particular from the availability of cord blood for transfusion, because they only need small volumes of blood.

burden of mortality in sub-Saharan Africa and potentially might benefit substantially from more evidence about the role of transfusion in prevention of high death rates.1 These young children might benefit in particular from the availability of cord blood for transfusion, because they only need small volumes of blood. The microbiological safety of cord blood provided by the donation programme that we have established at Coast Provincial General Hospital in Mombasa compares favourably with that of conventional blood from the same setting.15 Mothers who donate their infants' umbilical cord blood are selected rigorously (including self-reporting of antenatal testing for syphilis and HIV), and aseptic cord blood collection is done by trained fieldworkers and not the midwives who manage the deliveries.14, 15 Furthermore, all cord blood donations in this study were screened for bacterial contamination. These rigorous techniques can be difficult to replicate outside of a research setting without additional resources.

c cord blood collection is done by trained fieldworkers and not the midwives who manage the deliveries.14, 15 Furthermore, all cord blood donations in this study were screened for bacterial contamination. These rigorous techniques can be difficult to replicate outside of a research setting without additional resources. Our findings suggest that further trials of umbilical cord red blood cell transfusions are warranted (panel), but the challenges of doing such trials and the barriers to potential scale up of such an intervention should not be underestimated. Attributing effects to the intervention is difficult in such a sick group of children. Despite this limitation, further clinical trials should also include children with signs of critical illness who potentially have the most to gain from an improved blood supply. The infrastructure and training needed to set up collection and administration of umbilical cord blood is complex, and such trials would need meticulous monitoring during and after the transfusion. Poor haemovigilance systems in these settings means that very little is known about the harms associated with conventional blood transfusion, which would be the comparator group in such trials.7

ration of umbilical cord blood is complex, and such trials would need meticulous monitoring during and after the transfusion. Poor haemovigilance systems in these settings means that very little is known about the harms associated with conventional blood transfusion, which would be the comparator group in such trials.7 Several improvements and additions could be made to the design of future trials. Better characterisation of the cause of anaemia could be included, in addition to assessment of any correlation with benefits and harms of cord blood transfusion. Immunological and genetic testing could be done to compare rates of alloimmunisation and microchimerism. Finally, operational analyses could be done to compare the availability of cord blood and adult-donated blood for urgent transfusion in children and to look at how using cord blood for transfusions in children affects the blood supply for adults who need larger volumes of transfused blood. In settings where demand for low-volume transfusions for children is high and supplies of conventional blood are low, umbilical cord blood could be a safe and effective supplementary source of blood for transfusion. Further trials comparing cord blood with conventional adult-donated blood transfusions are merited. Supplementary Material Supplementary appendix

In settings where demand for low-volume transfusions for children is high and supplies of conventional blood are low, umbilical cord blood could be a safe and effective supplementary source of blood for transfusion. Further trials comparing cord blood with conventional adult-donated blood transfusions are merited. Supplementary Material Supplementary appendix Acknowledgments This study was funded by a Wellcome Trust Training Fellowship (to OWH, number 073604). We thank Jay Berkley (local safety monitor); Victor Bandika, Michael Boele van Hensbroek, and Mike English (safety review committee); Trudie Lang, Kevin Marsh, Norbert Peshu, Sophie Uyoga, and Tom Williams; the Wazo Geni team; midwives and mothers at Coast Provincial General Hospital (Mombasa); and clinical and laboratory staff at Kilifi District Hospital. This report was published with the permission of the Director of KEMRI. Contributors OWH had the idea for the study, and all authors contributed to study design. OWH, JT, and FH managed the trial and data collection. SM, DD, KW, and KMan were responsible for laboratory procedures and quality management. OWH and GF undertook the initial data analysis. OWH wrote the first draft of the report, and all authors contributed to data interpretation and subsequent drafts of the report. Declaration of interests We declare that we have no competing interests. Figure Study profile Table 1 Selected characteristics of children at admission

Contributors OWH had the idea for the study, and all authors contributed to study design. OWH, JT, and FH managed the trial and data collection. SM, DD, KW, and KMan were responsible for laboratory procedures and quality management. OWH and GF undertook the initial data analysis. OWH wrote the first draft of the report, and all authors contributed to data interpretation and subsequent drafts of the report. Declaration of interests We declare that we have no competing interests. Figure Study profile Table 1 Selected characteristics of children at admission Infants aged ≤3 months (n=24) Infants aged >3 months (n=31) Total (n=55) Girls/boys 12/12 15/16 27/28 Sickle-cell genotype SS .. 6 6 Malaria parasites present 0 6 6 Weight-for-height Z score less than −3 .. 7 7 Preterm* 14 .. 14 Weight (kg) 1·6 (1·1–4·9) 8·6 (4·7–14·5) 5·3 (1·1–14·5) Pretransfusion haemoglobin concentration (g/L) 87 (55–100) 32 (19–40) 40 (19–100) Data are either number of participants or median (range). * Born before 37 weeks of pregnancy was completed. Table 2 Haemoglobin concentration in children receiving umbilical cord red blood cell transfusions, stratified by age

Infants aged ≤3 months (n=24) Infants aged >3 months (n=31) Total (n=55) Girls/boys 12/12 15/16 27/28 Sickle-cell genotype SS .. 6 6 Malaria parasites present 0 6 6 Weight-for-height Z score less than −3 .. 7 7 Preterm* 14 .. 14 Weight (kg) 1·6 (1·1–4·9) 8·6 (4·7–14·5) 5·3 (1·1–14·5) Pretransfusion haemoglobin concentration (g/L) 87 (55–100) 32 (19–40) 40 (19–100) Data are either number of participants or median (range). * Born before 37 weeks of pregnancy was completed. Table 2 Haemoglobin concentration in children receiving umbilical cord red blood cell transfusions, stratified by age Before transfusion Within 24 h From 24 h to 28 days Age ≤3 months Number of children 24 24 13 Haemoglobin concentration (g/L) 87 (78–92) 114 (105–124) 92 (84–99) Change in haemoglobin concentration (g/L) .. 27 (22–38) 5 (2–12) Age >3 months Number of children 31 31 20 Haemoglobin concentration (g/L) 32 (27–38) 58 (53–62) 93 (83–110) Change in haemoglobin concentration (g/L) .. 26 (21–31) 61 (53–82) All children Number of children 55 54* 33† Haemoglobin concentration (g/L) 40 (31–82) 65 (57–113) 90 (84–102) Change in haemoglobin concentration (g/L) .. 26 (21–31) 50 (10–68) Data are median (IQR), unless otherwise stated. * Excludes one child for whom consent for a blood test was declined. † Excludes 22 children: ten received further transfusions; 11 were followed up in the community; and one was not bled in error. Table 3 Serious adverse events and adverse events in children receiving umbilical cord red blood cell transfusions

Before transfusion Within 24 h From 24 h to 28 days Age ≤3 months Number of children 24 24 13 Haemoglobin concentration (g/L) 87 (78–92) 114 (105–124) 92 (84–99) Change in haemoglobin concentration (g/L) .. 27 (22–38) 5 (2–12) Age >3 months Number of children 31 31 20 Haemoglobin concentration (g/L) 32 (27–38) 58 (53–62) 93 (83–110) Change in haemoglobin concentration (g/L) .. 26 (21–31) 61 (53–82) All children Number of children 55 54* 33† Haemoglobin concentration (g/L) 40 (31–82) 65 (57–113) 90 (84–102) Change in haemoglobin concentration (g/L) .. 26 (21–31) 50 (10–68) Data are median (IQR), unless otherwise stated. * Excludes one child for whom consent for a blood test was declined. † Excludes 22 children: ten received further transfusions; 11 were followed up in the community; and one was not bled in error. Table 3 Serious adverse events and adverse events in children receiving umbilical cord red blood cell transfusions Serious adverse events (n=10) Adverse events (n=94) Children (n) 10 43 Event timing Before transfusion 4 4 During transfusion and within 24 h afterwards 1 12 From 24 h to 28 days after transfusion 5 78 Indicator of severity Fatal 1 N/A Life-threatening 7 N/A Admission 2 N/A Imputability level* Not assessable 0 0 Excluded (0) 4 15 Unlikely (0) 3 76 Possible (1) 3 3 Likely/probable (2) 0 0 Certain (3) 0 0 N/A=not applicable. * Numbers in parentheses refer to four-point imputability score. Table 4 Serious adverse events within 1 month of umbilical cord red blood cell transfusion

Serious adverse events (n=10) Adverse events (n=94) Children (n) 10 43 Event timing Before transfusion 4 4 During transfusion and within 24 h afterwards 1 12 From 24 h to 28 days after transfusion 5 78 Indicator of severity Fatal 1 N/A Life-threatening 7 N/A Admission 2 N/A Imputability level* Not assessable 0 0 Excluded (0) 4 15 Unlikely (0) 3 76 Possible (1) 3 3 Likely/probable (2) 0 0 Certain (3) 0 0 N/A=not applicable. * Numbers in parentheses refer to four-point imputability score. Table 4 Serious adverse events within 1 month of umbilical cord red blood cell transfusion Time after transfusion, and description of serious adverse event Indicator of severity Comments Imputability level (score) Girl, aged 13 months Acute leukaemia, probably acute lymphoblastic leukaemia Fatal Diagnosis of leukaemia made by microscopic examination of a peripheral blood film taken before cord blood transfusion but reported afterwards; child was referred to regional hospital but discharged against medical advice and taken home; child was pale, sick-looking, and febrile at discharge; child's grandfather reported she died the day after arriving home, about 1 week after cord blood transfusion Unlikely (0) Boy, aged 2 days At 7 days, abdominal distension, respiratory distress, jaundice Life-threatening Infant was born preterm (estimated gestation, 30 weeks; birthweight, 1440 g) with probable neonatal sepsis; full recovery after intervention with broad-spectrum antibiotics, nil by mouth, intravenous fluid, and phototherapy Unlikely (0) Boy, aged 4 years At 28 days, anaemia Admission Child had known sickle-cell disease; was discharged well 3 days after cord blood transfusion (haemoglobin 62 g/L) with haematinics; returned for follow-up at 21 days (haemoglobin 66 g/L); at 28-day follow-up, haemoglobin was 41 g/L; child admitted for conventional blood transfusion; discharged next day (haemoglobin 61 g/L) Unlikely (0) Boy, aged 24 days After 26 h, diarrhoea, dehydration, and metabolic acidosis Life-threatening Preterm infant (estimated gestation, 28 weeks; birthweight, 1080 g) with probable neonatal sepsis; full recovery after intervention with oxygen, broad-spectrum antibiotics, and intravenous fluid Possible (1) Girl, aged 9 days After 14 h, sepsis, pneumonia, and apnoea events Life-threatening Preterm infant (estimated gestation, 30 weeks; birthweight, 1200 g); pneumonia was confirmed by radiography; full recovery after intervention with broad-spectrum antibiotics, oxygen, aminophylline, and blood transfusion Possible (1) Girl, aged 5 days At 28 days, anaemia Admission Baby had a low birthweight (probable prematurity), neonatal jaundice, and possible sepsis; was discharged well 3 days after cord blood transfusion (haemoglobin 115 g/L); at 28-day follow-up,

-spectrum antibiotics, oxygen, aminophylline, and blood transfusion Possible (1) Girl, aged 5 days At 28 days, anaemia Admission Baby had a low birthweight (probable prematurity), neonatal jaundice, and possible sepsis; was discharged well 3 days after cord blood transfusion (haemoglobin 115 g/L); at 28-day follow-up, haemoglobin was 62 g/L; admission offered but declined; child presented again 1 week later and was admitted (haemoglobin 84 g/L); conventional blood transfusion given once donor found (initially no blood available); discharged next day with haematinics Possible (1) Ages of children are at enrolment to the study. Panel Research in context Systematic review

haemoglobin was 62 g/L; admission offered but declined; child presented again 1 week later and was admitted (haemoglobin 84 g/L); conventional blood transfusion given once donor found (initially no blood available); discharged next day with haematinics Possible (1) Ages of children are at enrolment to the study. Panel Research in context Systematic review A review of published work relating to cord blood transfusion was done before the study (April, 2007) and repeated before submission (September, 2014). We searched PubMed with the following search query: (“cord blood”[Title/Abstract] OR “placental blood”[Title/Abstract] AND “transfusion”[Title/Abstract]); we did not restrict by date, language, or article type. The reference lists of articles identified were also screened for relevant titles. Allogeneic cord blood transfusion was first reported in the 1930s, before the advent of modern blood transfusion services.23 Subsequently, most research and clinical activity relating to cord blood transfusion has concerned autologous cord blood transfusion in preterm neonates.11 In India, a series of about 200 mainly elderly patients with chronic and terminal disease has received allogeneic cord blood transfusions.9, 11 No adverse reactions were reported. To our knowledge, our study is the first clinical trial of allogeneic cord blood transfusion in children. Interpretation

A review of published work relating to cord blood transfusion was done before the study (April, 2007) and repeated before submission (September, 2014). We searched PubMed with the following search query: (“cord blood”[Title/Abstract] OR “placental blood”[Title/Abstract] AND “transfusion”[Title/Abstract]); we did not restrict by date, language, or article type. The reference lists of articles identified were also screened for relevant titles. Allogeneic cord blood transfusion was first reported in the 1930s, before the advent of modern blood transfusion services.23 Subsequently, most research and clinical activity relating to cord blood transfusion has concerned autologous cord blood transfusion in preterm neonates.11 In India, a series of about 200 mainly elderly patients with chronic and terminal disease has received allogeneic cord blood transfusions.9, 11 No adverse reactions were reported. To our knowledge, our study is the first clinical trial of allogeneic cord blood transfusion in children. Interpretation In children with severe anaemia in sub-Saharan Africa, transfusion of sedimented red blood cells obtained from umbilical cord blood has a low probability of adverse events. Haemoglobin recovery after cord blood transfusion is within expected limits. Umbilical cord blood could be a safe and efficacious supplementary source of blood for transfusion when demand for low-volume transfusions for children is high and supplies of conventional blood are limited. Further work needs to be undertaken by clinical researchers to establish the safety and efficacy of cord blood transfusion compared with conventional blood transfusion. Additional research is also needed on the operational aspects of cord blood collection, including costs, the effect on conventional blood supply, and scalability.

Introduction Worldwide inequalities in health and health care are reflected in global differences in life expectancy and overall mortality in both adults and children,1 and the findings from several studies have highlighted global differences in cancer incidence2 and survival.3, 4 Diagnostic techniques and treatment for childhood leukaemia have improved since the 1990s. Access to these techniques and treatment has, however, been limited in some countries, partly by a shortage of resources.5 Leukaemias, a heterogeneous group of diseases of mostly unknown origin, are globally the most common malignancies in children (aged 0–14 years), except in Africa.6 Unlike in adults, acute lymphoid leukaemias are the commonest subtype in children (accounting for approximately 80% of cases), and acute myeloid leukaemia (AML) represents about 15% of cases. For both types, incidence varies widely with age; in lymphoid leukaemias, incidence is slightly higher in boys than girls, and in industrialised high-income countries (HIC).7 In low-income and middle-income countries (LMIC), where the population is young, the incidence of childhood leukaemias is lower than in HIC, but these diseases are still responsible for many deaths.2, 5 Research in context Evidence before this study

Leukaemias, a heterogeneous group of diseases of mostly unknown origin, are globally the most common malignancies in children (aged 0–14 years), except in Africa.6 Unlike in adults, acute lymphoid leukaemias are the commonest subtype in children (accounting for approximately 80% of cases), and acute myeloid leukaemia (AML) represents about 15% of cases. For both types, incidence varies widely with age; in lymphoid leukaemias, incidence is slightly higher in boys than girls, and in industrialised high-income countries (HIC).7 In low-income and middle-income countries (LMIC), where the population is young, the incidence of childhood leukaemias is lower than in HIC, but these diseases are still responsible for many deaths.2, 5 Research in context Evidence before this study In 2015, the CONCORD-2 study initiated surveillance of survival trends for childhood leukaemia at a worldwide scale. Results from CONCORD-2 identified huge worldwide variation in 5-year net survival for children diagnosed with precursor-cell acute lymphoblastic leukaemia or lymphoma (ALL) during 1995-2009. Our analysis extends those results to cover survival trends for several subtypes of childhood leukaemia, grouped according to the third edition of the International Classification of Childhood Cancer. Added value of this study

In 2015, the CONCORD-2 study initiated surveillance of survival trends for childhood leukaemia at a worldwide scale. Results from CONCORD-2 identified huge worldwide variation in 5-year net survival for children diagnosed with precursor-cell acute lymphoblastic leukaemia or lymphoma (ALL) during 1995-2009. Our analysis extends those results to cover survival trends for several subtypes of childhood leukaemia, grouped according to the third edition of the International Classification of Childhood Cancer. Added value of this study We included 89 828 children diagnosed with leukaemia during 1995–2009 in 53 countries. Despite substantial improvements in survival from childhood acute myeloid leukaemia (AML) in most countries during 1995–2009, huge international disparities in 5-year survival persisted up to 2009, matching those previously reported in the 2015 CONCORD-2 paper for childhood ALL. 5-year age-standardised net survival from AML (ie, the probability of surviving at least 5 years after diagnosis) was consistently lower than 5-year age-standardised net survival from ALL, but the difference narrowed in most countries since the early 2000s. Children aged 1–9 years at diagnosis had higher 5-year net survival than older or younger children, both for ALL and AML. Survival for older children (10–14 years) improved by 2009, but infants (aged <1 year) diagnosed with either ALL or AML still had the lowest 5-year net survival. Implications of all the available evidence

We included 89 828 children diagnosed with leukaemia during 1995–2009 in 53 countries. Despite substantial improvements in survival from childhood acute myeloid leukaemia (AML) in most countries during 1995–2009, huge international disparities in 5-year survival persisted up to 2009, matching those previously reported in the 2015 CONCORD-2 paper for childhood ALL. 5-year age-standardised net survival from AML (ie, the probability of surviving at least 5 years after diagnosis) was consistently lower than 5-year age-standardised net survival from ALL, but the difference narrowed in most countries since the early 2000s. Children aged 1–9 years at diagnosis had higher 5-year net survival than older or younger children, both for ALL and AML. Survival for older children (10–14 years) improved by 2009, but infants (aged <1 year) diagnosed with either ALL or AML still had the lowest 5-year net survival. Implications of all the available evidence Data obtained in the CONCORD programme provide a unique opportunity to explore disparities in survival from childhood leukaemia at an unprecedented scale. The results suggest that good access to health care and appropriate treatment have a clear population effect on survival for children with leukaemia. The findings support the need for continuing international efforts to improve worldwide access to appropriate cancer care for children. They can also be used to assess the effect of cancer strategies targeting childhood cancers.

propriate treatment have a clear population effect on survival for children with leukaemia. The findings support the need for continuing international efforts to improve worldwide access to appropriate cancer care for children. They can also be used to assess the effect of cancer strategies targeting childhood cancers. Although cancer mortality trends provide a useful measure of the societal cancer burden, they depend on trends in both incidence and survival. Cancer survival is the probability that cancer patients survive up to a certain point after diagnosis. Observed survival and event-free survival are clinically important, but population-based net survival is the appropriate indicator for comparisons between populations. Net survival is the probability of surviving after controlling for mortality from other causes.

patients survive up to a certain point after diagnosis. Observed survival and event-free survival are clinically important, but population-based net survival is the appropriate indicator for comparisons between populations. Net survival is the probability of surviving after controlling for mortality from other causes. The CONCORD programme was designed to address the shortage of globally comparable data on population-based cancer survival.3, 4 Population-based cancer survival reflects several aspects of health care, from screening and early diagnosis, to access to effective treatment. This metric is increasingly used as a measure of the effectiveness of health-care systems in the management of cancer, and to assess the effectiveness of national cancer plans.8, 9 The second cycle of CONCORD (CONCORD-2)4 established global surveillance of population-based cancer survival for patients diagnosed with one of ten common cancers, or childhood leukaemia, during the 15-year period 1995–2009, using data from 279 population-based cancer registries in 67 countries. In this analysis, we examine worldwide trends in survival from precursor-cell acute lymphoblastic leukaemia (ALL) in children, by age and sex, alongside trends in survival from acute myeloid leukaemia (AML) and other types of childhood leukaemia.

, using data from 279 population-based cancer registries in 67 countries. In this analysis, we examine worldwide trends in survival from precursor-cell acute lymphoblastic leukaemia (ALL) in children, by age and sex, alongside trends in survival from acute myeloid leukaemia (AML) and other types of childhood leukaemia. Methods Search strategy and selection criteria Cancer registries participating in CONCORD-2 were asked to submit tumour registrations for all children (aged 0–14 years) diagnosed with a haematological malignancy between Jan 1, 1995, and Dec 31, 2009, including information about their vital status at Dec 31, 2009.4 Depending on the registry, patients were followed up actively, via direct investigation, or passively, using linkage to national or regional databases of death.4 Haematological malignancies were defined by morphology codes in the range 9590–9989 in the International Classification of Diseases for Oncology, third revision (ICD-O-3).10 215 registries in 60 countries submitted data on 126 830 children with a haematological malignancy. We excluded data from 13 registries from which the data were judged to be less reliable,4 or included only lymphomas (two), and data from countries for which fewer than 10 children were available for analysis (two). This left 124 015 records for children with a haematological malignancy.

th a haematological malignancy. We excluded data from 13 registries from which the data were judged to be less reliable,4 or included only lymphomas (two), and data from countries for which fewer than 10 children were available for analysis (two). This left 124 015 records for children with a haematological malignancy. We did standardised data cleaning in three phases, as detailed previously.4 Records that were ineligible (eg, for patients aged 15 years or older), inaccurate or inappropriate for survival analysis (eg, incoherent date sequence, or registration only from a death certificate or autopsy report) were excluded.4, 11 For patients with more than one record of a haematological malignancy diagnosed during 1995–2009, we kept only the record of the first malignancy. A few registries submitted records coded to earlier revisions of ICD-O or to the first revision of ICD-O-3. In agreement with these registries, we recoded those morphology codes to be compliant with ICD-O-3. Data from Sétif (Algeria), Arkhangelsk (Russia), Wrocław (Poland), and Northern Ireland (UK) were included after their data were recoded to ICD-O-3.

oded to earlier revisions of ICD-O or to the first revision of ICD-O-3. In agreement with these registries, we recoded those morphology codes to be compliant with ICD-O-3. Data from Sétif (Algeria), Arkhangelsk (Russia), Wrocław (Poland), and Northern Ireland (UK) were included after their data were recoded to ICD-O-3. Data analysis We estimated 5-year net survival with the Pohar-Perme estimator12 using the STNS command implemented in Stata 13. We used the life tables of background mortality rates by sex, single year of age, calendar year and—for USA, Israel, Malaysia, and New Zealand—by race or ethnic group, produced for CONCORD-2.13 For each country, we estimated net survival by calendar period of diagnosis (1995–99, 2000–04, and 2005–09). We used the classic cohort approach for children diagnosed during 1995–99 and 2000–04, because 5 years of follow-up data were available for all children. We used the period approach to predict 5-year survival for leukaemias diagnosed more recently (2005–09), as this approach allows for the prediction of survival where 5 years of follow-up are not yet available.14 Survival estimates for each country were based on data from a national registry or from one or several subnational registries. We excluded data from some regional registries from the pooled estimate for a given country if data quality or information about vital status were deemed unsatisfactory.4 Country estimates were flagged if data quality was considered less reliable.

from a national registry or from one or several subnational registries. We excluded data from some regional registries from the pooled estimate for a given country if data quality or information about vital status were deemed unsatisfactory.4 Country estimates were flagged if data quality was considered less reliable. We estimated 5-year survival by sex and age at diagnosis (<1, 1–4, 5–9, 10–14 years, inclusive). Exact age at diagnosis was calculated from the dates of birth and diagnosis. The rules adopted to impute missing components of dates have been described previously.4, 11 Age-standardised survival was calculated from three equally weighted age-specific estimates (0–4, 5–9, and 10–14 years).15 Data for age groups with fewer than ten patients were pooled with data for the adjacent age group; we then re-estimated survival for both age groups combined, and the pooled estimate was attributed to each age group. Leukaemias were grouped according to the International Classification of Childhood Cancer (ICCC-3).16 We estimated survival for all lymphoid leukaemias combined (ICCC-3 group Ia), for acute myeloid leukaemias (AML; Ib), and for unspecified and other specified leukaemias (Ie) (appendix p 5). We also estimated survival separately for two subgroups of lymphoid leukaemia: precursor-cell lymphoid leukaemias (ALL; Ia1) and mature B-cell leukaemias (Ia2). We did not analyse survival for chronic myeloproliferative diseases (group Ic) or myelodysplastic syndrome and other myeloproliferative diseases (group Id).

). We also estimated survival separately for two subgroups of lymphoid leukaemia: precursor-cell lymphoid leukaemias (ALL; Ia1) and mature B-cell leukaemias (Ia2). We did not analyse survival for chronic myeloproliferative diseases (group Ic) or myelodysplastic syndrome and other myeloproliferative diseases (group Id). Ethical approval for access to the data was obtained from the Ethics and Confidentiality Committee of the UK's statutory National Information Governance Board (now the Health Research Authority; ECC 3-04(i)/2011) and the UK National Health Service (NHS) Research Ethics Service (South-East; 11/LO/0331), and from other jurisdictions as required.4 Role of the funding source The funders 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 data in the study and had final responsibility for the decision to submit for publication.

Ethical approval for access to the data was obtained from the Ethics and Confidentiality Committee of the UK's statutory National Information Governance Board (now the Health Research Authority; ECC 3-04(i)/2011) and the UK National Health Service (NHS) Research Ethics Service (South-East; 11/LO/0331), and from other jurisdictions as required.4 Role of the funding source The funders 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 data in the study and had final responsibility for the decision to submit for publication. Results Of the 124 015 children who were considered for analysis, we excluded 1623 (1%) as ineligible, usually because of missing information on date of birth, diagnosis, or last known vital status (table 1). More than 75% of records from the Tunisia Central Registry were ineligible because of incomplete data. Overall, only 0·5% of records were excluded because the registration was based solely on a death certificate or an autopsy report. We excluded five patients with synchronous leukaemia and lymphoma, and 106 children whose leukaemia followed a lymphoma, also diagnosed during 1995–2009. We also excluded 2222 children with chronic myeloproliferative disease (ICCC-3 group Ic) and 2002 children with myelodysplastic syndrome or other myeloproliferative disease (Id). Lymphomas (27 609) were not included.

ia and lymphoma, and 106 children whose leukaemia followed a lymphoma, also diagnosed during 1995–2009. We also excluded 2222 children with chronic myeloproliferative disease (ICCC-3 group Ic) and 2002 children with myelodysplastic syndrome or other myeloproliferative disease (Id). Lymphomas (27 609) were not included. Data quality was generally very high, and a high proportion of diagnoses were reported with a specific morphology. Improvements in diagnosis and registration are illustrated in Latvia by the drop in the number of unspecified and other specified leukaemias (group Ie) between 1995–99 and 2005–09. We focused our analyses on 89 828 children (73·8% of all haematological malignancies) from 198 registries in 53 countries who were diagnosed with lymphoid leukaemia (ICCC-3 group Ia), acute myeloid leukaemia (Ib), or unspecified or other specified leukaemia (Ie). For children diagnosed during 1995–2004 who were not known to have died, the median follow-up was at least 5 years in all participating countries in North, Central, and South America, Europe, Asia, and Oceania. In Africa, the median follow-up of surviving children diagnosed up to 2004 was 0·7 years (IQR 0·2–3·0) in the two Algerian registries and 1·7 years (1·0–2·2) in the Tunisia Central Registry, although the maximum follow-up was 8·9 and 13·3 years, respectively.

orth, Central, and South America, Europe, Asia, and Oceania. In Africa, the median follow-up of surviving children diagnosed up to 2004 was 0·7 years (IQR 0·2–3·0) in the two Algerian registries and 1·7 years (1·0–2·2) in the Tunisia Central Registry, although the maximum follow-up was 8·9 and 13·3 years, respectively. All lymphoid leukaemias combined represented 81% of leukaemias, AML 16%, and unspecified and other specified leukaemias the remaining 3% (table 1). In Lesotho, no AML was registered during 1995–2009. Information from the cancer registries in Belarus, Argentina, and Colombia (Cali) was only available for lymphoid leukaemias. Precursor-cell ALL was by far the most common type of lymphoid leukaemia, but some registries submitted data on rarer types, such as mature B-cell (mostly Burkitt leukaemia), mature T and Natural-Killer (NK) cell leukaemias (figure 1). Due to the rarity of mature T-cell and NK-cell leukaemia (n=94, ICCC-3 group Ia3) and unspecified lymphoid leukaemias (n=446, Ia4), we did not estimate survival separately for these morphological groups. In the data from Indonesia (Jakarta), 17 Chinese registries, and Latvia, more than 25% of the childhood leukaemias were coded as unspecified or other specified leukaemia (ICCC-3 group Ie).

roup Ia3) and unspecified lymphoid leukaemias (n=446, Ia4), we did not estimate survival separately for these morphological groups. In the data from Indonesia (Jakarta), 17 Chinese registries, and Latvia, more than 25% of the childhood leukaemias were coded as unspecified or other specified leukaemia (ICCC-3 group Ie). During 1995–99, 5-year age-standardised net survival for all lymphoid leukaemias combined ranged from 10·6% (95% CI 3·1–18·2) in the Chinese registries to 86·8% (81·6–92·0) in Austria (table 2). This wide range in survival narrowed over time, with survival in 2005–09 ranging from 52·4% (42·8–61·9) in Cali, Colombia, to 91·6% (89·5–93·6) in the German registries. Survival from precursor-cell ALL was very close to that of all lymphoid leukaemias combined, with similar variation (figure 2). The greatest absolute difference in survival between all lymphoid leukaemias and ALL was noted in Iceland (80·9% and 90·1%, respectively) but these estimates were not age-standardised because of small numbers. Survival from precursor-cell ALL increased between 1995–99 and 2005–09 in most countries (figure 3). We estimated survival for mature B-cell leukaemia for 17 countries. Many of these estimates were based on a pooled analysis for children diagnosed throughout 1995–2009, because the number of patients diagnosed in each 5-year period was small. In France and the US registries, 5-year age-standardised net survival from mature B-cell leukaemia was lower than that of precursor-cell ALL in 1995–99, but the difference had disappeared by 2005–09.

s for children diagnosed throughout 1995–2009, because the number of patients diagnosed in each 5-year period was small. In France and the US registries, 5-year age-standardised net survival from mature B-cell leukaemia was lower than that of precursor-cell ALL in 1995–99, but the difference had disappeared by 2005–09. 5-year age-standardised net survival for AML was consistently lower than that for ALL (table 2, figure 2). Age-standardised survival for AML in 1995–99 ranged from 4·2% (0·0–8·6) in the Chinese registries to 72·2% (61·6–82·7) in Sweden, and from 33·3% (18·9–47·7) in Bulgaria to 78·2% (72·0–84·3) in German registries for 2005–09. In most countries, survival from childhood AML increased quite remarkably over time (figure 3). Age-standardised 5-year survival for unspecified and other specified leukaemia (group Ie) showed wide variation, ranging from 13·5% (Bulgaria) to 69·6% (Latvia) in 1995–99, and from 26·0% (Chinese registries) to 89·2% (Israel) in 2005–09.

ies, survival from childhood AML increased quite remarkably over time (figure 3). Age-standardised 5-year survival for unspecified and other specified leukaemia (group Ie) showed wide variation, ranging from 13·5% (Bulgaria) to 69·6% (Latvia) in 1995–99, and from 26·0% (Chinese registries) to 89·2% (Israel) in 2005–09. Survival from ALL in infants (aged <1 year) was much lower than for older children, including those aged 10–14 years (appendix pp 6–14). By contrast, survival for infants with AML in many countries was close to that for children aged 10–14 years (appendix pp 15–22). Children aged 1–4 and 5–9 years had the highest survival for both ALL and AML. The difference in survival between children with ALL aged 10–14 and those aged 1–4 and 5–9 years fell progressively between 1995–99 and 2005–09 in most countries. The pattern was not so clear for AML. Survival from AML and ALL was often slightly higher for girls than for boys, but this pattern was not consistent across all countries (appendix pp 6–22). Survival trends for precursor-cell ALL and AML were not markedly different between 1995–99 and 2000–04 (figure 4); however, between 2000–04 and 2005–09, survival from AML increased more than survival from ALL in most countries (figure 4B), particularly in Thailand and Switzerland.

Survival from ALL in infants (aged <1 year) was much lower than for older children, including those aged 10–14 years (appendix pp 6–14). By contrast, survival for infants with AML in many countries was close to that for children aged 10–14 years (appendix pp 15–22). Children aged 1–4 and 5–9 years had the highest survival for both ALL and AML. The difference in survival between children with ALL aged 10–14 and those aged 1–4 and 5–9 years fell progressively between 1995–99 and 2005–09 in most countries. The pattern was not so clear for AML. Survival from AML and ALL was often slightly higher for girls than for boys, but this pattern was not consistent across all countries (appendix pp 6–22). Survival trends for precursor-cell ALL and AML were not markedly different between 1995–99 and 2000–04 (figure 4); however, between 2000–04 and 2005–09, survival from AML increased more than survival from ALL in most countries (figure 4B), particularly in Thailand and Switzerland. Discussion This study provides the largest population-based comparison of survival from childhood leukaemia. It covers trends over 15 years between 1995–2009 in 53 countries, of which 17 were classified by the World Bank as low-income or middle-income countries (LMIC) in 2011. The results highlight the very wide international differences in 5-year net survival for children with acute lymphoid leukaemia and for children with myeloid leukaemia. 5-year net survival has been increasing for both precursor-cell ALL and AML, but it remains consistently higher for precursor-cell ALL than for AML; this worldwide pattern is consistent with previous studies in various regions of the world.7, 17, 18, 19, 20 For many LMICs, analyses and interpretation are limited by sparse data. Several survival estimates could not be age-standardised. However, our results show that, overall, survival has been increasing for ALL and for AML in most countries during 1995–2009, including both high-income and LMICs.

he world.7, 17, 18, 19, 20 For many LMICs, analyses and interpretation are limited by sparse data. Several survival estimates could not be age-standardised. However, our results show that, overall, survival has been increasing for ALL and for AML in most countries during 1995–2009, including both high-income and LMICs. 5-year survival for both ALL and AML was very high in Germany and Austria. This might be attributable to the very tight adherence of paediatric haematologists and oncologists in those countries to the protocols and trials of the BFM Group (Berlin, Frankfurt, Muenster), within the framework of a national paediatric cancer registry and reference laboratories, imaging review and tumour boards.21 In most countries, the difference in survival between ALL and AML tended to narrow over the period 1995–2009. This happened both in countries where survival from ALL was very high throughout 1995–2009, and in countries where survival from ALL was low in 1995–99, but increased up to 2005–09. The greater improvement in 5-year survival from AML than from ALL might be related to recent improvements in the care of childhood AML. Our results might reflect the effect of better diagnostic characterisation, risk stratification and subsequent adaptation of treatment, and restriction of indications to cranial radiotherapy and haemopoietic stem-cell transplantation.22

than from ALL might be related to recent improvements in the care of childhood AML. Our results might reflect the effect of better diagnostic characterisation, risk stratification and subsequent adaptation of treatment, and restriction of indications to cranial radiotherapy and haemopoietic stem-cell transplantation.22 Net survival from AML seemed to approach the level of survival from precursor-cell ALL earlier in some countries than in others. This suggests that improvements in clinical practice have not been implemented in all countries at the same time. In Switzerland, there was a large increase in survival from AML between 1995–2004 and 2005–09, by which time it approached the level of survival from ALL. In Thailand, national protocols were introduced for childhood leukaemia in 2006. This may have contributed to the marked improvement in survival from AML, but did not lead to substantial improvements in survival for ALL.23 In China, despite the impressive increases in survival from AML and ALL throughout 1995–2009, there was no reduction in the difference in survival between ALL and AML.

a in 2006. This may have contributed to the marked improvement in survival from AML, but did not lead to substantial improvements in survival for ALL.23 In China, despite the impressive increases in survival from AML and ALL throughout 1995–2009, there was no reduction in the difference in survival between ALL and AML. This large study offered a unique opportunity to examine age-standardised trends in population-based survival from some of the rarer childhood leukaemias, such as mature B-cell leukaemias, mostly Burkitt's leukaemia. In France and the USA, where survival from mature B-cell leukaemia could be age-standardised, the most recent estimates showed that survival from mature B-cell leukaemia was close to that of precursor-cell ALL. A degree of misclassification between precursor-cell ALL and Burkitt's leukaemia is likely because of their common historical classification and the remaining ambiguity in coding, but this is unlikely to explain the increasing survival trends for mature B-cell leukaemia in France and the USA. Leukaemia survival has been reported as higher in girls than in boys, both in Europe from 1970 up to the 1990s,24 and still recently in North America,18 but this is not a consistent feature worldwide.

This large study offered a unique opportunity to examine age-standardised trends in population-based survival from some of the rarer childhood leukaemias, such as mature B-cell leukaemias, mostly Burkitt's leukaemia. In France and the USA, where survival from mature B-cell leukaemia could be age-standardised, the most recent estimates showed that survival from mature B-cell leukaemia was close to that of precursor-cell ALL. A degree of misclassification between precursor-cell ALL and Burkitt's leukaemia is likely because of their common historical classification and the remaining ambiguity in coding, but this is unlikely to explain the increasing survival trends for mature B-cell leukaemia in France and the USA. Leukaemia survival has been reported as higher in girls than in boys, both in Europe from 1970 up to the 1990s,24 and still recently in North America,18 but this is not a consistent feature worldwide. As expected, age at diagnosis was an important determinant of 5-year survival: children diagnosed with precursor-cell ALL aged 1–4 years consistently had the best prognosis. Survival from leukaemia in infants (under 1 year) was usually lower than at older ages. Survival was lower for infants with precursor-cell ALL than for infants with AML. This might be because unfavourable prognostic factors such as MLL gene rearrangements are more frequent in infants with ALL than AML. The new coding rules to individualise ALLs and AMLs with specific genetic rearrangements should enable analysis of more specific subgroups in the future.25

ALL than for infants with AML. This might be because unfavourable prognostic factors such as MLL gene rearrangements are more frequent in infants with ALL than AML. The new coding rules to individualise ALLs and AMLs with specific genetic rearrangements should enable analysis of more specific subgroups in the future.25 Interpretation might be restricted by changes in the clinical definition of leukaemias and lymphomas over time, and differences in coding between registries. The classifications for grouping the types of leukaemia and lymphoma have also changed over decades. ICCC-3 was proposed in 2005,16 and, in 2010, the HAEMACARE Working Group proposed a classification for haematological malignancies in both adults and children.26 Results shown here were based on ICCC-3, which is commonly used for childhood cancer surveillance.7, 17 Using the HAEMACARE classification, we noted similar results for precursor-cell ALL and AML, but some differences for Burkitt's leukaemia, probably due to the broader grouping defined by HAEMACARE (data not shown). We did not have data to show the effect on survival estimates of including endemic Burkitt's lymphoma in Africa in that group.

lassification, we noted similar results for precursor-cell ALL and AML, but some differences for Burkitt's leukaemia, probably due to the broader grouping defined by HAEMACARE (data not shown). We did not have data to show the effect on survival estimates of including endemic Burkitt's lymphoma in Africa in that group. Children are more likely to be diagnosed and registered with a poorly specified type of leukaemia in settings where access to a pathologist is less than optimal. This could have clinical implications, with children not gaining access to the most appropriate treatment for their particular leukaemia.27 Where the estimates for this poorly specified group of leukaemias could be age-standardised, they indicate that in high-income countries, survival was between that of precursor-cell ALL and AML. For the Chinese registries, where the proportion of unspecified and other specified leukaemias remained higher than 25% throughout 1995–2009, survival for those leukaemias was lower than that for AML. This suggests uncertain diagnosis and insufficient or inappropriate treatment. Implementation of the recent resource-stratified guidelines for the management of childhood ALL in Asia should lead to better diagnostic characterisation, more appropriate treatment and higher survival.28 In Colombia, the low survival estimate for lymphoid leukaemia (52%) suggests the need for further investigation, since childhood cancer became a national priority in 2010.

nes for the management of childhood ALL in Asia should lead to better diagnostic characterisation, more appropriate treatment and higher survival.28 In Colombia, the low survival estimate for lymphoid leukaemia (52%) suggests the need for further investigation, since childhood cancer became a national priority in 2010. Leukaemia is the most common malignancy in children in most countries,6 and we used population-based data, but comparisons of survival by age and type of leukaemia were sometimes limited by low numbers, especially in small populations. Some age-standardised survival estimates have wide CIs. This was particularly true for AML and mature B-cell leukaemia, and for all types of leukaemia in infants. Survival estimates for Tunisian and Algerian registries were less reliable than for other countries. We used the age range 0–14 years for our analysis because this range has been the standard in childhood cancer studies for many years. Adolescents and young adults are increasingly being treated under paediatric protocols, but our choice was agreed with 100 collaborators during a 2-day meeting of the CONCORD Working Group (Cork, Ireland, 2012), at which the protocol was finalised. Detailed analyses of survival for adolescents, young adults, and older adults with leukaemia are under preparation for publication elsewhere.

iatric protocols, but our choice was agreed with 100 collaborators during a 2-day meeting of the CONCORD Working Group (Cork, Ireland, 2012), at which the protocol was finalised. Detailed analyses of survival for adolescents, young adults, and older adults with leukaemia are under preparation for publication elsewhere. Other studies of childhood cancer survival have not used net survival. The US SEER programme presents relative survival.18 The European programmes, EUROCARE17 and ACCIS,7 present observed survival, because background mortality in children does not vary greatly between European populations. The CONCORD-2 study has worldwide coverage, so we estimated net survival, because infant mortality varied particularly widely between participating countries.4 For example, infant mortality rates in males covered a 25-fold range in 2007, from 3·0 per 1000 livebirths in Finland to 82·5 per 1000 livebirths in Lesotho.13

The CONCORD-2 study has worldwide coverage, so we estimated net survival, because infant mortality varied particularly widely between participating countries.4 For example, infant mortality rates in males covered a 25-fold range in 2007, from 3·0 per 1000 livebirths in Finland to 82·5 per 1000 livebirths in Lesotho.13 Population-based cancer registry data include all or nearly all cases of a given malignancy in each registry's jurisdiction. By contrast with the best achievable survival estimates obtained from patients included in clinical trials, population-based survival estimates reflect the survival of all cancer patients in the population, irrespective of socioeconomic status and disease features. They reflect the overall effectiveness of the health system, from parents' perception of how to respond to symptoms suggestive of malignancy in their child, as well as the efficiency of referral, the quality of investigation and treatment, and the resourcing and organisation of the health service.

and disease features. They reflect the overall effectiveness of the health system, from parents' perception of how to respond to symptoms suggestive of malignancy in their child, as well as the efficiency of referral, the quality of investigation and treatment, and the resourcing and organisation of the health service. In many high-income countries, the incidence of ALL is rising by an average of about 1% every year.7, 18 Whether this is due to a true increase, improved registration, or both, is still debated.29 It is unlikely that improved diagnosis of the less aggressive forms of leukaemia can explain the very widespread rises in survival that we report. In many countries, childhood cancer treatment is provided in specialised centres, which makes cancer registration for children easier than for adults, although active follow-up can be more challenging in children and adolescents. Population-based cancer registry data might still be restricted by underdiagnosis and under-registration of cancer patients, both of which are difficult to quantify. This might be a more important issue for participating registries in the 17 LMICs, but it is nevertheless important to capture the available information as a guide to policy.

cer registry data might still be restricted by underdiagnosis and under-registration of cancer patients, both of which are difficult to quantify. This might be a more important issue for participating registries in the 17 LMICs, but it is nevertheless important to capture the available information as a guide to policy. There is still room for improvement in the management of childhood leukaemia in many countries. In some LMICs, even the most basic treatment for leukaemia,28, 30 or pain relief, was still not consistently available until recently.31 Abandonment of treatment is also a major issue in some settings.32, 33 5-year survival for children with precursor-cell lymphoblastic leukaemia can be as high as 90%, and up to 80% for children with AML, but in some countries, survival remains below 60% for both types of leukaemia. Interventions that have been proven to improve outcomes in childhood malignancy include enrolment in clinical trials, international collaboration, and treatment guidelines.5, 34, 35 Wider implementation of these initiatives, together with mobilisation of additional resources, especially in poorer countries, would be likely to improve the delivery of effective treatments, and to reduce worldwide inequalities in survival.8, 9, 28, 36, 37 Supplementary Material Supplementary appendix

There is still room for improvement in the management of childhood leukaemia in many countries. In some LMICs, even the most basic treatment for leukaemia,28, 30 or pain relief, was still not consistently available until recently.31 Abandonment of treatment is also a major issue in some settings.32, 33 5-year survival for children with precursor-cell lymphoblastic leukaemia can be as high as 90%, and up to 80% for children with AML, but in some countries, survival remains below 60% for both types of leukaemia. Interventions that have been proven to improve outcomes in childhood malignancy include enrolment in clinical trials, international collaboration, and treatment guidelines.5, 34, 35 Wider implementation of these initiatives, together with mobilisation of additional resources, especially in poorer countries, would be likely to improve the delivery of effective treatments, and to reduce worldwide inequalities in survival.8, 9, 28, 36, 37 Supplementary Material Supplementary appendix Acknowledgments This work was supported by the Canadian Partnership Against Cancer, Cancer Focus Northern Ireland, Cancer Institute New South Wales, Cancer Research UK (C1336/A16148), US Centers for Disease Control and Prevention (CDC: 12FED03123, ACO12036), Swiss Re, Swiss Cancer Research Foundation, Swiss Cancer League, and the University of Kentucky (3049024672-12-568).

Partnership Against Cancer, Cancer Focus Northern Ireland, Cancer Institute New South Wales, Cancer Research UK (C1336/A16148), US Centers for Disease Control and Prevention (CDC: 12FED03123, ACO12036), Swiss Re, Swiss Cancer Research Foundation, Swiss Cancer League, and the University of Kentucky (3049024672-12-568). Contributors AB, CA, and MPC designed the analyses for this study. CAS, JC, DCS, RM-G, RP-B, CEK, and MFGM contributed to data acquisition. AB, RH, HC, DS, MPC, and CA had access to all raw data. AB, RH, HC, DS, MPC, and CA contributed to the data preparation, quality control and analyses, and checked the results. AB drafted the initial report. All authors contributed to the data interpretation, critically revised the manuscript, and approved the version to be published. All members of the CONCORD Working Group had access to the results and contributed to interpretation of the findings. Declaration of interests We declare no competing interests. Figure 1 Distribution (%) of leukaemia subtypes in children diagnosed during 1995–2009 and included in survival analyses, by continent Leukaemias were classified according to the third edition of the International Classification of Childhood Cancer. Figure 2 Age-standardised 5-year net survival (%) for children diagnosed with acute lymphoblastic leukaemia (ALL) and acute myeloid leukaemia (AML) during 1995–2009

Figure 1 Distribution (%) of leukaemia subtypes in children diagnosed during 1995–2009 and included in survival analyses, by continent Leukaemias were classified according to the third edition of the International Classification of Childhood Cancer. Figure 2 Age-standardised 5-year net survival (%) for children diagnosed with acute lymphoblastic leukaemia (ALL) and acute myeloid leukaemia (AML) during 1995–2009 The number of countries for which survival estimates are shown in each box-plot is given in parentheses. Box-plots in light blue are for ALL (group Ia1 according to the third edition of the International Classification of Childhood Cancer [ICCC-3]), and dark blue for AML (ICCC-3 group Ib). The vertical line inside each box denotes the median survival value, and the box shows the IQR between the lower and upper quartiles. The extreme limits of the box-plot are 1·5 times the IQR below the lower quartile and above the upper quartile. Open circles indicate outlier values, outside this range Survival estimates for African countries are not shown because they were either not standardised or less reliable. Figure 3 Trends in age-standardised 5-year net survival (%) for children diagnosed with acute lymphoblastic leukaemia (ALL) and acute myeloid leukaemia (AML), during 1995–1999, 2000–2004, and 2005–2009

The number of countries for which survival estimates are shown in each box-plot is given in parentheses. Box-plots in light blue are for ALL (group Ia1 according to the third edition of the International Classification of Childhood Cancer [ICCC-3]), and dark blue for AML (ICCC-3 group Ib). The vertical line inside each box denotes the median survival value, and the box shows the IQR between the lower and upper quartiles. The extreme limits of the box-plot are 1·5 times the IQR below the lower quartile and above the upper quartile. Open circles indicate outlier values, outside this range Survival estimates for African countries are not shown because they were either not standardised or less reliable. Figure 3 Trends in age-standardised 5-year net survival (%) for children diagnosed with acute lymphoblastic leukaemia (ALL) and acute myeloid leukaemia (AML), during 1995–1999, 2000–2004, and 2005–2009 Countries have been grouped into ten geographical regions. Survival estimates for African countries are not shown because they were either not standardised or less reliable. ALL: group Ia1 according to the third edition of the International Classification of Childhood Cancer (ICCC-3). AML: ICCC-3 group Ib. ARG=Argentina. AUS=Australia. AUT=Austria. BEL=Belgium. BGR=Bulgaria. BLR=Belarus. BRA=Brazil. CAN=Canada. CHE=Switzerland. CHN=China. COL=Colombia. CYP=Cyprus. DEU=Germany. DNK=Denmark. ECU=Ecuador. ESP=Spain. EST=Estonia. FIN=Finland. FRA=France. GBR=United Kingdom. HRV=Croatia. IRL=Ireland. ISR=Israel. ITA=Italy. JPN=Japan. KOR=Republic of Korea. LTU=Lithuania. LVA=Latvia. MYS=Malaysia. NLD=Netherlands. NOR=Norway. NZL=New Zealand. POL=Poland. PRI=Puerto Rico. PRT=Portugal. SVK=Slovakia. SVN=Slovenia. SWE=Sweden. TWN=Taiwan. THA=Thailand. TUR=Turkey. USA=United States of America.

HRV=Croatia. IRL=Ireland. ISR=Israel. ITA=Italy. JPN=Japan. KOR=Republic of Korea. LTU=Lithuania. LVA=Latvia. MYS=Malaysia. NLD=Netherlands. NOR=Norway. NZL=New Zealand. POL=Poland. PRI=Puerto Rico. PRT=Portugal. SVK=Slovakia. SVN=Slovenia. SWE=Sweden. TWN=Taiwan. THA=Thailand. TUR=Turkey. USA=United States of America. Figure 4 Change (absolute difference, %) in age-standardised 5-year net survival for acute lymphoblastic leukaemia (ALL) and acute myeloid leukaemia (AML), between (A) 1995–99 and 2000–04 and (B) between 2000–04 and 2005–09 Each datapoint represents one of the participating countries. Datapoints above the diagonal indicate that survival from AML increased more than survival from ALL between the two calendar periods. Countries are represented only if 5-year age-standardised estimates were available for ALL and for AML in successive calendar periods. ALL: group Ia1 according to the third edition of the International Classification of Childhood Cancer (ICCC-3). AML: ICCC-3 group Ib. AUS=Australia. AUT=Austria. BGR=Bulgaria. CAN=Canada. CHE=Switzerland. CHN=China. DEU=Germany. ESP=Spain. FRA=France. GBR=United Kingdom. ISR=Israel. ITA=Italy. JPN=Japan. KOR=Republic of Korea. NLD=Netherlands. NOR=Norway. NZL=New Zealand. PRT=Portugal. SWE=Sweden. THA=Thailand. TWN=Taiwan. USA=United States of America. Table 1 Data quality indicators for children aged 0–14 years diagnosed with leukaemia between 1995–2009: by continent and country

Each datapoint represents one of the participating countries. Datapoints above the diagonal indicate that survival from AML increased more than survival from ALL between the two calendar periods. Countries are represented only if 5-year age-standardised estimates were available for ALL and for AML in successive calendar periods. ALL: group Ia1 according to the third edition of the International Classification of Childhood Cancer (ICCC-3). AML: ICCC-3 group Ib. AUS=Australia. AUT=Austria. BGR=Bulgaria. CAN=Canada. CHE=Switzerland. CHN=China. DEU=Germany. ESP=Spain. FRA=France. GBR=United Kingdom. ISR=Israel. ITA=Italy. JPN=Japan. KOR=Republic of Korea. NLD=Netherlands. NOR=Norway. NZL=New Zealand. PRT=Portugal. SWE=Sweden. THA=Thailand. TWN=Taiwan. USA=United States of America. Table 1 Data quality indicators for children aged 0–14 years diagnosed with leukaemia between 1995–2009: by continent and country Calendar period Registries (n) Patients submitted (n) Eligible patients (n [%]) DCO (n [%]) Any haematological malignancy (n [% of eligible]) Leukaemia Included in analyses (n [% of any haematological malignancy]) Microscopically verified (n [%]) Follow-up time (years) of alive patients* (median [IQR]) Lymphoid leukaemia (n [%]) Acute myeloid leukaemia (n [%]) Unspecified and other specified leukaemia (n [%]) Africa ·· ·· 373 223 (60%) 1 (<1%) 193 (87%) 153 (79%) 152 (99%) 3·5 (0·5–4·7) 111 (73%) 33 (22%) 9 (6%) Algerian registries 2000–09 2 131 128 (98%) 0 99 (77%) 62 (63%) 61 (98%) 0·7 (0·2–3·0) 37 (60%) 20 (32%) 5 (8%) Lesotho† 1995–2009 1 27 27 (100%) 0 27 (100%) 27 (100%) 27 (100%) 6·6 (0·9–8·4) 25 (93%) 0 2 (7%) Libya (Benghazi) 2003–04 1 23 23 (100%) 1 (4%) 22 (96%) 21 (95%) 21 (100%) 4·9 (0·0–5·2) 14 (67%) 5 (24%) 2 (10%) Tunisia (Central) 1996–2007 1 192 45 (23%) 0 45 (100%) 43 (96%) 43 (100%) 1·7 (1·0–2·2) 35 (81%) 8 (19%) 0 America (Central and South) ·· ·· 5731 5722 (>99%) 175 (3%) 5488 (96%) 5086 (93%) 5063 (>99%) 9·2 (7·3–11·5) 4788 (94%) 195 (4%) 103 (2%) Argentina† 2000–09 1 3671 3671 (100%) 128 (3%) 3496 (95%) 3496 (100%) 3496 (100%) 6·5 (4·9–8·3) 3496 (100%) 0 0 Brazilian registries 1996–2009 5 687 678 (99%) 41 (6%) 626 (92%) 497 (79%) 475 (96%) 11·2 (9·8–13·1) 387 (78%) 79 (16%) 31 (6%) Chilean registries 1998–2008 2 116 116 (100%) 0 116 (100%) 96 (83%) 96 (100%) 9·2 (7·7–12·4) 66 (69%) 14 (15%) 16 (17%) Colombia (Cali) 1995–2009 1 383 383 (100%) 0 382 (>99%) 381 (>99%) 381 (100%) 6·8 (3·6–9·2) 381 (100%) 0 0 Ecuador (Quito) 1995–2009 1 507 507 (100%) 4 (1%) 503 (99%) 375 (75%) 375 (100%) 13·8 (11·5–17·3) 295 (79%) 67 (18%) 13 (3%) Puerto Rico† 2000–09 1 367 367 (100%) 2 (1%) 365 (99%) 241 (66%) 240 (>99%) 7·4 (6·2–8·6) 163 (68%) 35 (15%) 43 (18%) America (North) ·· ·· 51 195 51 110 (>99%) 144 (<1%) 50 920 (>99%) 36 970 (73%) 36 156 (98%) 10·7 (8·3–13·3) 30 088 (81%) 5894 (16%) 988 (3%) Canada† 1995–2009 13 5259 5200 (99%) 13 (<1%) 5186 (>99%) 4152 (80%) 3983 (96%) 11·0 (8·6–13·5) 3363 (81%) 623 (15%) 166 (4%) US registries 1995–2009 38 45 936 45 910 (>99%) 131 (<1%) 45 734 (>99%) 32 818 (72%) 32 173 (98%) 10·3 (8·0–13·1) 26 725 (81%) 5271 (16

156 (98%) 10·7 (8·3–13·3) 30 088 (81%) 5894 (16%) 988 (3%) Canada† 1995–2009 13 5259 5200 (99%) 13 (<1%) 5186 (>99%) 4152 (80%) 3983 (96%) 11·0 (8·6–13·5) 3363 (81%) 623 (15%) 166 (4%) US registries 1995–2009 38 45 936 45 910 (>99%) 131 (<1%) 45 734 (>99%) 32 818 (72%) 32 173 (98%) 10·3 (8·0–13·1) 26 725 (81%) 5271 (16 %) 822 (3%) Asia ·· ·· 17 371 17 176 (99%) 74 (<1%) 17 094 (>99%) 12 552 (73%) 12 309 (98%) 9·6 (7·4–11·8) 8553 (68%) 3051 (24%) 948 (8%) Chinese registries 1995–2009 17 741 728 (98%) 0 728 (100%) 632 (87%) 619 (98%) 7·1 (6·3–8·2) 309 (49%) 149 (24%) 174 (28%) Cyprus† 2004–09 1 49 49 (100%) 2 (4%) 47 (96%) 46 (98%) 46 (100%) 5·4 (5·1–5·8) 35 (76%) 10 (22%) 1 (2%) India (Karunagappally) 1995–2009 1 54 54 (100%) 0 54 (100%) 53 (98%) 53 (100%) 11·2 (7·4–14·8) 43 (81%) 8 (15%) 2 (4%) Indonesia (Jakarta) 2005–07 1 29 29 (100%) 0 28 (97%) 28 (100%) 26 (93%) ·· 14 (50%) 6 (21%) 8 (29%) Israel† 1995–2009 1 1880 1775 (94%) 13 (1%) 1759 (99%) 1002 (57%) 997 (>99%) 10·3 (8·1–12·8) 731 (73%) 185 (18%) 86 (9%) Japanese registries 1995–2009 9 1729 1727 (>99%) 46 (3%) 1681 (97%) 1441 (86%) 1407 (98%) 10·1 (6·2–11·3) 969 (67%) 414 (29%) 58 (4%) Korea†‡ 1995–2009 1 7569 7507 (99%) 0 7507 (100%) 5422 (72%) 5279 (97%) 11·5 (9·2–14·1) 3583 (66%) 1419 (26%) 420 (8%) Malaysia (Penang) 1995–2009 1 274 269 (98%) 3 (1%) 263 (98%) 261 (99%) 260 (>99%) 9·4 (7·6–11·9) 163 (62%) 67 (26%) 31 (12%) Mongolia† 2005–09 1 47 42 (89%) 0 41 (98%) 41 (100%) 31 (76%) ·· 25 (61%) 8 (20%) 8 (20%) Taiwan† 1995–2009 1 3642 3642 (100%) 0 3642 (100%) 2641 (73%) 2612 (99%) 11·5 (9·2–14·0) 1935 (73%) 614 (23%) 92 (3%) Thai registries 1995–2009 3 537 537 (100%) 6 (1%) 531 (99%) 469 (88%) 463 (99%) 10·8 (7·8–12·8) 327 (70%) 85 (18%) 57 (12%) Turkey (Izmir) 1995–2009 1 820 817 (>99%) 4 (<1%) 813 (>99%) 516 (63%) 516 (100%) 9·2 (7·3–12·3) 419 (81%) 86 (17%) 11 (2%) Europe ·· ·· 45 127 44 003 (98%) 165 (<1%) 43 815 (>99%) 31 797 (73%) 31 256 (98%) 11·0 (8·9–13·2) 26 262 (83%) 4856 (15%) 679 (2%) Austria† 1995–2009 2 1248 1209 (97%) 0 1203 (>99%) 826 (69%) 823 (>99%) 12·2 (9·5–14·4) 681 (82%) 122 (15%) 23 (3%) Belarus† 1995–2009 1 745 745 (100%) 0 745 (100%) 745 (100%) 744 (>99%) 12·6 (10·0–15·0) 745 (100%) 0 0 Belgium† 2004–09 1 747 747 (100%) 0 747 (100%) 466 (62%) 465 (>99%) 8·0 (7·8–8·3) 396 (85%) 68 (15%) 2 (<1%) Bulgaria† 1995–2009 1 1079 1079 (100%) 0 1079 (100%) 714 (66%) 678 (95%) 13·0 (9·9–15·5) 558 (78%) 96 (13%) 60 (8%) Croatia† 1998–2009 1 651 651 (100%) 0 651 (100%) 446 (69%) 446 (100%) 9·9 (8·8–11·8

(100%) 0 0 Belgium† 2004–09 1 747 747 (100%) 0 747 (100%) 466 (62%) 465 (>99%) 8·0 (7·8–8·3) 396 (85%) 68 (15%) 2 (<1%) Bulgaria† 1995–2009 1 1079 1079 (100%) 0 1079 (100%) 714 (66%) 678 (95%) 13·0 (9·9–15·5) 558 (78%) 96 (13%) 60 (8%) Croatia† 1998–2009 1 651 651 (100%) 0 651 (100%) 446 (69%) 446 (100%) 9·9 (8·8–11·8 ) 375 (84%) 61 (14%) 10 (2%) Denmark† 1995–2009 1 695 695 (100%) 0 695 (100%) 680 (98%) 660 (97%) 10·4 (8·3–12·9) 551 (81%) 112 (16%) 17 (3%) Estonia† 1995–2008 1 160 160 (100%) 1 (1%) 159 (99%) 99 (62%) 99 (100%) 11·9 (9·1–14·4) 73 (74%) 24 (24%) 2 (2%) Finland† 1995–2009 1 1007 1007 (100%) 6 (1%) 1001 (99%) 728 (73%) 726 (>99%) 11·0 (8·5–13·5) 593 (81%) 101 (14%) 34 (5%) France† 1995–2009 1 10 619 10 032 (95%) 0 10 030 (>99%) 6875 (69%) 6873 (>99%) 9·4 (7·2–12·5) 5611 (82%) 1109 (16%) 155 (2%) German registries 1995–2009 13 3801 3781 (99%) 96 (3%) 3684 (97%) 2661 (72%) 2460 (92%) 10·7 (8·4–13·4) 2162 (81%) 436 (16%) 63 (2%) Iceland† 1995–2009 1 36 36 (100%) 0 36 (100%) 36 (100%) 36 (100%) 10·9 (10·1–13·0) 29 (81%) 6 (17%) 1 (3%) Ireland† 1995–2009 1 817 797 (98%) 2 (<1%) 794 (>99%) 571 (72%) 570 (>99%) 10·6 (8·2–13·2) 458 (80%) 99 (17%) 14 (2%) Italian registries 1995–2009 32 3955 3602 (91%) 10 (<1%) 3584 (>99%) 2313 (65%) 2247 (97%) 9·7 (7·3–12·6) 1930 (83%) 336 (15%) 47 (2%) Latvia† 1995–2009 1 205 205 (100%) 0 204 (>99%) 196 (96%) 194 (99%) 13·4 (11·5–16·6) 97 (49%) 35 (18%) 64 (33%) Lithuania† 1995–2009 1 532 529 (99%) 2 (<1%) 527 (>99%) 358 (68%) 358 (100%) 10·9 (9·1–13·6) 301 (84%) 53 (15%) 4 (1%) Malta† 1995–2009 1 92 92 (100%) 0 92 (100%) 61 (66%) 60 (98%) 10·8 (7·9–13·2) 53 (87%) 8 (13%) 0 Netherlands† 1995–2009 1 2860 2858 (>99%) 3 (<1%) 2855 (>99%) 1988 (70%) 1988 (100%) 12·4 (10·1–15·0) 1652 (83%) 307 (15%) 29 (1%) Norway† 1995–2009 1 665 665 (100%) 1 (<1%) 664 (>99%) 646 (97%) 646 (100·0) 10·8 (7·9–13·3) 531 (82%) 108 (17%) 7 (1%) Poland (Wroclaw) 1995–2009 1 205 205 (100%) 0 203 (99%) 120 (59%) 120 (100%) 10·2 (9·2–11·4) 96 (80%) 24 (20%) 0 Portugal† 1998–2009 4 872 855 (98%) 0 854 (>99%) 665 (78%) 565 (85%) 9·3 (8·2–10·9) 511 (77%) 134 (20%) 20 (3%) Russia (Arkhangelsk) ‡ 2000–09 1 60 60 (100%) 1 (2%) 59 (98%) 59 (100%) 59 (100%) 10·7 (9·0–11·8) 51 (86%) 0 8 (14%) Slovakia† 2000–07 1 426 423 (99%) 6 (1%) 417 (99%) 284 (68%) 284 (100%) 9·7 (8·6–10·9) 223 (79%) 59 (21%) 2 (1%) Slovenia† 1995–2009 1 280 280 (100%) 2 (1%) 278 (99%) 181 (65%) 181 (100%) 13·1 (10·9–15·2) 147 (81%) 33 (18%) 1 (1%) Spanish registries 1995–2009 11 1487 1479 (99%) 9 (1%)

0·7 (9·0–11·8) 51 (86%) 0 8 (14%) Slovakia† 2000–07 1 426 423 (99%) 6 (1%) 417 (99%) 284 (68%) 284 (100%) 9·7 (8·6–10·9) 223 (79%) 59 (21%) 2 (1%) Slovenia† 1995–2009 1 280 280 (100%) 2 (1%) 278 (99%) 181 (65%) 181 (100%) 13·1 (10·9–15·2) 147 (81%) 33 (18%) 1 (1%) Spanish registries 1995–2009 11 1487 1479 (99%) 9 (1%) 1470 (99%) 1152 (78%) 1136 (99%) 10·9 (8·6–13·3) 997 (87%) 123 (11%) 32 (3%) Sweden† 1995–2009 1 1162 1162 (100%) 0 1162 (100%) 1145 (99%) 1145 (100%) 11·1 (8·2–13·6) 984 (86%) 154 (13%) 7 (1%) Switzerland† 1995–2009 1 853 852 (>99%) 1 (<1%) 851 (>99%) 813 (96%) 812 (>99%) 11·2 (9·2–13·2) 684 (84%) 119 (15%) 10 (1%) UK† 1995–2009 3 9868 9797 (99%) 25 (<1%) 9771 (>99%) 6969 (71%) 6881 (99%) 12·5 (10·0–14·9) 5773 (83%) 1129 (16%) 67 (1%) Oceania ·· ·· 4218 4158 (99%) 6 (<1%) 4151 (>99%) 3270 (79%) 3143 (96%) 11·2 (9·1–13·9) 2672 (82%) 544 (17%) 54 (2%) Australian registries 1995–2009 6 3537 3477 (98%) 2 (<1%) 3474 (>99%) 2607 (75%) 2480 (95%) 9·7 (7·5–12·4) 2145 (82%) 424 (16%) 38 (1%) New Zealand† 1995–2009 1 681 681 (100%) 4 (1%) 677 (99%) 663 (98%) 663 (100%) 12·8 (10·7–15·4) 527 (79%) 120 (18%) 16 (2%) Total ·· 198 124 015 122 392 (99%) 565 (<1%) 121 661 (99%) 89 828 (74%) 88 079 (98%) ·· 72 474 (81%) 14 573 (16%) 2781 (3%) Calendar period shows the maximum time span covered by the data. DCO=patients diagnosed from autopsy or Death Certificate Only (% of eligible patients). Microscopic verification=cytology or histology, or morphological verification unspecified (% of patients included in analyses). * Median follow-up of patients diagnosed during 1995–2004, and IQR.

1470 (99%) 1152 (78%) 1136 (99%) 10·9 (8·6–13·3) 997 (87%) 123 (11%) 32 (3%) Sweden† 1995–2009 1 1162 1162 (100%) 0 1162 (100%) 1145 (99%) 1145 (100%) 11·1 (8·2–13·6) 984 (86%) 154 (13%) 7 (1%) Switzerland† 1995–2009 1 853 852 (>99%) 1 (<1%) 851 (>99%) 813 (96%) 812 (>99%) 11·2 (9·2–13·2) 684 (84%) 119 (15%) 10 (1%) UK† 1995–2009 3 9868 9797 (99%) 25 (<1%) 9771 (>99%) 6969 (71%) 6881 (99%) 12·5 (10·0–14·9) 5773 (83%) 1129 (16%) 67 (1%) Oceania ·· ·· 4218 4158 (99%) 6 (<1%) 4151 (>99%) 3270 (79%) 3143 (96%) 11·2 (9·1–13·9) 2672 (82%) 544 (17%) 54 (2%) Australian registries 1995–2009 6 3537 3477 (98%) 2 (<1%) 3474 (>99%) 2607 (75%) 2480 (95%) 9·7 (7·5–12·4) 2145 (82%) 424 (16%) 38 (1%) New Zealand† 1995–2009 1 681 681 (100%) 4 (1%) 677 (99%) 663 (98%) 663 (100%) 12·8 (10·7–15·4) 527 (79%) 120 (18%) 16 (2%) Total ·· 198 124 015 122 392 (99%) 565 (<1%) 121 661 (99%) 89 828 (74%) 88 079 (98%) ·· 72 474 (81%) 14 573 (16%) 2781 (3%) Calendar period shows the maximum time span covered by the data. DCO=patients diagnosed from autopsy or Death Certificate Only (% of eligible patients). Microscopic verification=cytology or histology, or morphological verification unspecified (% of patients included in analyses). * Median follow-up of patients diagnosed during 1995–2004, and IQR. † National coverage—the data are derived from a population-based cancer registry (registries) covering the entire country. ‡ Korea: Republic of Korea; Russia: Russian Federation. Table 2 5-year age-standardised net survival in children aged 0–14 years diagnosed with leukaemia

* Median follow-up of patients diagnosed during 1995–2004, and IQR. † National coverage—the data are derived from a population-based cancer registry (registries) covering the entire country. ‡ Korea: Republic of Korea; Russia: Russian Federation. Table 2 5-year age-standardised net survival in children aged 0–14 years diagnosed with leukaemia Lymphoid leukaemia Acute myeloid leukaemia (ICCC-3 group Ib) Unspecified & other leukaemias (ICCC-3 group Ie) All lymphoid (Ia) Precursor cell (Ia1) Mature B cell (Ia2) n Net survival (%), 95% CI n Net survival (%), 95% CI n Net survival (%), 95% CI n Net survival (%), 95% CI n Net survival (%), 95% CI Africa Algerian registries 2000–04 19 21·6%*(0·0–45·6) 19 21·6%*(0·0–45·6) .. .. 13 23·9%*(0·0–48·5) 3 .. 2005–09 18 .. 17 .. .. .. 7 .. 2 .. Lesotho† 1995–2009 25 43·0%‡(20·9–65·1) 22 39·5%‡(16·3–62·7) 3 .. .. .. 2 .. Libya (Benghazi) 2003–04 14 70·2% (43·4–96·9) 14 70·2% (43·4–96·9) .. .. 5 .. 2 .. Tunisia (Central) 1996–99 20 46·1%*(13·2–79·1) 20 46·1%*(13·2–79·1) .. .. 3 .. .. .. 2000–04 6 .. 6 .. .. .. 2 .. .. .. 2005–07 9 .. 9 .. .. .. 3 .. .. .. America (Central and South) Argentina† 2000–04 1785 64·6% (62·3–67·0) 1785 64·6% (62·3–67·0) .. .. .. .. .. .. 2005–09 1711 66·9% (64·4–69·3) 1711 66·9% (64·4–69·3) .. .. .. .. .. .. Brazilian registries 1996–99 67 72·9% (61·8–84·0) 64 71·8% (60·0–83·6) .. .. 13 38·6% (13·8–63·5) 4 .. 2000–04 172 67·9% (60·5–75·3) 168 67·1% (59·6–74·7) .. .. 41 46·1% (32·1–60·1) 16 75·5% (55·0–96·0) 2005–09 148 66·2% (58·5–73·9) 132 66·4% (58·4–74·3) 10 80·3%‡(48·7–100·0) 25 53·1% (36·9–69·4) 11 55·8% (31·7–79·9) Chilean registries 1998–99 17 41·2% (19·1–63·3) 17 41·2% (19·1–63·3) .. .. 3 .. .. .. 2000–04 21 71·5% (52·7–90·3) 21 71·5% (52·7–90·3) .. .. 2 .. 12 100·0% 2005–08 28 83·9%§(73·3–94·5) 23 77·5% (60·5–94·6) .. .. 9 .. 4 .. Colombia (Cali) 1995–99 125 40·7% (31·6–49·9) 124 40·4% (31·2–49·6) .. .. .. .. .. .. 2000–04 137 48·4% (39·4–57·4) 136 48·8% (39·7–57·8) .. .. .. .. .. .. 2005–09 119 52·4% (42·8–61·9) 117 52·4% (42·8–61·9) .. .. .. .. .. .. Ecuador (Quito) 1995–99 85 64·3% (54·2–74·4) 81 63·7% (53·3–74·1) .. .. 33 49·0%§(35·9–62·1) 5 .. 2000–04 112 63·5% (54·2–72·8) 110 64·1% (54·7–73·6) .. .. 18 50·2% (28·2–72·1) 6 .. 2005–09 98 62·5% (53·5–71·5) 95 63·1% (54·0–72·1) .. .. 16 59·3% (36·8–81·7) 2 .. Puerto Rico† 2000–04 79 79·4% (70·5–88·2) 73 79·0% (69·7–88·3) 3 .. 17 53·0% (30·3–75·8) 34 72·4%§(60·8–83·9) 2005–09 84 78·7% (69·5–87·9) 81 78·9% (69·4–88·4) 3 .. 18 48·1% (26·4–69·8) 9 ..

·1% (54·7–73·6) .. .. 18 50·2% (28·2–72·1) 6 .. 2005–09 98 62·5% (53·5–71·5) 95 63·1% (54·0–72·1) .. .. 16 59·3% (36·8–81·7) 2 .. Puerto Rico† 2000–04 79 79·4% (70·5–88·2) 73 79·0% (69·7–88·3) 3 .. 17 53·0% (30·3–75·8) 34 72·4%§(60·8–83·9) 2005–09 84 78·7% (69·5–87·9) 81 78·9% (69·4–88·4) 3 .. 18 48·1% (26·4–69·8) 9 .. America (North) Canada† 1995–99 1156 86·0% (83·6–88·4) 1134 85·9% (83·4–88·3) 11 81·9% (60·2–100·0) 235 56·1% (49·5–62·7) 53 68·0% (55·2–80·8) 2000–04 1092 91·0% (89·0–93·0) 1074 91·0% (89·0–93·0) 10 90·0% (72·4–100·0) 188 64·0% (57·0–70·9) 51 72·7% (60·6–84·9) 2005–09 1115 90·5% (88·4–92·6) 1097 90·7% (88·6–92·9) 14 80·0% (60·5–99·6) 200 71·8% (65·1–78·5) 62 83·1% (72·9–93·3) US registries 1995–99 7801 82·9% (81·9–83·9) 7670 82·9% (81·9–83·9) 73 77·0% (67·7–86·4) 1574 51·5% (48·9–54·1) 241 68·8% (62·4–75·2) 2000–04 9025 86·5% (85·7–87·4) 8842 86·6% (85·8–87·5) 128 86·5% (80·1–92·9) 1799 59·7% (57·3–62·1) 285 63·8% (57·7–69·8) 2005–09 9899 87·7% (86·9–88·5) 9735 87·7% (86·9–88·5) 132 88·7% (83·0–94·5) 1898 63·3% (60·9–65·6) 296 69·8% (64·2–75·4) Asia Chinese registries 1995–99 29 10·6%§(3·1–18·2) 28 11·1%§(3·3–19·0) .. .. 27 4·2%§(0·0–8·6) 23 8·7% (0·0–18·9) 2000–04 98 45·6% (36·3–54·9) 84 48·8% (38·8–58·8) .. .. 61 20·9% (10·1–31·7) 69 15·0% (6·7–23·2) 2005–09 182 69·2% (61·6–76·8) 151 69·4% (61·1–77·8) 10 68·7%‡(40·8–96·6) 61 41·1% (27·8–54·4) 82 26·0% (15·9–36·0) Cyprus† 2004–09 35 83·0%‡ (70·9–95·0) 34 80·8%‡ (68·9–92·6) 1 .. 10 60·1%‡(32·2–88·0) 1 .. India (Karunagappally) 1995–99 17 59·3% (36·7–81·9) 17 59·3% (36·7–81·9) .. .. 1 .. .. .. 2000–04 14 57·4% (32·6–82·1) 14 57·4% (32·6–82·1) .. .. 4 .. .. .. 2005–09 12 80·2% (60·3–100·0) 12 80·2% (60·3–100·0) .. .. 3 .. 2 .. Indonesia (Jakarta) 2005–07 14 44·3% (13·4–75·3) 14 44·3% (13·4–75·3) .. .. 6 .. 8 .. Israel† 1995–99 192 81·4% (74·9–87·8) 188 81·0% (74·5–87·6) 2 .. 40 63·2%§(51·6–74·9) 27 74·1% (57·9–90·3) 2000–04 264 86·3% (81·6–91·1) 258 86·3% (81·5–91·1) 4 .. 67 67·5% (56·6–78·4) 27 77·8% (62·5–93·2) 2005–09 275 84·4% (79·6–89·1) 271 84·5% (79·7–89·3) 1 .. 78 66·2% (56·2–76·3) 32 89·2%§(80·7–97·7) Japanese registries 1995–99 294 78·0% (72·9–83·1) 294 78·0% (72·9–83·1) .. .. 122 62·4% (53·7–71·0) 25 60·1% (41·4–78·8) 2000–04 409 78·0% (73·2–82·9) 406 77·9% (73·1–82·8) 2 .. 181 61·2% (53·5–68·9) 24 87·5% (74·6–100·0) 2005–09 266 82·5% (78·0–86·9) 259 82·2% (77·6–86·8) 4 .. 111 69·4% (62·0–76·9) 9 .. Korea†;¶ 1995–99 1191 63·1% (60·2–66·0) 1171 63·2% (60·3–66·1) .. ..

78·0% (72·9–83·1) .. .. 122 62·4% (53·7–71·0) 25 60·1% (41·4–78·8) 2000–04 409 78·0% (73·2–82·9) 406 77·9% (73·1–82·8) 2 .. 181 61·2% (53·5–68·9) 24 87·5% (74·6–100·0) 2005–09 266 82·5% (78·0–86·9) 259 82·2% (77·6–86·8) 4 .. 111 69·4% (62·0–76·9) 9 .. Korea†;¶ 1995–99 1191 63·1% (60·2–66·0) 1171 63·2% (60·3–66·1) .. .. 482 40·1% (35·6–44·5) 167 44·2% (36·8–51·6) 2000–04 1221 73·0% (70·3–75·6) 1197 73·2% (70·5–75·8) .. .. 504 49·7% (45·4–54·1) 128 55·4% (46·9–63·9) 2005–09 1171 76·4% (73·9–78·9) 1137 77·1% (74·5–79·6) 43 65·2%‡ (52·9–77·5) 433 53·9% (49·3–58·6) 125 63·9% (55·7–72·1) Malaysia (Penang) 1995–99 51 77·3% (65·5–89·1) 51 77·3% (65·5–89·1) .. .. 22 68·3% (49·3–87·2) 1 .. 2000–04 53 82·8%§(74·2–91·3) 51 81·3%§(72·1–90·4) .. .. 23 74·0% (56·5–91·5) 5 .. 2005–09 59 70·1% (58·4–81·9) 57 69·5% (57·4–81·5) 1 .. 22 49·1% (27·8–70·5) 25 76·4% (58·5–94·4) Mongolia† 2005–09 25 18·7% (0·0–40·8) 24 19·5% (0·0–42·5) .. .. 8 .. 8 .. Taiwan† 1995–99 630 62·8% (58·8–66·9) 601 62·5% (58·4–66·6) 18 72·3% (52·3–92·4) 195 40·8% (33·8–47·8) 52 50·8% (38·2–63·3) 2000–04 682 72·2% (68·5–76·0) 665 72·0% (68·2–75·8) 15 80·0% (60·5–99·6) 214 48·6% (41·8–55·5) 22 77·3% (60·3–94·4) 2005–09 623 77·6% (74·1–81·2) 610 78·1% (74·6–81·7) 11 62·1% (39·2–85·0) 205 55·6% (48·8–62·3) 18 63·9% (43·2–84·7) Thai registries 1995–99 102 51·1% (39·4–62·8) 102 51·1% (39·4–62·8) .. .. 18 21·8% (2·1–41·5) 20 52·0% (28·7–75·4) 2000–04 120 58·9%(49·2–68·5) 120 58·9% (49·2–68·5) .. .. 31 30·5%§(18·0–43·0) 20 37·3% (16·6–58·1) 2005–09 105 55·0% (45·4–64·7) 102 55·6% (46·0–65·3) .. .. 36 44·7%§(31·8–57·5) 17 57·0% (31·5–82·5) Turkey (Izmir) 1995–99 118 62·8% (53·0–72·7) 116 63·5% (53·7–73·4) 2 .. 34 59·2%§(46·1–72·4) 2 .. 2000–04 135 71·0% (62·6–79·3) 131 70·9% (62·5–79·3) 3 .. 24 31·2% (13·1–49·4) 1 .. 2005–09 166 73·6% (66·3–81·0) 161 74·2% (66·8–81·7) 4 .. 28 50·6%§(35·7–65·4) 8 .. Europe Austria† 1995–99 242 86·8% (81·6–92·0) 240 86·6% (81·3–91·9) .. .. 47 60·1%§(49·1–71·1) 7 .. 2000–04 213 90·1% (85·7–94·5) 208 89·7% (85·2–94·3) 1 .. 40 65·0%§(53·4–76·5) 12 66·7% (41·4–92·1) 2005–09 226 91·1% (86·9–95·3) 221 91·4% (87·2–95·7) 3 .. 35 72·6%§(61·2–84·1) 4 .. Belarus† 1995–99 286 74·3% (69·1–79·6) 282 74·7% (69·4–79·9) .. .. .. .. .. .. 2000–04 241 77·5% (72·0–83·0) 235 78·4% (72·9–83·9) .. .. .. .. .. .. 2005–09 218 88·1% (83·4–92·7) 209 88·3% (83·6–93·0) 17 74·8%‡(54·0–95·7) .. .. .. .. Belgium† 2004–09 396 87·2%‡ (82·9–91·4) 386 86·9%‡ (82·6–91·3) 10 90·0%‡(72·4–100·0) 68 56·6%‡ (43·5–69·6) 2 ..

% (69·1–79·6) 282 74·7% (69·4–79·9) .. .. .. .. .. .. 2000–04 241 77·5% (72·0–83·0) 235 78·4% (72·9–83·9) .. .. .. .. .. .. 2005–09 218 88·1% (83·4–92·7) 209 88·3% (83·6–93·0) 17 74·8%‡(54·0–95·7) .. .. .. .. Belgium† 2004–09 396 87·2%‡ (82·9–91·4) 386 86·9%‡ (82·6–91·3) 10 90·0%‡(72·4–100·0) 68 56·6%‡ (43·5–69·6) 2 .. Bulgaria† 1995–99 184 57·0% (49·8–64·2) 166 57·6% (50·0–65·2) .. .. 28 27·8%§(15·7–39·9) 37 13·5%§(5·6–21·3) 2000–04 159 61·8% (53·9–69·6) 154 62·3% (54·3–70·3) .. .. 35 23·1%§(12·7–33·5) 17 29·4% (9·6–49·3) 2005–09 215 72·0% (65·3–78·8) 215 71·8% (64·9–78·6) .. .. 33 33·3% (18·9–47·7) 6 .. Croatia† 1998–99 51 68·7% (56·9–80·6) 44 69·8%§(59·2–80·4) .. .. 13 46·2% (20·7–71·8) 2 .. 2000–04 177 84·1% (77·6–90·6) 160 81·6% (74·2–89·0) .. .. 20 65·0% (44·8–85·3) 4 .. 2005–09 147 85·6% (79·8–91·5) 146 85·7% (79·8–91·7) 17 94·1%‡(83·3–100·0) 28 55·9% (37·6–74·2) 4 .. Denmark† 1995–99 166 85·9% (79·7–92·0) 163 85·6% (79·4–91·8) 3 .. 32 59·4% (42·9–75·9) 4 .. 2000–04 212 84·2% (78·6–89·8) 209 84·5% (78·9–90·1) 2 .. 42 69·1% (55·3–82·9) 10 70·0% (43·3–96·8) 2005–09 173 87·4% (81·9–93·0) 171 87·2% (81·5–92·9) 2 .. 38 68·9% (54·5–83·3) 3 .. Estonia† 1995–99 29 53·6%§(40·0–67·3) 29 53·6%§(40·0–67·3) .. .. 7 .. 2 .. 2000–04 31 61·0%§(47·8–74·2) 31 61·0%§(47·8–74·2) .. .. 11 36·4% (10·3–62·6) .. .. 2005–08 13 75·4% (57·0–93·7) 13 75·4% (57·0–93·7) .. .. 6 .. .. .. Finland† 1995–99 193 82·3% (76·3–88·4) 193 82·3% (76·3–88·4) .. .. 37 78·4% (65·4–91·5) 4 .. 2000–04 192 84·8% (78·1–91·4) 191 84·7% (78·0–91·4) .. .. 29 65·4%§(52·1–78·8) 12 58·4% (32·0–84·8) 2005–09 208 82·0% (75·4–88·5) 205 81·8% (75·2–88·4) 2 .. 35 69·2% (54·3–84·2) 18 66·9% (45·5–88·4) France† 1995–99 1806 82·4% (80·4–84·3) 1728 82·4% (80·4–84·4) 78 78·7% (69·2–88·2) 380 60·2% (55·1–65·3) 36 52·8% (36·8–68·8) 2000–04 1883 88·0% (86·3–89·6) 1793 88·1% (86·4–89·8) 89 86·0% (78·7–93·3) 392 62·4% (57·7–67·1) 48 59·8% (46·1–73·5) 2005–09 1922 89·2% (87·6–90·8) 1828 89·2% (87·6–90·9) 93 88·8% (82·2–95·4) 337 69·4% (64·5–74·2) 71 64·9% (53·0–76·8) German registries 1995–99 481 86·3% (83·0–89·6) 468 86·5% (83·1–89·8) .. .. 107 61·4% (52·3–70·5) 27 74·1% (57·9–90·3) 2000–04 661 87·3% (84·7–90·0) 642 87·1% (84·4–89·8) .. .. 139 71·0% (63·2–78·8) 16 81·3% (62·8–99·8) 2005–09 1020 91·6% (89·5–93·6) 989 91·6% (89·5–93·6) 39 93·7%‡ (85·7–100·0) 190 78·2% (72·0–84·3) 20 74·3% (53·0–95·5) Iceland† 1995–99 9 .. 9 .. .. .. .. .. 1 .. 2000–04 9 .. 9 .. .. .. 2 .. .. .. 2005–09 11 80·9% (58·1–100·0) 10 90·1% (72·4–100·0) 1 .. 4 ..

(84·4–89·8) .. .. 139 71·0% (63·2–78·8) 16 81·3% (62·8–99·8) 2005–09 1020 91·6% (89·5–93·6) 989 91·6% (89·5–93·6) 39 93·7%‡ (85·7–100·0) 190 78·2% (72·0–84·3) 20 74·3% (53·0–95·5) Iceland† 1995–99 9 .. 9 .. .. .. .. .. 1 .. 2000–04 9 .. 9 .. .. .. 2 .. .. .. 2005–09 11 80·9% (58·1–100·0) 10 90·1% (72·4–100·0) 1 .. 4 .. .. .. Ireland† 1995–99 149 78·9% (71·5–86·3) 146 79·0% (71·6–86·5) 1 .. 29 65·6% (48·7–82·5) 3 .. 2000–04 146 82·9% (76·6–89·2) 145 82·8% (76·5–89·2) 1 .. 35 60·0% (44·1–76·0) 7 .. 2005–09 163 84·7% (78·3–91·1) 162 84·7% (78·3–91·1) 1 .. 35 65·8% (51·0–80·7) 4 .. Italian registries 1995–99 677 83·8% (80·7–86·9) 661 83·7% (80·6–86·9) .. .. 109 60·9% (51·7–70·1) 16 75·1% (54·6–95·5) 2000–04 740 82·4% (79·1–85·7) 725 82·5% (79·1–85·8) .. .. 124 66·7% (58·5–74·9) 18 77·8% (59·2–96·5) 2005–09 513 87·9% (85·0–90·9) 507 88·0% (85·0–90·9) 24 76·7%‡(59·0–94·3) 103 68·9% (60·4–77·4) 13 77·2% (55·3–99·2) Latvia† 1995–99 16 50·2% (26·7–73·6) 15 46·8% (22·8–70·9) .. .. 15 40·1% (16·6–63·5) 43 69·6%§(58·4–80·9) 2000–04 36 91·8% (82·9–100·0) 35 91·6% (82·5–100·0) .. .. 11 45·5% (18·1–72·9) 19 63·3% (42·2–84·3) 2005–09 45 77·0%§(66·5–87·6) 45 76·5%§(65·8–87·2) .. .. 9 .. 2 .. Lithuania† 1995–99 103 59·4% (49·3–69·6) 102 59·2% (49·0–69·4) .. .. 12 41·8% (15·9–67·7) 3 .. 2000–04 112 73·7% (64·9–82·5) 106 74·3% (65·3–83·4) 2 .. 27 22·3% (7·3–37·2) .. .. 2005–09 86 68·2% (58·1–78·4) 86 69·0% (58·4–79·6) .. .. 14 44·0% (20·1–67·9) 1 .. Malta† 1995–99 19 63·3% (42·3–84·3) 19 63·3% (42·3–84·3) .. .. .. .. .. .. 2000–04 16 81·4% (62·9–99·8) 16 81·4% (62·9–99·8) .. .. 4 .. .. .. 2005–09 18 83·1% (66·0–100·0) 17 82·5% (64·9–100·0) .. .. 4 .. .. .. Netherlands† 1995–99 529 81·1% (77·2–84·9) 527 81·0% (77·1–84·9) 2 .. 92 58·6% (48·0–69·2) 12 75·1% (51·8–98·5) 2000–04 586 84·0% (80·6–87·4) 582 84·0% (80·6–87·4) 1 .. 109 53·7% (43·7–63·8) 11 81·9% (60·2–100·0) 2005–09 537 86·2% (82·9–89·6) 533 86·2% (82·9–89·6) 4 .. 106 58·6% (48·2–69·0) 6 .. Norway† 1995–99 180 79·2% (71·6–86·9) 177 79·1% (71·4–86·8) 2 .. 42 54·3% (39·6–69·0) 3 .. 2000–04 182 87·7% (82·5–93·0) 178 87·7% (82·3–93·1) 4 .. 37 65·5%§(53·1–77·9) 2 .. 2005–09 169 89·7% (84·5–94·9) 169 89·7% (84·5–94·9) .. .. 29 67·3% (50·1–84·5) 2 .. Poland (Wroclaw) 2000–04 33 68·9%§(56·6–81·2) 33 68·9%§(56·6–81·2) .. .. 8 .. .. .. 2005–09 63 80·9% (70·8–90·9) 62 80·9% (70·8–90·9) 1 .. 16 66·7% (43·6–89·9) .. .. Portugal† 1998–99 45 66·0%§(54·3–77·6) 45 66·0%§(54·3–77·6) .. .. 17 53·0% (30·2–75·8) 1 ..

·5–94·9) 169 89·7% (84·5–94·9) .. .. 29 67·3% (50·1–84·5) 2 .. Poland (Wroclaw) 2000–04 33 68·9%§(56·6–81·2) 33 68·9%§(56·6–81·2) .. .. 8 .. .. .. 2005–09 63 80·9% (70·8–90·9) 62 80·9% (70·8–90·9) 1 .. 16 66·7% (43·6–89·9) .. .. Portugal† 1998–99 45 66·0%§(54·3–77·6) 45 66·0%§(54·3–77·6) .. .. 17 53·0% (30·2–75·8) 1 .. 2000–04 257 79·3% (73·8–84·8) 238 79·2% (73·5–84·8) .. .. 65 53·3% (41·7–65·0) 7 .. 2005–09 209 84·0% (78·8–89·2) 191 83·7% (78·1–89·2) 21 94·5%‡(84·2–100·0) 52 60·4% (48·6–72·3) 12 69·7% (42·5–96·8) Russia (Arkhangelsk)¶ 2000–04 27 55·7% (37·3–74·0) 23 52·2% (32·4–72·1) 1 .. .. .. 6 .. 2005–09 24 74·8% (57·7–92·0) 23 73·1% (55·0–91·2) .. .. .. .. 2 .. Slovakia† 2000–04 132 79·4% (72·1–86·7) 131 79·3% (72·0–86·6) .. .. 35 45·0% (30·0–59·9) 2 .. 2005–07 91 79·1% (70·4–87·7) 89 78·3% (69·5–87·2) .. .. 24 46·7% (23·2–70·3) .. .. Slovenia† 1995–99 50 83·4%§(74·4–92·3) 50 83·4%§(74·4–92·3) .. .. 7 .. .. .. 2000–04 49 89·7%§(81·8–97·6) 49 89·7%§(81·8–97·6) .. .. 11 63·7% (36·8–90·5) .. .. 2005–09 48 75·8%§(65·4–86·2) 43 81·3% (69·7–92·9) 5 .. 15 79·5% (59·5–99·5) 1 .. Spanish registries 1995–99 296 74·4% (68·9–79·9) 290 74·1% (68·5–79·7) .. .. 38 47·3%§(34·7–59·9) 17 58·9% (36·4–81·4) 2000–04 319 81·6% (77·0–86·1) 308 81·9% (77·3–86·5) .. .. 39 64·0% (50·0–78·0) 10 70·0% (43·3–96·8) 2005–09 382 83·7% (79·5–87·9) 362 84·2% (80·0–88·4) 30 80·0%‡ (67·4–92·7) 46 60·2% (47·2–73·2) 5 .. Sweden† 1995–99 368 84·2% (79·8–88·7) 344 85·0% (80·5–89·5) 2 .. 50 72·2%§(61·6–82·7) .. .. 2000–04 333 85·9% (81·7–90·0) 313 86·8% (82·6–90·9) .. .. 56 56·3%§(45·8–66·8) 3 .. 2005–09 283 84·5% (80·1–88·9) 230 85·5% (80·9–90·1) 2 .. 48 65·9%§(55·5–76·4) 4 .. Switzerland† 1995–99 224 86·0% (80·9–91·1) 220 85·6% (80·3–90·8) .. .. 42 53·4% (38·7–68·1) .. .. 2000–04 229 87·6% (82·8–92·4) 222 87·2% (82·3–92·1) .. .. 43 53·7%§(42·3–65·2) 6 .. 2005–09 231 87·9% (83·3–92·6) 225 88·3% (83·7–92·9) 14 100·0%‡(100·0–100·0) 34 75·2%§(64·0–86·3) 4 .. UK† 1995–99 1896 79·1% (77·0–81·2) 1871 79·2% (77·0–81·3) .. .. 371 58·6% (53·4–63·8) 27 70·5% (53·6–87·3) 2000–04 1976 85·9% (84·2–87·7) 1954 86·0% (84·3–87·8) .. .. 390 65·6% (60·7–70·5) 18 55·6% (33·7–77·5) 2005–09 1901 89·3% (87·7–90·9) 1893 89·2% (87·6–90·8) 51 77·8%‡ (66·4–89·1) 368 68·1% (63·2–73·1) 22 76·3% (58·4–94·1) Oceania Australian registries 1995–99 685 82·8% (79·6–86·0) 682 82·9% (79·7–86·1) .. .. 125 53·4% (44·6–62·2) 13 46·2% (20·7–71·7) 2000–04 833 86·0% (83·3–88·7) 825 86·2% (83·4–88·9) .. ..

(33·7–77·5) 2005–09 1901 89·3% (87·7–90·9) 1893 89·2% (87·6–90·8) 51 77·8%‡ (66·4–89·1) 368 68·1% (63·2–73·1) 22 76·3% (58·4–94·1) Oceania Australian registries 1995–99 685 82·8% (79·6–86·0) 682 82·9% (79·7–86·1) .. .. 125 53·4% (44·6–62·2) 13 46·2% (20·7–71·7) 2000–04 833 86·0% (83·3–88·7) 825 86·2% (83·4–88·9) .. .. 170 70·8% (63·8–77·8) 16 68·9% (47·0–90·8) 2005–09 627 88·8% (86·0–91·6) 617 89·0% (86·2–91·8) 17 82·0%‡(64·0–99·9) 129 68·5% (60·3–76·6) 9 .. New Zealand† 1995–99 168 82·8% (76·4–89·3) 167 82·6% (76·0–89·1) 1 .. 42 67·6%§(56·3–78·9) 5 .. 2000–04 183 85·2% (79·3–91·2) 182 85·8% (79·9–91·7) 1 .. 44 68·5%§(57·7–79·4) 6 .. 2005–09 176 89·3% (83·8–94·8) 176 89·3% (83·8–94·8) .. .. 34 74·9%§(63·7–86·1) 5 .. Data stratified by continent, country, and calendar period of diagnosis (1995–99, 2000–04, and 2005–09). Underlined estimates are not age-standardised. ICCC-3= International Classification of Childhood Cancer, 3rd edition. * Estimate judged as less reliable. † National coverage—the data are derived from a population-based cancer registry (registries) covering the entire country. ‡ Estimate based on merging data for 2 (or all 3) calendar periods. § Age-standardised estimate computed by pooling two age-specific estimates and re-estimation. ¶ Korea: Republic of Korea; Russia: Russian Federation.

Introduction Peripheral T-cell lymphoma is a rare and heterogeneous subgroup of non-Hodgkin lymphomas, and comprises approximately 10% of all non-Hodgkin lymphomas in Europe and America.1 Randomised prospective trials in patients with peripheral T-cell lymphoma are scarce and hence there is no consensus on the optimal chemotherapy for previously untreated patients, although combination chemotherapy with cyclophosphamide, doxorubicin, vincristine, and prednisolone (CHOP) is widely used, with consolidative autologous stem-cell transplantation considered in eligible patients. However, for most patients, outcomes with CHOP are poor, with only 33% to 43% achieving a complete response2, 3, 4 and 38·5% achieving 5-year overall survival;5 therefore, a superior upfront regimen for peripheral T-cell lymphoma is urgently required. Previous evidence suggested that the addition of etoposide to CHOP (CHOEP) might improve event-free survival for younger patients (aged ≤60 years) without increased lactate dehydrogenase,6 but no benefit for overall survival was apparent and this approach is not widely applicable. Research in context Evidence before this study

Introduction Peripheral T-cell lymphoma is a rare and heterogeneous subgroup of non-Hodgkin lymphomas, and comprises approximately 10% of all non-Hodgkin lymphomas in Europe and America.1 Randomised prospective trials in patients with peripheral T-cell lymphoma are scarce and hence there is no consensus on the optimal chemotherapy for previously untreated patients, although combination chemotherapy with cyclophosphamide, doxorubicin, vincristine, and prednisolone (CHOP) is widely used, with consolidative autologous stem-cell transplantation considered in eligible patients. However, for most patients, outcomes with CHOP are poor, with only 33% to 43% achieving a complete response2, 3, 4 and 38·5% achieving 5-year overall survival;5 therefore, a superior upfront regimen for peripheral T-cell lymphoma is urgently required. Previous evidence suggested that the addition of etoposide to CHOP (CHOEP) might improve event-free survival for younger patients (aged ≤60 years) without increased lactate dehydrogenase,6 but no benefit for overall survival was apparent and this approach is not widely applicable. Research in context Evidence before this study CHOP combination chemotherapy (cyclophosphamide, doxorubicin, vincristine, and prednisolone) is widely used for treatment of peripheral T-cell lymphoma; however, outcomes with CHOP are poor for most patients. We investigated GEM-P chemotherapy (gemcitabine, methylprednisolone, and cisplatin) compared with CHOP in previously untreated patients with peripheral T-cell lymphoma. We searched PubMed on Dec 4, 2017, for English-language articles published from January, 1998, to October, 2017, and searched abstracts from the American Society of Hematology and American Society of Clinical Oncology published between 2015–17 with the search terms “T-cell lymphoma”, “chemotherapy”, and “gemcitabine”, excluding studies in which patients with non-nodal peripheral T-cell lymphoma forms were assessed exclusively. We identified 30 reports showing activity of gemcitabine in peripheral T-cell lymphoma; three with gemcitabine as monotherapy and nine with a combination of gemcitabine with platinum and steroids predominantly in pretreated populations, including three retrospective reports on GEM-P specifically in peripheral T-cell lymphoma. Different combinations of gemcitabine with other novel drugs, with or without the addition of platinum, in the treatment of peripheral T-cell lymphoma were also reported in the scientific literature, including one randomised trial that assessed the combination of gemcitabine, cisplatin, prednisolone and thalidomide versus CHOP in treatment-naive patients. However, there were no reported randomised studies on the combination of gemcitabine, platinum, and steroids alone versus CHOP in the front-line setting.

erature, including one randomised trial that assessed the combination of gemcitabine, cisplatin, prednisolone and thalidomide versus CHOP in treatment-naive patients. However, there were no reported randomised studies on the combination of gemcitabine, platinum, and steroids alone versus CHOP in the front-line setting. Added value of this study

erature, including one randomised trial that assessed the combination of gemcitabine, cisplatin, prednisolone and thalidomide versus CHOP in treatment-naive patients. However, there were no reported randomised studies on the combination of gemcitabine, platinum, and steroids alone versus CHOP in the front-line setting. Added value of this study Our phase 2 trial is one of the few randomised studies in previously untreated patients with peripheral T-cell lymphoma and is an important addition to the evidence-base. The findings confirm the poor outcomes for peripheral T-cell lymphoma within a prospective trial cohort and indicate that CHOP should remain the reference regimen at present. Our trial was the first prospective study to assess an 18F-FDG-PET-CT response in patients with peripheral T-cell lymphoma as part of a pre-planned substudy, and the data suggest that 18F-FDG-PET-CT might be a more sensitive tool than contrast-enhanced CT for determining response in peripheral T-cell lymphoma, as it might better distinguish between a residual fibrotic mass present after chemotherapy versus viable tumour. Furthermore, obtaining a complete response by 18F-FDG-PET-CT was independently prognostic for superior progression-free survival in multivariable analysis, whereas complete response by CT was not. Additionally, we reported the incidence and pattern of CNS relapse in peripheral T-cell lymphoma from a prospective trial. Determination of the subtype of peripheral T-cell lymphoma in our study was revised for around a fifth of the patients, highlighting the diagnostic challenges in peripheral T-cell lymphoma. Despite GEM-P showing non-significant inferiority for the endpoint of CT-based confirmed and unconfirmed complete response compared with CHOP, there were no differences in either progression-free survival or overall survival between the groups; future study design in this indication should be powered for primary endpoints of survival rather than complete response, which might not be an accurate surrogate endpoint (particularly when assessed by contrast-enhanced CT) in peripheral T-cell lymphoma.

rogression-free survival or overall survival between the groups; future study design in this indication should be powered for primary endpoints of survival rather than complete response, which might not be an accurate surrogate endpoint (particularly when assessed by contrast-enhanced CT) in peripheral T-cell lymphoma. Implications of all the available evidence Taken together, current evidence supports the use of gemcitabine as an effective therapy in the management of peripheral T-cell lymphoma. However, our randomised phase 2 study did not suggest superiority of GEM-P over CHOP in the front-line setting for previously untreated patients and therefore, CHOP should remain the reference regimen in this indication for previously untreated patients. Although GEM-P has shown efficacy in patients with peripheral T-cell lymphoma, at present it should be reserved for patients with relapsed or refractory disease. A superior upfront induction regimen is urgently required for patients with treatment-naive peripheral T-cell lymphoma, and future incorporation of novel drugs could enhance the efficacy of front-line therapy in this indication. Trials are ongoing (ClinicalTrials.gov NCT01777152, NCT01796002, and NCT02561273) to address this research question.

ion regimen is urgently required for patients with treatment-naive peripheral T-cell lymphoma, and future incorporation of novel drugs could enhance the efficacy of front-line therapy in this indication. Trials are ongoing (ClinicalTrials.gov NCT01777152, NCT01796002, and NCT02561273) to address this research question. The nucleoside analogue gemcitabine is not effluxed by the multidrug resistance gene-1–P glycoprotein (MDR-1–Pgp), which is overexpressed in some peripheral T-cell lymphomas7, 8 on the tumour cells, residual lymphocytes, or in the endothelium8 and has shown activity both as monotherapy9, 10, 11 and in combination with platinum and steroids12, 13, 14, 15, 16, 17, 18, 19, 20 in patients with relapsed or refractory peripheral T-cell lymphoma. The regimen of intravenous gemcitabine 1000 mg/m2 on days 1, 8, and 15 of a cycle, intravenous cisplatin 100 mg/m2 on day 15, and oral or intravenous methylprednisolone 1000 mg on days 1–5 (GEM-P) administered every 28 days is associated with 69% to 100% of pretreated patients with peripheral T-cell lymphoma achieving an objective response, and 19% to 50% achieving a complete response;14, 15, 16 a median progression-free survival of 12 months has been reported in the largest retrospective series.16

s 1–5 (GEM-P) administered every 28 days is associated with 69% to 100% of pretreated patients with peripheral T-cell lymphoma achieving an objective response, and 19% to 50% achieving a complete response;14, 15, 16 a median progression-free survival of 12 months has been reported in the largest retrospective series.16 Given the poor outcomes associated with CHOP as front-line therapy in peripheral T-cell lymphoma and the promising data in relapsing and refractory peripheral T-cell lymphoma with GEM-P, the UK National Cancer Research Institute (NCRI) Lymphoma Clinical Study Group started the CHEMO-T trial in 2012 to investigate the potential superiority of GEM-P compared with CHOP in previously untreated patients with peripheral T-cell lymphoma. Methods Study design and participants We did a phase 2, parallel-group, multicentre, open-label randomised trial at 47 hospitals: 46 in the UK and one in Australia (appendix pp 1–2). Eligible participants were patients aged 18 years and older with previously untreated histologically confirmed peripheral T-cell lymphoma of the WHO 2008 subtypes:21 peripheral T-cell lymphoma not otherwise specified, angioimmunoblastic T-cell lymphoma, anaplastic lymphoma kinase (ALK) negative anaplastic large cell T-cell lymphoma, enteropathy-associated T-cell lymphoma, and hepatosplenic γδ T-cell lymphoma.

ogically confirmed peripheral T-cell lymphoma of the WHO 2008 subtypes:21 peripheral T-cell lymphoma not otherwise specified, angioimmunoblastic T-cell lymphoma, anaplastic lymphoma kinase (ALK) negative anaplastic large cell T-cell lymphoma, enteropathy-associated T-cell lymphoma, and hepatosplenic γδ T-cell lymphoma. For participation in the study, patients were required to have bulky stage I (tumour mass diameter >10 cm) to stage IV disease; a WHO performance status of 0–3 (patients with a performance status of 3 were only eligible if this was deemed to be related to a lymphoma); adequate cardiac, renal, hepatic, and bone marrow function (absolute neutrophil count ≥1·0 × 109/L, white blood cell count ≥3·0 × 109/L, platelet count ≥100·0 × 109/L, and haemoglobin concentration ≥9·0 g/dL unless related to disease). Patients with CNS or leptomeningeal involvement, positive serology for HIV-1, active hepatitis B or C, or a history of malignancy within the preceding 5 years (with the exception of curatively treated skin cancers or carcinoma in situ of the cervix) and poorly controlled diabetes or other comorbidities which would preclude the safe delivery of treatment within the trial, were ineligible, as were patients without at least one site of measurable disease at baseline (measurable in two perpendicular dimensions and ≥1 cm on the longest diameter on contrast-enhanced CT scan, except for patients with enteropathy-associated T-cell lymphoma following complete surgical resection). Corticosteroids, not exceeding a maximum of prednisolone 100 mg/day (or equivalent corticosteroid dose) for 7 days, for symptoms related to lymphoma before starting study treatment were permitted, but preferably not instituted before an 18F-fluoro-deoxyglucose (FDG) PET-CT scan at baseline. If corticosteroids were instituted before the baseline scan, they were required to be discontinued 24 h beforehand and patients were required to have a serum glucose concentration of 8·0 mmol/L or lower immediately before the scan.

eferably not instituted before an 18F-fluoro-deoxyglucose (FDG) PET-CT scan at baseline. If corticosteroids were instituted before the baseline scan, they were required to be discontinued 24 h beforehand and patients were required to have a serum glucose concentration of 8·0 mmol/L or lower immediately before the scan. The protocol was approved by the UK Medicines and Healthcare products Regulatory Agency and London Riverside South West Research Ethics Committee. Ethics approval in Australia was from The Eastern Health Human Research Ethics Committee. Patients provided written informed consent. Randomisation We assigned patients (1:1) to either CHOP or GEM-P. Randomisation was done centrally by the clinical trials unit at the Institute for Cancer Research (ICR) independently of the trial team and investigators using a minimisation procedure from the beginning of the trial and first patient, without a burn-in period. The stratification variables of locally-determined histological subtype and International Prognostic Index risk group (low 0–1 vs intermediate 2–3 vs high 4–5) were used for computer-based minimisation. We factored in a probability component (ie, 80% chance of an incoming patient being allocated to an unbalanced group). Given the differences in the administration schedules between the chemotherapy regimens under assessment, it was not possible to mask patients to the treatment they were receiving.

-based minimisation. We factored in a probability component (ie, 80% chance of an incoming patient being allocated to an unbalanced group). Given the differences in the administration schedules between the chemotherapy regimens under assessment, it was not possible to mask patients to the treatment they were receiving. Procedures Patients in the CHOP group received cyclophosphamide 750 mg/m2, doxorubicin 50 mg/m2, and vincristine 1·4 mg/m2 (up to a maximum of 2 mg) intravenously on day 1, and oral prednisolone 100 mg on days 1–5 every 21 days, for six cycles. Patients in the GEM-P group received intravenous gemcitabine 1000 mg/m2 on days 1, 8, and 15, intravenous cisplatin 100 mg/m2 on day 15, and oral or intravenous methylprednisolone 1000 mg on days 1–5 every 28 days, for four cycles. The rationale for administering four cycles of GEM-P (rather than six cycles) was that because GEM-P is a platinum-based regimen used mainly for patients with relapsed or refractory disease, we anticipated that most patients would have achieved a maximum response with this type of regimen by the fourth cycle.

r cycles. The rationale for administering four cycles of GEM-P (rather than six cycles) was that because GEM-P is a platinum-based regimen used mainly for patients with relapsed or refractory disease, we anticipated that most patients would have achieved a maximum response with this type of regimen by the fourth cycle. For both treatment groups, dose modifications were required for treatment-related toxicities in accordance with the study protocol. Dose banding as per local guidelines was permitted. Administration of supportive medication, including antiemetic therapy, prophylaxis for tumour lysis, granulocyte-colony stimulating factor (G-CSF), prophylaxis for Pneumocystis jirovecii pneumonia, fluid and electrolyte administration relating to cisplatin, and administration of prophylaxis for CNS relapse, was given in accordance with local practice. Criteria for withdrawing a patient from study treatment were interruption of study treatment for more than 3 weeks for reasons of toxicity (or for >4 weeks for reasons other than toxicity), disease progression, withdrawal of consent, unacceptable treatment-related toxicity, pregnancy, patient non-compliance, or other events precluding further administration of study drugs as judged by the study chief investigator (DC).

than 3 weeks for reasons of toxicity (or for >4 weeks for reasons other than toxicity), disease progression, withdrawal of consent, unacceptable treatment-related toxicity, pregnancy, patient non-compliance, or other events precluding further administration of study drugs as judged by the study chief investigator (DC). On completion of study treatment, patients with a complete response or unconfirmed complete response could proceed to high-dose chemotherapy and autologous stem-cell transplantation at the local investigator's discretion. Consolidation radiotherapy to sites of initial bulk or residual disease was also permitted at the end of treatment at the investigator's discretion.

complete response or unconfirmed complete response could proceed to high-dose chemotherapy and autologous stem-cell transplantation at the local investigator's discretion. Consolidation radiotherapy to sites of initial bulk or residual disease was also permitted at the end of treatment at the investigator's discretion. Patients were assessed at baseline, at each attendance for treatment, 30 days after completion of treatment, and subsequently every 3 months for 1 year, then every 6 months until year 2, and every year thereafter for a maximum follow-up of 5 years. Laboratory monitoring with a full blood-cell count (haemoglobin, white cells, neutrophils, and platelets) and biochemistry (sodium, potassium, urea, creatinine, calcium, magnesium, phosphate, lactate dehydrogenase, total bilirubin, liver transaminases, alkaline phosphatase, serum albumin, and blood glucose concentrations) were done at baseline, on each day of treatment, and at 30 days after completion of study treatment. At baseline all patients had a contrast enhanced CT scan of the neck, thorax, abdomen and pelvis and an 18F-FDG-PET-CT scan done within 28 days of randomisation, and a bone marrow biopsy was required within 6 weeks of randomisation. CT scans were also required during treatment (after cycles 2 and 4 for patients on CHOP and after cycles 1 and 3 for patients on GEM-P). Both a contrast enhanced CT scan and an 18F-FDG-PET-CT scan were done at the end of treatment. Any clinical suspicion of relapse or progression was confirmed radiologically or on bone marrow aspirate or trephine if disease was marrow-based only.

nd 4 for patients on CHOP and after cycles 1 and 3 for patients on GEM-P). Both a contrast enhanced CT scan and an 18F-FDG-PET-CT scan were done at the end of treatment. Any clinical suspicion of relapse or progression was confirmed radiologically or on bone marrow aspirate or trephine if disease was marrow-based only. All patients enrolled were required to submit their diagnostic tissue for central histopathology review within 28 days of randomisation; CT and 18F-FDG-PET-CT responses were also centrally assessed. Site accreditation, data collation, and quality control for 18F-FDG-PET-CT imaging was done by the UK PET Core Lab (London, UK). Outcomes The primary endpoint of the trial was the proportion of patients who achieved an investigator-assessed, CT-based complete response or unconfirmed complete response on completion of study chemotherapy, according to the International Workshop Standardized Response Criteria for non-Hodgkin lymphoma.22 The primary endpoint was locally assessed and analysed as randomised in the assessable population, defined as all eligible patients who received at least one cycle of treatment and had an end of treatment scan or reported clinical progression as a reason for stopping trial treatment. A patient with no CT imaging but a report of clinical progression as reason for stopping trial treatment was categorised as having progressed.

as all eligible patients who received at least one cycle of treatment and had an end of treatment scan or reported clinical progression as a reason for stopping trial treatment. A patient with no CT imaging but a report of clinical progression as reason for stopping trial treatment was categorised as having progressed. A preplanned primary endpoint sensitivity analysis was done in the intention-to-treat (ITT) population including all eligible patients who received at least one cycle of treatment; patients with no end-of-treatment-CT assessment were counted as non-responders. A second preplanned primary endpoint sensitivity analysis was done as treated in the per-protocol population defined as all patients who received the planned number of cycles and were either assessable by CT or had reported clinical progression at the end of treatment.

ment-CT assessment were counted as non-responders. A second preplanned primary endpoint sensitivity analysis was done as treated in the per-protocol population defined as all patients who received the planned number of cycles and were either assessable by CT or had reported clinical progression at the end of treatment. The secondary endpoints of the trial were overall survival, progression-free survival, partial response, stable disease, progressive disease, and toxicity, and the proportion of patients with metabolic complete response determined by 18F-FDG-PET-CT, according to the Revised Response Criteria for Malignant Lymphoma.23 A secondary endpoint assessing 18F-FDG-PET-CT response specifically was deemed to be of particular importance at the time of study design, as the role of this outcome in peripheral T-cell lymphoma had not previously been assessed in a dedicated, prospective cohort of patients with peripheral T-cell lymphoma. All patients receiving at least one dose of study medication were included in the safety analysis and analysed as treated. The severity of adverse events was defined according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0. The worst toxicity per organ per patient was considered.

atients receiving at least one dose of study medication were included in the safety analysis and analysed as treated. The severity of adverse events was defined according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0. The worst toxicity per organ per patient was considered. A planned subgroup analysis was done to determine the factors associated with complete response or unconfirmed complete response by treatment arm in the assessable population with the following factors: peripheral T-cell lymphoma subtype, International Prognostic Index, sex, B symptoms, age, disease stage, WHO performance status, presence of extranodal disease, and raised lactate dehydrogenase concentration. Statistical analysis We calculated the patient sample size for this trial estimating that 50% of patients in the CHOP group and 70% of patients in the GEM-P group would achieve a complete response or unconfirmed complete response by CT, with an odds ratio of 2·33. We calculated that 93 patients per group were required to detect this difference with 5% significance (two-sided) and 80% power. Continuity correction was not used.

the CHOP group and 70% of patients in the GEM-P group would achieve a complete response or unconfirmed complete response by CT, with an odds ratio of 2·33. We calculated that 93 patients per group were required to detect this difference with 5% significance (two-sided) and 80% power. Continuity correction was not used. A planned formal interim analysis with 90% power to show non-inferiority for review by an independent data monitoring committee was due after the first 51 patients in each group had been assessed for end-of-treatment response; however, this milestone was not reached as the trial closed before accruing 51 patients per group. A lower 90% one-sided CI of difference in the number of patients achieving a confirmed or unconfirmed complete response between treatments was expected to be greater than 25% (ie, a non-inferiority margin of 25%). The number of patients achieving a complete response or unconfirmed complete response at the end of study chemotherapy was reported by treatment group with 95% CI using normal approximation and compared between groups using logistic regression, with odds ratio (OR [95% CI]) for CHOP versus GEM-P. Additionally, a logistic regression model was fitted to adjust for the stratification variables, with OR (95% CI) for CHOP versus GEM-P.

rapy was reported by treatment group with 95% CI using normal approximation and compared between groups using logistic regression, with odds ratio (OR [95% CI]) for CHOP versus GEM-P. Additionally, a logistic regression model was fitted to adjust for the stratification variables, with OR (95% CI) for CHOP versus GEM-P. In the planned subgroup analysis, we used logistic regression to assess treatment differences in responses within the predefined subgroups, with results displayed as ORs in a Forest plot. Progression events were defined as clinical or radiological documented disease progression or death from any cause. Patients recording no event were censored at the last follow-up date. Progression-free survival and overall survival were calculated from the date of randomisation until a progression or death event occurred, respectively, using Kaplan-Meier methods and compared between groups using a log-rank test. Additionally, a Cox proportional hazards model was fitted to adjust for the stratification variables and to calculate a hazard ratio (HR [95% CI]) for CHOP versus GEM-P. The ITT population was used for survival analyses.

curred, respectively, using Kaplan-Meier methods and compared between groups using a log-rank test. Additionally, a Cox proportional hazards model was fitted to adjust for the stratification variables and to calculate a hazard ratio (HR [95% CI]) for CHOP versus GEM-P. The ITT population was used for survival analyses. We assessed the following factors affecting overall survival and progression-free survival using Cox regression analysis in the ITT population combining both treatment groups: age, sex, disease stage, WHO performance status, B symptoms, presence of raised lactate dehydrogenase, International Prognostic Index risk group, subtype of peripheral T-cell lymphoma, number of extranodal sites, treatment arm, local CT response after study chemotherapy, central 18F-FDG-PET-CT response after study chemotherapy, and autologous stem-cell transplantation following first-line treatment. For safety analyses, the proportion of patients reporting toxicities of grade 3 or worse was compared between groups using a χ2 test. The data analysis was generated using Stata version 14. The trial was overseen by trial management and trial steering committees and an independent data monitoring committee. This trial is registered with ClinicalTrials.gov (NCT01719835) and the European Clinical Trials Database (EudraCT 2011-004146-18).

We assessed the following factors affecting overall survival and progression-free survival using Cox regression analysis in the ITT population combining both treatment groups: age, sex, disease stage, WHO performance status, B symptoms, presence of raised lactate dehydrogenase, International Prognostic Index risk group, subtype of peripheral T-cell lymphoma, number of extranodal sites, treatment arm, local CT response after study chemotherapy, central 18F-FDG-PET-CT response after study chemotherapy, and autologous stem-cell transplantation following first-line treatment. For safety analyses, the proportion of patients reporting toxicities of grade 3 or worse was compared between groups using a χ2 test. The data analysis was generated using Stata version 14. The trial was overseen by trial management and trial steering committees and an independent data monitoring committee. This trial is registered with ClinicalTrials.gov (NCT01719835) and the European Clinical Trials Database (EudraCT 2011-004146-18). Role of the funding source The trial sponsor (The Royal Marsden Hospital, London, UK) was responsible for randomisation, data gathering, entry, and validation, reports of serious adverse events, organisation of the central histopathology review and central response assessment, statistical analysis, and production of the report. 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.

tical analysis, and production of the report. 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 June 18, 2012, and Nov 16, 2016, we randomly assigned 87 patients to study treatment, 43 to CHOP and 44 to GEM-P (figure 1). Before the formal interim analysis, interim results from a planned unmasked independent review of efficacy data by the independent data monitoring committee in November, 2016, showed that fewer patients had achieved a complete response or unconfirmed complete response with GEM-P than with CHOP. The committee concluded there was a high likelihood that GEM-P would be inferior to CHOP at the end of the trial, as indicated by the one-sided 80% CI for difference. The trial was subsequently closed to recruitment and all patients treated in the GEM-P group were offered the option to change to CHOP off-study at the time of study closure as per the recommendations of the data monitoring committee and trial management group. The last dose of study treatment on trial was administered on Feb 23, 2017.Figure 1 Trial profile

cruitment and all patients treated in the GEM-P group were offered the option to change to CHOP off-study at the time of study closure as per the recommendations of the data monitoring committee and trial management group. The last dose of study treatment on trial was administered on Feb 23, 2017.Figure 1 Trial profile Toxicity reasons for early study withdrawal were neutropenic sepsis (n=1) in CHOP; and tinnitus (n=2), hearing loss (n=1), peripheral neuropathy (n=1), and thrombocytopenia and anaemia (n=1) in GEM-P. Two deaths occurred in the GEM-P group, both due to lung infections. One patient chose to discontinue GEM-P and commence CHOP off-study at the time of study closure. GEM-P=gemcitabine, cisplatin, and methylprednisolone. CHOP=cyclophosphamide, doxorubicin, vincristine, and prednisolone. ITT=intention-to-treat. *Includes three patients with clinical progression. Patient baseline and disease characteristics were well balanced between groups (table 1). All 71 patients with sufficient tissue for central histopathology review had the diagnosis of peripheral T-cell lymphoma confirmed centrally, although the subtype of peripheral T-cell lymphoma was revised for 16 patients (23%).Table 1 Baseline characteristics of all randomly assigned patients

between groups (table 1). All 71 patients with sufficient tissue for central histopathology review had the diagnosis of peripheral T-cell lymphoma confirmed centrally, although the subtype of peripheral T-cell lymphoma was revised for 16 patients (23%).Table 1 Baseline characteristics of all randomly assigned patients CHOP (six cycles; n=43) GEM-P (four cycles; n=44) Sex Men 30 (70%) 32 (73%) Women 13 (30%) 12 (27%) Age (years; median IQR) 64 (54–69) 61 (50–70) Aged >60 years 27 (63%) 24 (55%) IPI score 0–1 9 (21%) 8 (18%) 2–3 26 (60%) 25 (57%) 4–5 8 (19%) 11 (25%) Local histology PTCL not otherwise specified 19 (44%) 18 (41%) ALK-negative ALCL 6 (14%) 8 (18%) AITL 17 (40%) 17 (39%) EATL 1 (2%) 1 (2%) Hepatosplenic γδ T-cell lymphoma 0 0 Central histology PTCL not otherwise specified 11 (26%) 10 (23%) ALK-negative ALCL 4 (9%) 4 (9%) AITL 22 (51%) 17 (39%) EATL 1 (2%) 0 Panniculitis-like T-cell lymphoma 0 1 (2%) PTCL not otherwise specified or panniculitis-like T-cell lymphoma 0 1 (2%) Not assessable 5 11 (25%) Stage I 1 (2%) 0 II 8 (19%) 3 (7%) III 16 (37%) 16 (36%) IV 18 (42%) 25 (57%) B symptoms present 26 (60%) 27 (61%) Increased lactate dehydrogenase 27 (63%) 27 (61%) Extranodal sites present 25 (58%) 30 (68%) WHO performance status 0 17 (40%) 21 (48%) 1 19 (44%) 18 (41%) 2 7 (16%) 5 (11%) Data are n (%) except where indicated otherwise. CHOP=cyclophosphamide, doxorubicin, vincristine, and prednisolone. GEM-P=gemcitabine, cisplatin, and methylprednisolone. IPI=International Prognostic Index. PTCL=peripheral T-cell lymphoma. ALK-negative ALCL=anaplastic lymphoma kinase-negative anaplastic large cell lymphoma. AITL=angioimmunoblastic T-cell lymphoma. EATL=enteropathy-associated T-cell lymphoma.

istine, and prednisolone. GEM-P=gemcitabine, cisplatin, and methylprednisolone. IPI=International Prognostic Index. PTCL=peripheral T-cell lymphoma. ALK-negative ALCL=anaplastic lymphoma kinase-negative anaplastic large cell lymphoma. AITL=angioimmunoblastic T-cell lymphoma. EATL=enteropathy-associated T-cell lymphoma. After receiving four cycles of GEM-P, one patient in the GEM-P group received an additional two cycles for persistent bone marrow infiltration before autologous stem cell transplantation. 16 patients (38%) of 42 treated with CHOP and 23 (52%) of 44 treated with GEM-P had at least one treatment delay, and 18 patients (43%) of 42 treated with CHOP and 21 (48%) of 44 treated with GEM-P had at least one dose reduction on study. Median total dose received and dose intensities for each drug by treatment group are in table 2.Table 2 Total chemotherapy dose received by patients and dose intensity achieved CHOP GEM-P Cyclophosphamide Doxorubicin Vincristine Prednisolone Gemcitabine Cisplatin Methylprednisolone Total dose received (mg) 7960 (5080–8760) 540 (342–588) 12 (7–12) 2400 (1500–3000) 18262·5 (7060–21900) 361·5 (201·5–707·5) 20 000 (7500–20 000) Relative dose intensity 98·1 (90·5–100) 97·9 (91·9–100·0) 99·2 (85·7–100·0) 97·0 (89·7–100·0) 86·5 (69·2–95·9) 82·2 (41·6–97·4) 98·1 (83·3–100·0) Data are median (IQR).

solone Total dose received (mg) 7960 (5080–8760) 540 (342–588) 12 (7–12) 2400 (1500–3000) 18262·5 (7060–21900) 361·5 (201·5–707·5) 20 000 (7500–20 000) Relative dose intensity 98·1 (90·5–100) 97·9 (91·9–100·0) 99·2 (85·7–100·0) 97·0 (89·7–100·0) 86·5 (69·2–95·9) 82·2 (41·6–97·4) 98·1 (83·3–100·0) Data are median (IQR). The proportions of CHOP and GEM-P treatments requiring G-CSF support were 135 (62%) of 219 and 70 (54%) of 130, respectively. Two patients (5%) of 43 in the CHOP group and four (9%) of 44 in the GEM-P group received prophylaxis for CNS relapse during the study: four (9%) received intrathecal methotrexate, one (2%) received intrathecal methotrexate plus methotrexate, cytarabine, and hydrocortisone, and one (2%) received unspecified therapy.

y. Two patients (5%) of 43 in the CHOP group and four (9%) of 44 in the GEM-P group received prophylaxis for CNS relapse during the study: four (9%) received intrathecal methotrexate, one (2%) received intrathecal methotrexate plus methotrexate, cytarabine, and hydrocortisone, and one (2%) received unspecified therapy. Investigator-assessed end of treatment response by contrast-enhanced CT was available for 74 patients (37 in each group; table 3). At a median follow-up of 27·4 months (IQR 16·6–38·4), 23 patients (62%) of 37 assigned to CHOP had achieved a complete response or unconfirmed complete response compared with 17 (46%) of 37 assigned to GEM-P (OR 0·52, 95% CI 0·21–1·31; p=0·164; adjusted OR for stratification factors 0·53 (0·19–1·46, p=0·22). Given the higher proportion of patients with stage IV disease in the GEM-P group versus CHOP, the logistic regression was adjusted for stage as well as for stratification variables, but this did not make any difference to the complete response outcome by randomisation result (OR 0·50, 95% CI 0·18–1·40; p=0·19). Sensitivity analyses of the primary endpoint in both the ITT and per-protocol populations showed no differences between the treatment groups (appendix p 3). The inferiority of complete response outcomes with GEM-P was consistent across subgroups (figure 2).Figure 2 Factors potentially predictive of a complete response or unconfirmed complete response

ary endpoint in both the ITT and per-protocol populations showed no differences between the treatment groups (appendix p 3). The inferiority of complete response outcomes with GEM-P was consistent across subgroups (figure 2).Figure 2 Factors potentially predictive of a complete response or unconfirmed complete response Data are from an analysis of outcomes at end of treatment. CHOP=cyclophosphamide, doxorubicin, vincristine, and prednisolone. GEM-P=gemcitabine, cisplatin, and methylprednisolone. ALK-negative ALCL= anaplastic lymphoma kinase-negative anaplastic large cell T-cell lymphoma. AITL=angioimmunoblastic T-cell lymphoma. PTCL NOS=peripheral T-cell lymphoma not otherwise specified. EATL=enteropathy-associated T-cell lymphoma. IPI=International Prognostic Index. ··=data not obtainable. Table 3 End of treatment response by CT CHOP (six cycles; n=43) GEM-P (four cycles; n=44) Overall response 28 (75·7) 25 (67·6) Complete response or unconfirmed complete response* 23 (62·2) 17 (45·9) Partial response 5 (13·5) 8 (21·6) Stable disease 2 (5·4) 3 (8·1) Progressive disease 4 (10·8) 6 (16·2) Progressive disease assessed clinically 3 (8·1) 3 (8·1) Not done or not assessable 6 7 Data are n (%) of those who had an assessment. CHOP=cyclophosphamide, doxorubicin, vincristine, and prednisolone. GEM-P=gemcitabine, cisplatin, and methylprednisolone. * p=0·164.

CHOP (six cycles; n=43) GEM-P (four cycles; n=44) Overall response 28 (75·7) 25 (67·6) Complete response or unconfirmed complete response* 23 (62·2) 17 (45·9) Partial response 5 (13·5) 8 (21·6) Stable disease 2 (5·4) 3 (8·1) Progressive disease 4 (10·8) 6 (16·2) Progressive disease assessed clinically 3 (8·1) 3 (8·1) Not done or not assessable 6 7 Data are n (%) of those who had an assessment. CHOP=cyclophosphamide, doxorubicin, vincristine, and prednisolone. GEM-P=gemcitabine, cisplatin, and methylprednisolone. * p=0·164. After chemotherapy, four (9%) of 43 patients in the CHOP group underwent radiotherapy; no patients in the GEM-P arm received radiotherapy. Stem cells were collected for 16 patients (37%) of 43 in the CHOP group and 15 (34%) of 44 in the GEM-P group, with 11 (26%) in CHOP and 13 (30%) in GEM-P proceeding to autologous stem-cell transplantation in first remission. Second-line chemotherapy was administered to 15 patients (35%) of 43 in the CHOP group and 20 (45%) of 44 in the GEM-P group.

6 patients (37%) of 43 in the CHOP group and 15 (34%) of 44 in the GEM-P group, with 11 (26%) in CHOP and 13 (30%) in GEM-P proceeding to autologous stem-cell transplantation in first remission. Second-line chemotherapy was administered to 15 patients (35%) of 43 in the CHOP group and 20 (45%) of 44 in the GEM-P group. Survival was assessed in the ITT population (84 patients [97%] of 87). Three patients (two in the CHOP group, one in the GEM-P group) were excluded from the ITT population following treatment assignment, two because of a change in diagnosis from peripheral T-cell lymphoma not otherwise specified to Hodgkin's lymphoma after randomisation and a third patient was deemed to be ineligible following treatment assignment because of cardiac impairment. 2-year progression-free survival was 38·0% (95% CI 22·9–52·9) in the GEM-P group and 36·6% (21·4–52·0) in the CHOP group (HR 1·07, 95% CI 0·61–1·85, p=0·82; appendix p 4). Two patients (2%) of 84 had CNS relapses (both isolated CNS recurrences); both patients had peripheral T-cell lymphoma not otherwise specified, were treated with CHOP, and developed progression in the CNS after cycle 1 and cycle 5, respectively. 2-year overall survival was 63·9% (95% CI 45·7–77·4) in the GEM-P group and 51·0% (32·8–67·4) in the CHOP group (HR 0·69, 95% CI 0·35–1·38, p=0·30; appendix p 5).

had peripheral T-cell lymphoma not otherwise specified, were treated with CHOP, and developed progression in the CNS after cycle 1 and cycle 5, respectively. 2-year overall survival was 63·9% (95% CI 45·7–77·4) in the GEM-P group and 51·0% (32·8–67·4) in the CHOP group (HR 0·69, 95% CI 0·35–1·38, p=0·30; appendix p 5). Significant risk factors associated with superior overall survival in univariable analysis (appendix p 6) were a complete response or unconfirmed complete response on completion of treatment determined by CT, a complete response on completion of treatment determined by 18F-FDG-PET-CT, and autologous stem-cell transplantation consolidation in first remission. No factor remained independently prognostic for overall survival in multivariable analysis (appendix p7). Presence of raised lactate dehydrogenase, an International Prognostic Index score of more than 1, and presence of more than one extranodal site was associated with inferior progression-free survival in univariable analysis (appendix p 8), while patients with a complete response or unconfirmed complete response to induction determined by CT or a complete response to induction determined by 18F-FDG-PET-CT or undergoing autologous stem-cell transplantation consolidation in first remission had superior progression-free survival. In multivariable analysis presence of raised lactate dehydrogenase concentrations at diagnosis was independently associated with inferior progression-free survival; by contrast, attainment of a complete response by 18F-FDG-PET-CT to study treatment or presence of a low or low-intermediate risk international prognostic index score was independently associated with superior progression-free survival (appendix p9)

was independently associated with inferior progression-free survival; by contrast, attainment of a complete response by 18F-FDG-PET-CT to study treatment or presence of a low or low-intermediate risk international prognostic index score was independently associated with superior progression-free survival (appendix p9) Two patients (5%) died during the study, both in the GEM-P group, from lung infections. Up to Nov 7, 2017, 33 patients (39%) of 84 had died in the ITT population, 18 (44%) of 41 in the CHOP group (13 from disease progression and five from sepsis) and 15 (35%) of 43 in the GEM-P group (nine from disease progression, three from sepsis, one cardiac event, one from ischaemic colitis and disease progression, and one from metastatic cancer). Safety was assessed for 86 patients (99%) who received a dose of study treatment. CHOP was associated with more febrile neutropenia of all grades (table 4). GEM-P was associated with more thrombocytopenia of all grades and grade 3 or worse; however, thrombocytopenia was not associated with increased bleeding. GEM-P was also associated with more tinnitus. There were 38 serious adverse events in the CHOP group and 43 in the GEM-P group.Table 4 Adverse events by treatment

4). GEM-P was associated with more thrombocytopenia of all grades and grade 3 or worse; however, thrombocytopenia was not associated with increased bleeding. GEM-P was also associated with more tinnitus. There were 38 serious adverse events in the CHOP group and 43 in the GEM-P group.Table 4 Adverse events by treatment CHOP (n=42) GEM-P (n=44) Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 Acute renal toxicity 0 0 0 1 (2%) 1 (2%) 0 Alopecia* 27 (64%) 0 0 12 (27%) 0 0 Anaemia 23 (55%) 4 (10%) 0 22 (50%) 5 (11%) 1 (2%) Constipation 18 (43%) 0 0 20 (45%) 1 (2%) 0 Cerebrovascular accident 0 0 0 0 0 1 (2%) Diarrhoea 15 (36%) 0 0 13 (30%) 2 (5%) 0 Dyspnoea 4 (10%) 0 0 6 (14%) 0 0 Fatigue 29 (69%) 2 (5%) 0 35 (80%) 0 0 Febrile neutropenia† 0 10 (24%) 2 (5%) 1 (2%) 3 (7%) 0 Fever 11 (26%) 3 (7%) 0 11 (25%) 2 (5%) 1 (2%) Haemorrhage 4 (10%) 0 0 7 (16%) 0 1 (2%) Headache* 4 (10%) 0 0 13 (30%) 0 0 Hyperglycaemia 1 (2%) 1 (2%) 0 3 (7%) 1 (2%) 0 Hypertension 1 (2%) 1 (2%) 0 3 (7%) 1 (2%) 0 Hypotension 3 (7%) 1 (2%) 0 5 (11%) 1 (2%) 0 Indigestion 14 (33%) 0 0 8 (18%) 0 0 Infection 8 (19%) 7 (17%) 0 14 (32%) 4 (9%) 2 (5%) Infusion reaction 1 (2%) 0 0 1 (2%) 1 (2%) 0 Mood disturbance 6 (14%) 0 0 7 (16%) 1 (2%) 0 Mucositis* 20 (48%) 1 (2%) 0 11 (25%) 0 0 Nausea 17 (40%) 0 0 22 (50%) 1 (2%) 0 Neuropathy 11 (26%) 1 (2%) 0 7 (16%) 0 0 Neutropenia 5 (12%) 2 (5%) 15 (36%) 7 (16%) 5 (11%) 4 (9%) Pulmonary embolus 0 2 (5%) 1 (2%) 0 1 (2%) 0 Pruritus 7 (17%) 0 0 9 (20%) 0 0 Skin rash 7 (17%) 0 0 12 (27%) 0 0 Thrombocytopenia† 8 (19%) 1 (2%) 3 (7%) 16 (36%) 8 (18%) 5 (11%) Tinnitus* 1 (2%) 0 0 12 (27%) 0 0 Vomiting 12 (29%) 1 (2%) 0 13 (30%) 2 (5%) 0 Weight loss 10 (24%) 1 (2%) 0 9 (20%) 1 (2%) 0 Other toxicity 23 (55%) 5 (12%) 1 (2%) 27 (61%) 9 (20%) 2 (5%) Data are n (%) for adverse events of grade 1–2 occurring in at least 10% of patients and all grade 3–5 events. Two patients (5%) died during the study, both in the GEM-P group, from lung infections. Other adverse events with CHOP of grade 3 or worse were superior mesenteric artery thrombosis (grade 3), muscle weakness (grade 3), insomnia (grade 3), hyponatraemia (grade 4), tumour lysis (grade 3), pain left hip and legs (grade 3), swelling in left knee (grade 3), and squamous cell carcinoma cheek (grade 3).

ctions. Other adverse events with CHOP of grade 3 or worse were superior mesenteric artery thrombosis (grade 3), muscle weakness (grade 3), insomnia (grade 3), hyponatraemia (grade 4), tumour lysis (grade 3), pain left hip and legs (grade 3), swelling in left knee (grade 3), and squamous cell carcinoma cheek (grade 3). Other adverse events with GEM-P of grade 3 or worse were right flank pain (grade 3), abdominal pain (grade 3), colonic perforation (grade 4), hip pain (grade 3), chest pain (grade 3), chest/abdominal pain (grade 3), paraneoplastic occurrence (grade 4), raised alanine transaminase (grade 3), bone pain (grade 3), dehydration (grade 3), hypokalaemia (grade 3), raised alanine transaminase (grade 3), and cellulitis (grade 3). CHOP=cyclophosphamide, doxorubicin, vincristine, and prednisolone. GEM-P=gemcitabine, cisplatin, and methylprednisolone. * p<0.05 for all grades. † p<0.05 for all grades and grade 3 or worse.

Other adverse events with GEM-P of grade 3 or worse were right flank pain (grade 3), abdominal pain (grade 3), colonic perforation (grade 4), hip pain (grade 3), chest pain (grade 3), chest/abdominal pain (grade 3), paraneoplastic occurrence (grade 4), raised alanine transaminase (grade 3), bone pain (grade 3), dehydration (grade 3), hypokalaemia (grade 3), raised alanine transaminase (grade 3), and cellulitis (grade 3). CHOP=cyclophosphamide, doxorubicin, vincristine, and prednisolone. GEM-P=gemcitabine, cisplatin, and methylprednisolone. * p<0.05 for all grades. † p<0.05 for all grades and grade 3 or worse. In the 18F-FDG-PET-CT substudy, 79 (91%) of 87 patients assessed had FDG-avid disease at baseline on central review of imaging. End-of-treatment response by 18F-FDG-PET-CT or CT was assessable in 70 patients (89%) of 79 (appendix p 10); although the incidence of complete response as determined by 18F-FDG-PET-CT was lower in the GEM-P group than in the CHOP group, the difference between groups was less marked than between-group differences for complete response assessment by contrast-enhanced CT. The proportion of agreement between 18F-FDG-PET-CT and contrast-enhanced CT for determining a complete response was 77·3%. For 40 patients with a complete response by 18F-FDG-PET-CT on completion of study treatment, the corresponding contrast-enhanced CT data were: 24 (60%) with a complete response, seven (18%) with an unconfirmed complete response, five (13%) with a partial response, two (5%) with stable disease, and two (5%) with progressive disease.

ts with a complete response by 18F-FDG-PET-CT on completion of study treatment, the corresponding contrast-enhanced CT data were: 24 (60%) with a complete response, seven (18%) with an unconfirmed complete response, five (13%) with a partial response, two (5%) with stable disease, and two (5%) with progressive disease. Discussion The number of complete responses or unconfirmed complete responses in the GEM-P group was non-significantly inferior to the CHOP group, indicating that the primary endpoint of the trial would not be met. A planned subgroup analysis showed that the effect was consistent across subgroups.

ts with a complete response by 18F-FDG-PET-CT on completion of study treatment, the corresponding contrast-enhanced CT data were: 24 (60%) with a complete response, seven (18%) with an unconfirmed complete response, five (13%) with a partial response, two (5%) with stable disease, and two (5%) with progressive disease. Discussion The number of complete responses or unconfirmed complete responses in the GEM-P group was non-significantly inferior to the CHOP group, indicating that the primary endpoint of the trial would not be met. A planned subgroup analysis showed that the effect was consistent across subgroups. As part of a preplanned analysis, we also assessed the proportion of agreement between contrast-enhanced CT and 18F-FDG-PET-CT for determining a complete response, and found it to be around 77%. This discrepancy was probably mainly due to the presence of persistent non-FDG-avid nodes discernible on contrast-enhanced CT after chemotherapy and 18F-FDG-PET-CT might be better able to distinguish between a residual fibrotic mass following chemotherapy versus viable tumour in peripheral T-cell lymphoma. The rate of early withdrawal before completing study treatment for reasons other than disease progression was higher in the GEM-P group with more participants withdrawing for toxicity and by choice; there were also two deaths on treatment with GEM-P, both due to lung infections. Patients treated in the CHOP group had more febrile neutropenia than those in the GEM-P group; by contrast, patients treated with GEM-P had more thrombocytopenia of all grades and grade 3 or worse than those treated with CHOP, although this did not lead to a substantial increase in bleeding events. GEM-P was also associated with a significant increased risk of grade 1-2 tinnitus, which led two patients to withdraw early from the study.

eated with GEM-P had more thrombocytopenia of all grades and grade 3 or worse than those treated with CHOP, although this did not lead to a substantial increase in bleeding events. GEM-P was also associated with a significant increased risk of grade 1-2 tinnitus, which led two patients to withdraw early from the study. Relative dose intensity was lower in the GEM-P group than in the CHOP group, although this might mainly reflect the dose modifications that were required with the GEM-P regimen according to blood results on the day of treatment. However, there were two separate instances where a day of GEM-P treatment was incorrectly omitted rather than delayed, resulting in two protocol violations, although it seems unlikely that this modification would have significantly negatively affected the primary endpoint assessment in the GEM-P arm. At the median follow-up of 27·4 months there was no difference in either 2-year progression-free survival or overall survival between the treatment groups. In a Cox regression analysis of the ITT population, significant risk factors associated with superior overall survival included a complete response at end of treatment determined either by CT or by 18F-FDG-PET-CT, and autologous stem-cell transplantation consolidation in first remission in univariable analysis. However, no factors remained independently significant for overall survival in multivariable analysis. In multivariable analysis of the factors associated with progression-free survival, presence of a raised lactate dehydrogenase concentration at presentation was prognostic of inferior progression-free survival, whereas a low or low-intermediate risk international prognostic index score or a complete response to study treatment determined by 18F-FDG-PET-CT was independently associated with superior progression-free survival.

ctate dehydrogenase concentration at presentation was prognostic of inferior progression-free survival, whereas a low or low-intermediate risk international prognostic index score or a complete response to study treatment determined by 18F-FDG-PET-CT was independently associated with superior progression-free survival. The number of patients who withdrew from the study for reasons other than disease progression was higher in the GEM-P group than in the CHOP group, and consequently more patients receiving GEM-P proceeded to second-line chemotherapy. This might explain the absence of a survival difference between arms at this time, as the numbers undergoing autologous stem-cell transplantation in first remission were similar between groups, although this remains to be further investigated. However, given that the inferior incidence of complete responses and unconfirmed complete responses in the GEM-P group was not associated with inferior progression-free survival or overall survival for these patients, complete response (particularly when assessed by contrast-enhanced CT) might not be an accurate surrogate endpoint in peripheral T-cell lymphoma. Future studies in this area should therefore better investigate this point, and perhaps should be designed with primary endpoints that incorporate survival. Indeed, measuring endpoints such as 24-month event-free survival in peripheral T-cell lymphoma might be an important predictor of overall survival; one study showed that patients achieving 24-month event-free survival had superior overall survival compared with those who did not reach this milestone.24

porate survival. Indeed, measuring endpoints such as 24-month event-free survival in peripheral T-cell lymphoma might be an important predictor of overall survival; one study showed that patients achieving 24-month event-free survival had superior overall survival compared with those who did not reach this milestone.24 Although our study closed early to recruitment, this phase 2 trial is an important addition to the prospective evidence-base for this rare subtype of non-Hodgkin lymphoma and is, to our knowledge, the third randomised trial reported with treatment-naive patients with peripheral T-cell lymphoma. All participants had the diagnosis and subtype of peripheral T-cell lymphoma confirmed centrally by two expert lymphoma pathologists; for around 23% of patients with peripheral T-cell lymphoma the subtype was revised, highlighting that there are challenges to accurately diagnosing peripheral T-cell lymphoma. Additionally, although extensive future work is necessary, our study reports, to our knowledge, the first dedicated prospective trial to assess 18F-FDG-PET-CT responses in patients with peripheral T-cell lymphoma, all of which were done at accredited sites, underwent central quality control, and were centrally assessed by an expert 18F FDG PET-CT physician. Our study also adds important data from a prospective trial cohort on the incidence of CNS relapse in peripheral T-cell lymphoma, which at the time of writing has been reported in two patients with peripheral T-cell lymphoma not otherwise specified and occurred during study treatment.

n expert 18F FDG PET-CT physician. Our study also adds important data from a prospective trial cohort on the incidence of CNS relapse in peripheral T-cell lymphoma, which at the time of writing has been reported in two patients with peripheral T-cell lymphoma not otherwise specified and occurred during study treatment. However, the limitations of our study also need to be acknowledged. Firstly, and most importantly, as the trial closed early to recruitment it was not adequately powered to assess the primary study endpoint. Another potential limitation of the study is in the interpretation of dose intensity for GEM-P; this regimen was administered weekly for 3 weeks out of 4, with the dose adjusted on each day of treatment in accordance with the blood counts on the day of treatment, which is in contrast to CHOP where treatment is administered once every 3 weeks allowing more time for count recovery.

tion of dose intensity for GEM-P; this regimen was administered weekly for 3 weeks out of 4, with the dose adjusted on each day of treatment in accordance with the blood counts on the day of treatment, which is in contrast to CHOP where treatment is administered once every 3 weeks allowing more time for count recovery. Despite the absence of international consensus on the gold-standard therapy for previously untreated patients with peripheral T-cell lymphoma, CHOP has been adopted by many countries as the reference regimen.25 With the exception of ALK-positive anaplastic large cell T-cell lymphoma26 the outcomes with CHOP or CHOP-like therapy appear to be suboptimum, with a 5-year overall survival of only 38·5%, as reported in a meta-analysis.5 Attempts to surpass outcomes with CHOP have proven largely unsuccessful, except a subgroup analysis by the German High-Grade NHL Study Group that showed improved event-free survival but no overall survival benefit with CHOEP for younger patients (≤60 years) with peripheral T-cell lymphoma with normal lactate dehydrogenase concentrations treated within aggressive non-Hodgkin lymphoma clinical trials.6 Therefore, optimising therapy for newly diagnosed patients remains an important unmet medical need.

ut no overall survival benefit with CHOEP for younger patients (≤60 years) with peripheral T-cell lymphoma with normal lactate dehydrogenase concentrations treated within aggressive non-Hodgkin lymphoma clinical trials.6 Therefore, optimising therapy for newly diagnosed patients remains an important unmet medical need. MDR-1/Pgp is known to be overexpressed in peripheral T-cell lymphoma7, 8 in the lymphoma cells, residual lymphocytes, and endothelium,8 and this might account for the reported efficacy of the nucleoside analogue gemcitabine in peripheral T-cell lymphoma, which is not a substrate of the efflux pump.27 The activity of intravenous gemcitabine monotherapy (1200 mg/m2 on days 1, 8, and 15 of a 28-day cycle) was first reported in 1998 in a phase 2 study9 of 13 patients with relapsed or refractory peripheral T-cell lymphoma unspecified or patients with mycosis fungoides with an objective response of 69% and complete response of 8%. Similar efficacy was shown in another study10 (objective response 60% and complete response 20%) in a cohort of ten patients with relapsed or refractory T-cell lymphoma using the same dosing schedule. A subsequent study in 2010 confirmed this efficacy (objective response 51%, complete response 23%) in a larger number of relapsed or refractory patients (n=39) with some durable remissions.11

d complete response 20%) in a cohort of ten patients with relapsed or refractory T-cell lymphoma using the same dosing schedule. A subsequent study in 2010 confirmed this efficacy (objective response 51%, complete response 23%) in a larger number of relapsed or refractory patients (n=39) with some durable remissions.11 Various combinations of gemcitabine with platinum and steroid in relapsed or refractory peripheral T-cell lymphoma have also been reported in the scientific literature with encouraging results (objective response 36%–100%) and acceptable toxicity.12, 13, 14, 15, 16, 17, 18, 19, 20, 28 Specifically, the GEM-P regimen has been assessed in three retrospective studies including patients with peripheral T-cell lymphoma, predominantly with relapsed or refractory disease, with reported objective responses of 69%–100% (complete response 19%–50%) and some durable remisisons.14, 15, 16 Furthermore, some novel regimens incorporating gemcitabine have been assessed in treatment-naive patients in the peripheral T-cell lymphoma setting,4, 8, 29, 30 including one randomised trial,4 and all have shown encouraging results for gemcitabine in this indication excepting one.8 In one trial, patients were assigned either to CHOP (n=51) or a combination of gemcitabine 800 mg/m2 on days 1 and 8, cisplatin 25 mg/m2 intravenously on days 1–3, prednisolone 60 mg/m2 orally on days 1–5, and thalidomide 200 mg orally once per day continuously (n=52); this regimen was associated with a significant improvement in the proportion of patients achieving a response, 2-year progression-free survival, and 2-year overall survival.4

tin 25 mg/m2 intravenously on days 1–3, prednisolone 60 mg/m2 orally on days 1–5, and thalidomide 200 mg orally once per day continuously (n=52); this regimen was associated with a significant improvement in the proportion of patients achieving a response, 2-year progression-free survival, and 2-year overall survival.4 Our study confirms the poor outcomes for patients with peripheral T-cell lymphoma in the setting of a prospective randomised trial.5, 25 Recruitment to the trial was closed early as there was strong evidence that the primary endpoint—to detect superiority of GEM-P over CHOP by a comparison of the CT-based complete responses and unconfirmed complete responses at end of treatment—would not be met, although this was not reflected in inferior progression-free survival or overall survival at 2 years in the GEM-P group. More patients also withdrew from the GEM-P group than from the CHOP group for reasons other than disease progression. The dose of cisplatin 100 mg/m2 administered in GEM-P was associated with more grade 1–2 tinnitus (compared with CHOP), which led three patients to withdraw early before completing study treatment; this dose therefore appears to be at the upper limit of what is tolerable in terms of ototoxicity, but further assessment of the dosing limit would be necessary to establish tolerability. Nevertheless, we have revised the dose of cisplatin administered with GEM-P at The Royal Marsden Hospital (London, UK) to a total dose of 75 mg/m2, in line with the dose used in other regimens containing gemcitabine and platinum,4, 31 with the expectation that this might reduce the incidence of ototoxicity.

lerability. Nevertheless, we have revised the dose of cisplatin administered with GEM-P at The Royal Marsden Hospital (London, UK) to a total dose of 75 mg/m2, in line with the dose used in other regimens containing gemcitabine and platinum,4, 31 with the expectation that this might reduce the incidence of ototoxicity. Although our data show that GEM-P has efficacy in terms of response and survival in peripheral T-cell lymphoma, in our randomised study it was inferior to CHOP for treatment-naive patients with this disease. One possible exception for the use of this type of regimen upfront might be when front-line anthracycline-based chemotherapy is contraindicated to avoid cardiotoxicity. In conclusion, although further studies are warranted, our phase 2 randomised trial suggests that CHOP should, for the time being, remain the reference regimen for previously untreated patients with peripheral T-cell lymphoma and that GEM-P is best reserved for the relapsed and refractory setting at present. Supplementary Material Supplementary appendix

Although our data show that GEM-P has efficacy in terms of response and survival in peripheral T-cell lymphoma, in our randomised study it was inferior to CHOP for treatment-naive patients with this disease. One possible exception for the use of this type of regimen upfront might be when front-line anthracycline-based chemotherapy is contraindicated to avoid cardiotoxicity. In conclusion, although further studies are warranted, our phase 2 randomised trial suggests that CHOP should, for the time being, remain the reference regimen for previously untreated patients with peripheral T-cell lymphoma and that GEM-P is best reserved for the relapsed and refractory setting at present. Supplementary Material Supplementary appendix Acknowledgments Bloodwise provided funding (grant reference number 12070) and Cancer Research UK provided endorsement for this trial. Site accreditation, data collation, and quality control for 18F-FDG-PET-CT imaging was done by Lucy Pike on behalf of the UK PET Core lab (London, UK). The UK National Health Service provided funding to the UK National Institute for Health Research (NIHR) Biomedical Research Centres at The Royal Marsden Hospital/Institute of Cancer Research (London, UK). GPC acknowledges support from the Blood Theme of the Oxford National Institute for Health Research Biomedical Research Centre (Oxford, UK).

ervice provided funding to the UK National Institute for Health Research (NIHR) Biomedical Research Centres at The Royal Marsden Hospital/Institute of Cancer Research (London, UK). GPC acknowledges support from the Blood Theme of the Oxford National Institute for Health Research Biomedical Research Centre (Oxford, UK). Contributors DC designed the study, had study oversight, and contributed to data interpretation, and writing and approval of the report. MG gathered, analysed, and interpreted the data, did the literature searches, and wrote the report. CP designed the study, analysed and interpreted the data, produced figures, and wrote the report. AW and ADA did the central histopathological review. IZ and BS did the central CT response assessment and SC did the central 18F-FDG-PET-CT response assessments. YMT, LE, and JO gathered, interpreted, and analysed the data. RB coordinated the collection of biological specimens. IC, PJ, KMA, EAH, MPM, GPC, JR, AF, AH, SM, PM, KB, NM, NK, YH, and DT gathered and interpreted data. All authors reviewed and approved the final version of the report.

F-FDG-PET-CT response assessments. YMT, LE, and JO gathered, interpreted, and analysed the data. RB coordinated the collection of biological specimens. IC, PJ, KMA, EAH, MPM, GPC, JR, AF, AH, SM, PM, KB, NM, NK, YH, and DT gathered and interpreted data. All authors reviewed and approved the final version of the report. Declaration of interests IC has received research funding from Eli-Lilly, Janssen-Cilag, Sanofi Oncology, Merck Serono, and Novartis; participated on advisory boards for Sanofi Oncology, Eli-Lilly, Bristol-Myers Squibb, Merck Sharpe Dohme (MSD), Bayer, Roche, and Five Prime Therapeutics; and received honoraria from Taiho, Pfizer, Amgen, and Eli-Lilly. EAH has received research funding from Bristol-Myers Squibb, Celgene, Merck-Serono, and MSD; participated on advisory boards for Celgene and Janssen; and received speaker fees and travel expenses from Bristol-Myers Squibb, Roche, and Janssen and travel expenses from Takeda. SM has received honoraria from Roche and travel support from Gilead. NM has received conference support and participated on advisory boards for Roche. DC has received research funding from Amgen, AstraZeneca, Bayer, Celgene, Medimmune, Merck Serono, Merrimack, and Sanofi. All other authors declare no competing interests.

Introduction Venous thromboembolism (deep venous thrombosis and pulmonary embolism) is the most common preventable cause of hospital death.1 Prevention of hospital-acquired thrombosis is a major health service focus and ranks as the number one hospital strategy for patient safety improvement worldwide.2 People with cancer are at particular risk of venous thromboembolism and clinical guidelines recommend pharmacological thromboprophylaxis for all patients with cancer if hospitalised with acute illness.3, 4 The presence of cancer is an independent risk factor for venous thromboembolism that varies according to primary tumour, stage, and associated cancer-modifying treatments. However, guideline recommendations are extrapolated from thromboprophylaxis trials not done specifically in cancer cohorts and do not consider varying thrombogenicity across the cancer population, particularly as the cancer progresses.5 Patients with advanced cancer and a life expectancy of less than 3 months were also excluded systematically from these studies.

polated from thromboprophylaxis trials not done specifically in cancer cohorts and do not consider varying thrombogenicity across the cancer population, particularly as the cancer progresses.5 Patients with advanced cancer and a life expectancy of less than 3 months were also excluded systematically from these studies. Most people with advanced, incurable cancer will be admitted to hospital where they will receive thrombo-prophylaxis routinely.6 Only a small proportion will be admitted to a specialist palliative care unit (SPCU), where thromboprophylaxis is a matter of debate, the primary focus of palliative care being symptom control, not survival.7 There is a belief that venous thromboembolism is uncommon in the palliative care setting, that data supporting primary thromboprophylaxis are extrapolated from unrepresentative populations (ie, those with prognosis >3 months), and that outcomes from thromboprophylaxis studies (such as radiologically apparent venous thromboembolism), without consideration of symptom impact, are less relevant to people with advanced cancer.8, 9 Venous thromboembolism in the SPCU setting is considered of clinical relevance only if it confers a patient-reported symptom burden or contributes to distressing symptoms at the end of life. Around the world, few SPCUs (whether in hospital or hospice settings) practice routine thromboprophylaxis.10, 11, 12, 13 In the UK, most hospice SPCUs are independent from the National Health Service and therefore lie outside national patient safety initiatives.14 Research in context Evidence before this study

Most people with advanced, incurable cancer will be admitted to hospital where they will receive thrombo-prophylaxis routinely.6 Only a small proportion will be admitted to a specialist palliative care unit (SPCU), where thromboprophylaxis is a matter of debate, the primary focus of palliative care being symptom control, not survival.7 There is a belief that venous thromboembolism is uncommon in the palliative care setting, that data supporting primary thromboprophylaxis are extrapolated from unrepresentative populations (ie, those with prognosis >3 months), and that outcomes from thromboprophylaxis studies (such as radiologically apparent venous thromboembolism), without consideration of symptom impact, are less relevant to people with advanced cancer.8, 9 Venous thromboembolism in the SPCU setting is considered of clinical relevance only if it confers a patient-reported symptom burden or contributes to distressing symptoms at the end of life. Around the world, few SPCUs (whether in hospital or hospice settings) practice routine thromboprophylaxis.10, 11, 12, 13 In the UK, most hospice SPCUs are independent from the National Health Service and therefore lie outside national patient safety initiatives.14 Research in context Evidence before this study We searched PubMed for international and national clinical guidelines for the prevention of venous thromboembolism in patients with cancer published between Jan 1, 1980, and Dec 31, 2017, with no language restrictions. Excluding updates, of nine published clinical guidelines (one international and eight national), only one specifically addressed patients receiving palliative care with guidance based on level 5 evidence (grade D recommendation). The UK guideline (National Institute for Health and Clinical Excellence, clinical guideline CG92) is recommended consideration of thromboprophylaxis for potentially reversible causes of increased risk of venous thromboembolism unless the patient was in palliative care. We then searched PubMed for studies published in English between the same dates, with the terms “venous thromboembolism” AND “thromboprophylaxis” or “prophylactic” AND “palliative” OR “advanced cancer” AND “hospice” OR “inpatient”. We identified one randomised controlled trial comparing thromboprophylaxis with the low-molecular-weight heparin (LMWH) nadroparin versus no thromboprophylaxis in patients with advanced cancer in a specialist palliative care unit, but the study recruited only 20 patients. Three clinician surveys identified an inconsistent approach to thromboprophylaxis in specialist palliative care units (SPCUs) and showed that in the acute setting, people with advanced disease are managed differently according to clinical specialty. Before our study, the true prevalence of clinically relevant deep vein thrombosis and its natural history in patients with advanced cancer were unknown.

in specialist palliative care units (SPCUs) and showed that in the acute setting, people with advanced disease are managed differently according to clinical specialty. Before our study, the true prevalence of clinically relevant deep vein thrombosis and its natural history in patients with advanced cancer were unknown. Added value of this study Our findings showed that approximately one in three people with advanced incurable disease admitted to an SPCU had a femoral deep vein thrombosis, but the he incidence of new thrombosis during the 3 week follow-up was low. Previous venous thromboembolism and being bedbound in the previous 3 months independently predicted deep vein thrombosis. We found no statistically significant association between deep vein thrombosis on admission and survival. Leg oedema was the only venous thromboembolism-relevant sign or symptom associated with deep vein thrombosis. Serum albumin concentration or use of thromboprophylaxis were not related to presence of deep vein thrombosis. Implications of all the available evidence

Our findings showed that approximately one in three people with advanced incurable disease admitted to an SPCU had a femoral deep vein thrombosis, but the he incidence of new thrombosis during the 3 week follow-up was low. Previous venous thromboembolism and being bedbound in the previous 3 months independently predicted deep vein thrombosis. We found no statistically significant association between deep vein thrombosis on admission and survival. Leg oedema was the only venous thromboembolism-relevant sign or symptom associated with deep vein thrombosis. Serum albumin concentration or use of thromboprophylaxis were not related to presence of deep vein thrombosis. Implications of all the available evidence The high prevalence but low 2-week incidence of femoral deep vein thrombosis in people with advanced cancer on SPCU admission suggests thromboprophylaxis at this stage might be too late. The absence of observed association with survival, or symptoms or signs other than leg oedema questions whether thromboprophylaxis offers clinically meaningful benefit. Our data challenge current international thromboprophylaxis guidelines on a number of counts. First, the findings suggest that the hospital model of care in which patients are risk assessed and given prophylaxis upon admission might be inappropriate for those with advanced cancer and a poor performance status before admission. Second, they raise questions about the optimal timing for introduction of thromboprophylaxis; should it be earlier in advanced disease? Recent work has shown no survival benefit with LMWH prophylaxis in newly diagnosed lung cancer or high-dose LMWH prophylaxis in pancreatic cancer, although high-dose LMWH prophylaxis prevented fatal pulmonary emboli. Third, is thromboprophylaxis of any clinical benefit at all in advanced disease? Is venous thromboembolism merely another manifestation of the inflammatory state of advanced disease, known to be associated with worse survival, and which does the greater damage at this stage of disease?

prevented fatal pulmonary emboli. Third, is thromboprophylaxis of any clinical benefit at all in advanced disease? Is venous thromboembolism merely another manifestation of the inflammatory state of advanced disease, known to be associated with worse survival, and which does the greater damage at this stage of disease? The provision of thromboprophylaxis for patients receiving palliative care might therefore be determined by place of admission rather than clinical risk. It is unknown whether current practice represents over-treatment (hospital) or under-treatment (SPCU), with hospital patients exposed to the risks of anticoagulation when none is needed, or SPCU patients exposed to the risks of symptomatic venous thromboembolism. We therefore aimed to evaluate the true prevalence of proximal deep vein thrombosis, diagnosed by systematic venous compression ultrasound, in people with advanced, incurable cancer admitted to an SPCU. Secondary outcomes included incidence during admission, associated factors (thromboprophylaxis), and clinical outcomes (symptoms or signs of venous thromboembolism and survival).

l deep vein thrombosis, diagnosed by systematic venous compression ultrasound, in people with advanced, incurable cancer admitted to an SPCU. Secondary outcomes included incidence during admission, associated factors (thromboprophylaxis), and clinical outcomes (symptoms or signs of venous thromboembolism and survival). Methods Study design and participants We did a prospective, multicentre, longitudinal, observational prevalence study. Participants were enrolled between June 20, 2016, and Oct 16, 2017. Institutional and ethical (Yorkshire and the Humber–Leeds West Research Ethics Committee) approvals, including for method of consent and management of ultrasound scan results, were granted before recruitment. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. The protocol allowed inclusion of people with advanced non-malignant disease as an exploratory substudy. We report the cancer objectives here only; data relating to people with non-malignant disease are available elsewhere.

ith the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. The protocol allowed inclusion of people with advanced non-malignant disease as an exploratory substudy. We report the cancer objectives here only; data relating to people with non-malignant disease are available elsewhere. Eligible patients were consecutive adults with cancer, aged 18 years or older, admitted to one of five SPCUs ((four hospices and one palliative care unit) in England (n=1), Wales (n=1), and Northern Ireland (n=3; appendix), who were able to give fully informed written consent or, in the absence of mental capacity to provide consent, an appropriate consultee to provide written agreement, and had no physical impediment to femoral vein ultrasound examination (eg, fixed flexion of the hip). There was no upper age limit to participation. If consultee agreement was used, retrospective consent for use of collected data was sought if the participant regained capacity. Patients with a clinician-estimated prognosis of 5 days or less, insufficient mental capacity and no appropriate consultee, or insufficient English or Welsh to provide consent or to comply with study assessments were excluded.

rospective consent for use of collected data was sought if the participant regained capacity. Patients with a clinician-estimated prognosis of 5 days or less, insufficient mental capacity and no appropriate consultee, or insufficient English or Welsh to provide consent or to comply with study assessments were excluded. Eligible patients were invited to participate by the admitting clinician. Participants had baseline assessments performed by a research nurse within 48 h of admission including participant demographics, clinical characteristics, venous thromboembolism history, Wells' score,15 and blood tests available from routine care. Study assessments were done at baseline and then weekly until discharge or death for a maximum of 3 weeks.

nts performed by a research nurse within 48 h of admission including participant demographics, clinical characteristics, venous thromboembolism history, Wells' score,15 and blood tests available from routine care. Study assessments were done at baseline and then weekly until discharge or death for a maximum of 3 weeks. Procedures Study outcome measures at baseline were bedside femoral and popliteal vein assessment by ultrasound; Australian-modified Karnofsky performance status (AKPS) score;16 clinical examination for signs or symptoms of venous thromboembolism (and new or worsening signs or symptoms at follow-up); known (previously confirmed) venous thromboembolism; bleeding; and medication record including anticoagulation. Noted signs and symptoms of venous thromboembolism were leg oedema and prominent veins; tenderness along the distribution of the deep venous system; calf swelling (circumference at least 3 cm greater than the other calf, measured 10 cm below tibial tuberosity); and pleuritic chest pain or breathlessness. Bleeding was categorised as major or clinically relevant non-major.17 Participants were considered at high risk of bleeding if they had thrombocytopenia (platelets <50 × 109 per L), international normalised ratio greater than 1·3, known active gastric or duodenal ulcer, known cerebral metastases, severe and uncontrolled hypertension, renal impairment (creatinine clearance <20 mL/min), or severe liver impairment.

red at high risk of bleeding if they had thrombocytopenia (platelets <50 × 109 per L), international normalised ratio greater than 1·3, known active gastric or duodenal ulcer, known cerebral metastases, severe and uncontrolled hypertension, renal impairment (creatinine clearance <20 mL/min), or severe liver impairment. All measures were repeated weekly apart from the medication record. New or worsening clinical symptoms and signs were reported to the clinical team. Overall survival was measured using routinely collected clinical record data from date of study enrolment until date of death from any cause. Deaths recorded using routinely collected clinical record data until 6 weeks after the end of recruitment. Bilateral femoral and popliteal vein ultrasound scans were undertaken at the bedside by one of five trained research nurses, who was independent to the participant's clinical care. Training occurred at a 2-day ultrasonography course approvied by the Royal College of Physicians, comprising basic ultrasonography physics, practical teaching, and hands-on experience. Staff underwent practical and theory assessments and were required to complete a portfolio of scans before sign-off. Specific to the study, trial nurses had a further day's focused training to optimise skills. The scan process involved identification of the common femoral vein and a compressibility assessment performed at 2 cm increments to the level of the popliteal fossa.

e required to complete a portfolio of scans before sign-off. Specific to the study, trial nurses had a further day's focused training to optimise skills. The scan process involved identification of the common femoral vein and a compressibility assessment performed at 2 cm increments to the level of the popliteal fossa. Outcomes The primary endpoint of the study was the prevalence of deep vein thrombosis within 48 h of SPCU admission. Secondary endpoints were associated symptoms attributable to deep vein thrombosis, 3-week incidence of new deep thrombosis during admission (with associated symptoms), clinical characteristics associated with the presence of deep vein thrombosis, association between use of anticoagulation and presence of deep vein thrombosis on, and during, admission to an SPCU, impact of deep vein thrombosis on length of stay, and overall survival. With respect to the primary endpoint, one of three possible outcomes was recorded for each scan: no deep vein thrombosis (vein compressible throughout), deep vein thrombosis (vein not compressible at any point), or unevaluable. Further training was provided 3 months into the study to optimise compression technique, improve image quality, and increase the number of evaluable images. We categorised scans done between June 20, and Sept 30, 2016, as early study scans and those done between Oct 1, 2016, and Oct 16, 2017, following the further training of study nurses in bedside ultrasound, as later study scans. All scans were digitally recorded and reviewed by the study radiologist (EN), who was the final arbiter of the presence of deep vein thrombosis or no deep vein thrombosis.

ns and those done between Oct 1, 2016, and Oct 16, 2017, following the further training of study nurses in bedside ultrasound, as later study scans. All scans were digitally recorded and reviewed by the study radiologist (EN), who was the final arbiter of the presence of deep vein thrombosis or no deep vein thrombosis. Because screening for deep vein thrombosis is not routinely undertaken in the SPCU, participants and clinicians were masked to ultrasound findings. However, the scan result could be given on request to the treating clinician if there was a clinical suspicion of deep vein thrombosis and the scan had been undertaken within the previous 24 h. We report the cancer objectives here only; data relating to people with non-malignant disease are available elsewhere. Statistical analysis The sample size calculation was based on a previous study showing bilateral obstruction of venous return in the legs in 17% of SPCU inpatients.18 Assuming bilateral obstruction represented more extensive thrombosis, we calculated that a sample size of 217 patients with cancer was needed to estimate the prevalence of proximal lower limb deep vein thrombosis (5% precision; 95% confidence level). Recruitment beyond the sample size for the primary outcome was permitted to improve precision for secondary outcomes.

extensive thrombosis, we calculated that a sample size of 217 patients with cancer was needed to estimate the prevalence of proximal lower limb deep vein thrombosis (5% precision; 95% confidence level). Recruitment beyond the sample size for the primary outcome was permitted to improve precision for secondary outcomes. Participant characteristics are summarised using descriptive analyses with mean (SD; range) or number (%), as appropriate. Prevalence (within 48 h of SPCU admission) is expressed as a percentage with associated 95% CI. Participants with evaluable data for the baseline scan were included in our primary analysis, and the choice of the analysis population was not prespecified. Univariable logistic regression models were performed to create odds ratios (ORs) and 95% CIs for the following risk factors: age, sex, baseline venous thromboembolism risk factors, use of anticoagulants, AKPS score, venous thromboembolism history, bleeding history, and bleeding risk. All these variables were entered in a multivariable logistic regression model using backward selection with a retention criterion of p value of less than 0·05. Adjusted ORs with 95% CIs were calculated. Missing data were not imputed. Participants with evaluable data for the baseline scan were included in these analyses.

ll these variables were entered in a multivariable logistic regression model using backward selection with a retention criterion of p value of less than 0·05. Adjusted ORs with 95% CIs were calculated. Missing data were not imputed. Participants with evaluable data for the baseline scan were included in these analyses. A Kaplan Meier curve was used to compare survival and prevalence of proximal lower limb deep vein thrombosis within 48 h after the patient's admission to SPCU. A log-rank test was used to test for statistical significance. Participants who had not died by the end of survival data collection were right censored. We did post-hoc sensitivity analyses excluding early scans (those done between June and September, 2016), to account for a technical learning curve, and excluding patients with previous history of deep vein thrombosis. Analyses were done with SPSS version 25.0. This study is registered with the ISRCTN registry, number ISRCTN97567719. 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. MJJ, VLA, and FS had access to all the raw data in the study. EN had access to all scan images. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

n, data collection, data analysis, data interpretation, or writing of the report. MJJ, VLA, and FS had access to all the raw data in the study. EN had access to all scan images. 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 June 20, 2016, and Oct 16, 2017, 1390 patients were screened, of whom 343 were recruited (figure 1). 11 participants were admitted more than once during the study period; only the first admission in the study period (n=343) was used for analysis. There was no loss to follow-up.Figure 1 Flow diagram DVT=deep vein thrombosis.

Results Between June 20, 2016, and Oct 16, 2017, 1390 patients were screened, of whom 343 were recruited (figure 1). 11 participants were admitted more than once during the study period; only the first admission in the study period (n=343) was used for analysis. There was no loss to follow-up.Figure 1 Flow diagram DVT=deep vein thrombosis. The most common primary tumour site was lung (20%), followed by upper gastrointestinal, hepatobiliary, or pancreatic (18%), and colorectal (16%) cancer. Most participants (84%) had metastatic disease and 80% had at least one comorbidity (table 1). Almost a quarter (77 [22%] patients) had a history of venous thromboembolism (deep vein thrombosis, 36 [10%]; pulmonary embolism, 55 [16%]). A quarter of participants (82 [24%]) were receiving low-molecular-weight heparin (LMWH) thromboprophylaxis (11 [4%] were on a direct oral anticoagulant; nine [3%] were taking warfarin) and ten (3%) had antithromboembolism stockings alone. 40 (12%) participants were receiving full treatment doses of anticoagulation. Relevant symptoms at baseline included breathlessness (177 [52%] patients), leg oedema (left leg, 147 [43%]; right leg, 137 [40%]), leg pain (left leg, 65 [19%]; right leg, 66 [19%]), haemoptysis (23 [7%]), and chest pain (65 [19%]). Wells' deep vein thrombosis score was likely (≥2) in 174 (51%) participants.Table 1 Baseline characteristics

aseline included breathlessness (177 [52%] patients), leg oedema (left leg, 147 [43%]; right leg, 137 [40%]), leg pain (left leg, 65 [19%]; right leg, 66 [19%]), haemoptysis (23 [7%]), and chest pain (65 [19%]). Wells' deep vein thrombosis score was likely (≥2) in 174 (51%) participants.Table 1 Baseline characteristics Total population (n=343) Age (years) 68·2 (12·8; 25–102) Sex Male 179 (52%) Female 164 (48%) Family history of venous thromboembolism Yes 49 (14%) No or unknown 294 (86%) Primary cancer Lung 70 (20%) Upper gastrointestinal, hepatobiliary, or pancreatic 61 (18%) Colorectal 55 (16%) Prostate 28 (8%) Breast 26 (8%) Gynaecological 25 (7%) Head and neck 22 (6%) Urological 19 (6%) Unknown primary 16 (5%) Haematological 10 (3%) Brain 6 (2%) Skin 3 (1%) Bone 2 (1%) Metastatic disease None 56 (16%) Yes 287 (84%) Number of sites of metastases 1 98 (28%) 2 92 (27%) 3 65 (19%) 4 23 (7%) 5 7 (2%) 6 2 (1%) Comorbidities None 67 (20%) Cardiovascular 141 (41%) Gastrointestinal or hepatorenal 90 (27%) Respiratory 69 (20%) Neurological 48 (14%) Diabetes 45 (13%) Musculoskeletal 45 (13%) AKPS score 49 (16·6, 20–90) Data are mean (SD; range) or n (%). AKPS=Australian-modified Karnofsky performance status. Ultrasound scans were done in 343 patients on admission. According to the radiologist's assessment, 92 (27%) scans showed deep vein thrombosis and 181 (53%) showed no deep vein thrombosis. 52 (15%) scans were not evaluable, and 18 (5%) were missing. Of 273 patients with evaluable scans and available data, 92 (34%, 95% CI 28–40) had femoral vein thrombosis.

ents on admission. According to the radiologist's assessment, 92 (27%) scans showed deep vein thrombosis and 181 (53%) showed no deep vein thrombosis. 52 (15%) scans were not evaluable, and 18 (5%) were missing. Of 273 patients with evaluable scans and available data, 92 (34%, 95% CI 28–40) had femoral vein thrombosis. Of the 273 patients with evaluable scans, 28 had a previous history of deep vein thrombosis (data missing for one patient). Of the 343 patients with cancer, 290 had at least one follow-up scan with a definitive evaluation; 41 of these scans were unevaluable and 12 were done, but the data were missing. Four participants with a scan showing no deep vein thrombosis on admission had a deep vein thrombosis identified subsequently during their admission (maximum 3-weeks' follow-up, figure 1). A further eight participants with unknown deep vein thrombosis status at baseline (because of missing or unevaluable scans) had a deep vein thrombosis.

th a scan showing no deep vein thrombosis on admission had a deep vein thrombosis identified subsequently during their admission (maximum 3-weeks' follow-up, figure 1). A further eight participants with unknown deep vein thrombosis status at baseline (because of missing or unevaluable scans) had a deep vein thrombosis. Table 2 shows the univariable and multivariable regression analyses. Previous venous thromboembolism, being bedbound in the past 12 weeks, and lower limb oedema independently predicted the presence of deep vein thrombosis in the final multivariable model. We noted no association between the use of thromboprophylaxis and the presence of deep vein thrombosis on admission. We found no association between serum albumin concentration and the presence of deep vein thrombosis (mean serum albumin concentration 31·4 mmol/L [SD 6·6] in patients with deep vein thrombosis vs 30·6 mmol/L [5·7] in patients without deep vein thrombosis; OR 0·98, 95% CI 0·93–1·03; p=0·43). Mean average survival was 30·55 days (SD 5·65) for patients with deep vein thrombosis versus 31·38 days (6·56) for those without deep vein thrombosis (p=0·432). The presence of deep vein thrombosis on admission was not related to survival (hazard ratio [HR] 1·102 (95% CI 0·842–1·441; p=0·45; figure 2).Table 2 Univariable and multivariable logistic regression analysis

r patients with deep vein thrombosis versus 31·38 days (6·56) for those without deep vein thrombosis (p=0·432). The presence of deep vein thrombosis on admission was not related to survival (hazard ratio [HR] 1·102 (95% CI 0·842–1·441; p=0·45; figure 2).Table 2 Univariable and multivariable logistic regression analysis No deep vein thrombosis (n=181) Deep vein thrombosis (n=92) Univariable analysis Multivariable analysis Odds ratio (95% CI) p value Odds ratio (95% CI) p value Demographic Age (years) 68·4 (12·8; 25–102) 68·1 (13·5; 29–95) 0·99 (0·98–1·02) 0·86 .. .. Sex Male 95/181 (52%) 50/92 (54%) 1·08 (0·65–1·78) 0·77 .. .. Female 86/181 (48%) 42/92 (46%) 1 (ref) .. .. .. Smoking history Current smoker 25/181 (14%) 15/92 (16%) 1·24 (0·58–2·64) 0·85 .. .. Ex-smoker 86/181 (48%) 43/92 (47%) 1·03 (0·59–1·78) .. .. .. Never smoked 70/181 (39%) 34/92 (37%) 1 (ref) .. .. .. AKPS score 50·0 (15·8; 20–90) 45·2 (16·7; 20–90) 0·98 (0·97–0·99) 0·022 .. .. Medical history Family history of venous thromboembolism Yes 22/178 (12%) 13/89 (15%) 1·21 (0·58–2·54) 0·61 .. .. No 156/178 (88%) 76/89 (85%) 1 (ref) .. .. .. Previous pulmonary embolus Yes 30/181 (17%) 15/92 (16%) 0·98 (0·50–1·93) 0·96 .. .. No 151/181 (83%) 77/92 (84%) 1 (ref) .. .. .. Previous deep vein thrombosis Yes 12/181 (7%) 16/91 (18%) 3·00 (1·36–6·66) 0·007 3·00 (1·28–7·00) 0·011 No 169/181 (93%) 75/91 (82%) 1 (ref) .. 1 (ref) .. Previous arterial thrombosis Yes 6/181 (3%) 3/91 (3%) 0·99 (0·24–4·07) 0·99 .. .. No 175/181 (97%) 88/91 (97%) 1 (ref) .. .. .. Previous venous thromboembolism Yes 33/181 (18%) 29/92 (32%) 2·06 (1·16–3·69) 0·014 2·06 (1·16–3·69) 0·014 No 148/181 (82%) 63/92 (68%) 1 (ref) .. 1 (ref) .. Deep vein thrombosis risk factors within past 12 weeks Acute medical illness Yes 54/181 (30%) 40/92 (43%) 1·81 (1·08–3·05) 0·026 .. .. No 127/181 (70%) 52/92 (57%) 1 (ref) .. .. .. Surgery Yes 16/181 (9%) 10/92 (11%) 1·26 (0·55–2·89) 0·59 .. .. No 165/181 (91%) 82/92 (89%) 1 (ref) .. .. .. Bedbound in past 12 weeks Yes 27/181 (15%) 27/92 (29%) 2·37 (1·29–4·35) 0·005 2·66 (1·38–5·10) 0·003 No 154/181 (85%) 65/92 (71%) 1 (ref) .. 1 (ref) .. Bleeding within past 6 months Major Yes 7/181 (4%) 2/92 (2%) 0·55 (0·11–2·71) 0·46 .. .. No 174/181 (96%) 90/92 (98%) 1 (ref) .. .. .. Non-major Yes 15/181 (8%) 4/92 (4%) 0·50 (0·16–1·56) 0·23 .. .. No 166/181 (92%) 88/92 (96%) 1 (ref) .. .. .. Any Yes 22/181 (12%) 6/92 (7%) 0·50 (0·20–1·30) 0·15 .. .. No 159/181 (88%) 86/92 (93%) 1 (ref) .. .. ..

Major Yes 7/181 (4%) 2/92 (2%) 0·55 (0·11–2·71) 0·46 .. .. No 174/181 (96%) 90/92 (98%) 1 (ref) .. .. .. Non-major Yes 15/181 (8%) 4/92 (4%) 0·50 (0·16–1·56) 0·23 .. .. No 166/181 (92%) 88/92 (96%) 1 (ref) .. .. .. Any Yes 22/181 (12%) 6/92 (7%) 0·50 (0·20–1·30) 0·15 .. .. No 159/181 (88%) 86/92 (93%) 1 (ref) .. .. .. High risk of bleeding Yes 33/181 (18%) 19/92 (21%) 1·17 (0·62–2·19) 0·63 .. .. No 148/181 (82%) 73/92 (79%) 1 (ref) .. .. .. Venous thromboembolism signs and symptoms at baseline Lower limb oedema (either) Yes 64/180 (36%) 49/92 (53%) 2·07 (1·24–3·44) 0·005 2·08 (1·20–3·60) 0·009 No 116/180 (64%) 43/92 (47%) 1 (ref) .. 1 (ref) .. Lower limb pain Yes 33/180 (18%) 23/92 (25%) 1·48 (0·81–2·71) 0·20 .. .. No 147/180 (82%) 69/92 (75%) 1 (ref) .. .. .. Chest pain Yes 36/180 (20%) 15/92 (16%) 0·75 (0·39–1·47) 0·41 .. .. No 144/180 (80%) 77/92 (84%) 1 (ref) .. .. .. Breathlessness Yes 96/180 (53%) 45/92 (49%) 0·87 (0·52–1·43) 0·57 .. .. No 84/180 (47%) 47/92 (51%) 1 (ref) .. .. .. Haemoptysis Yes 13/172 (8%) 4/90 (4%) 0·53 (0·17–1·65) 0·27 .. .. No 159/172 (92%) 86/90 (96%) 1 (ref) .. .. .. Thromboprophylaxis Anticoagulation Yes 43/165 (26%) 26/75 (35%) 1·51 (0·84–2·71) 0·17 .. .. No 122/165 (74%) 49/75 (65%) 1 (ref) .. .. .. Antithromboembolism stockings Yes 6/169 (4%) 1/80 (1%) 0·34 (0·04–2·91) 0·32 .. .. No 163/169 (96%) 79/80 (99%) 1 (ref) .. .. .. Data are mean (SD; range) or n/N (%), unless otherwise stated. AKPS=Australian-modified Karnofsky performance status. Figure 2 Overall survival in patients with and without deep vein thrombosis on admission

High risk of bleeding Yes 33/181 (18%) 19/92 (21%) 1·17 (0·62–2·19) 0·63 .. .. No 148/181 (82%) 73/92 (79%) 1 (ref) .. .. .. Venous thromboembolism signs and symptoms at baseline Lower limb oedema (either) Yes 64/180 (36%) 49/92 (53%) 2·07 (1·24–3·44) 0·005 2·08 (1·20–3·60) 0·009 No 116/180 (64%) 43/92 (47%) 1 (ref) .. 1 (ref) .. Lower limb pain Yes 33/180 (18%) 23/92 (25%) 1·48 (0·81–2·71) 0·20 .. .. No 147/180 (82%) 69/92 (75%) 1 (ref) .. .. .. Chest pain Yes 36/180 (20%) 15/92 (16%) 0·75 (0·39–1·47) 0·41 .. .. No 144/180 (80%) 77/92 (84%) 1 (ref) .. .. .. Breathlessness Yes 96/180 (53%) 45/92 (49%) 0·87 (0·52–1·43) 0·57 .. .. No 84/180 (47%) 47/92 (51%) 1 (ref) .. .. .. Haemoptysis Yes 13/172 (8%) 4/90 (4%) 0·53 (0·17–1·65) 0·27 .. .. No 159/172 (92%) 86/90 (96%) 1 (ref) .. .. .. Thromboprophylaxis Anticoagulation Yes 43/165 (26%) 26/75 (35%) 1·51 (0·84–2·71) 0·17 .. .. No 122/165 (74%) 49/75 (65%) 1 (ref) .. .. .. Antithromboembolism stockings Yes 6/169 (4%) 1/80 (1%) 0·34 (0·04–2·91) 0·32 .. .. No 163/169 (96%) 79/80 (99%) 1 (ref) .. .. .. Data are mean (SD; range) or n/N (%), unless otherwise stated. AKPS=Australian-modified Karnofsky performance status. Figure 2 Overall survival in patients with and without deep vein thrombosis on admission Shaded areas represent 95% CIs. HR=hazard ratio.

High risk of bleeding Yes 33/181 (18%) 19/92 (21%) 1·17 (0·62–2·19) 0·63 .. .. No 148/181 (82%) 73/92 (79%) 1 (ref) .. .. .. Venous thromboembolism signs and symptoms at baseline Lower limb oedema (either) Yes 64/180 (36%) 49/92 (53%) 2·07 (1·24–3·44) 0·005 2·08 (1·20–3·60) 0·009 No 116/180 (64%) 43/92 (47%) 1 (ref) .. 1 (ref) .. Lower limb pain Yes 33/180 (18%) 23/92 (25%) 1·48 (0·81–2·71) 0·20 .. .. No 147/180 (82%) 69/92 (75%) 1 (ref) .. .. .. Chest pain Yes 36/180 (20%) 15/92 (16%) 0·75 (0·39–1·47) 0·41 .. .. No 144/180 (80%) 77/92 (84%) 1 (ref) .. .. .. Breathlessness Yes 96/180 (53%) 45/92 (49%) 0·87 (0·52–1·43) 0·57 .. .. No 84/180 (47%) 47/92 (51%) 1 (ref) .. .. .. Haemoptysis Yes 13/172 (8%) 4/90 (4%) 0·53 (0·17–1·65) 0·27 .. .. No 159/172 (92%) 86/90 (96%) 1 (ref) .. .. .. Thromboprophylaxis Anticoagulation Yes 43/165 (26%) 26/75 (35%) 1·51 (0·84–2·71) 0·17 .. .. No 122/165 (74%) 49/75 (65%) 1 (ref) .. .. .. Antithromboembolism stockings Yes 6/169 (4%) 1/80 (1%) 0·34 (0·04–2·91) 0·32 .. .. No 163/169 (96%) 79/80 (99%) 1 (ref) .. .. .. Data are mean (SD; range) or n/N (%), unless otherwise stated. AKPS=Australian-modified Karnofsky performance status. Figure 2 Overall survival in patients with and without deep vein thrombosis on admission Shaded areas represent 95% CIs. HR=hazard ratio. The presence of leg oedema (either leg) was an independent predictor of deep vein thrombosis (table 2). When analysed by left leg and left leg deep vein thrombosis (table 3), and right leg and right leg deep vein thrombosis (table 4), only left leg oedema was associated with the presence of ipsilateral deep vein thrombosis. There was no association between left leg pain and left leg deep vein thrombosis, or for right leg pain and right leg deep vein thrombosis.Table 3 Patient-reported left leg venous thromboembolism signs and symptoms by deep vein thrombosis on admission

as associated with the presence of ipsilateral deep vein thrombosis. There was no association between left leg pain and left leg deep vein thrombosis, or for right leg pain and right leg deep vein thrombosis.Table 3 Patient-reported left leg venous thromboembolism signs and symptoms by deep vein thrombosis on admission No left leg deep vein thrombosis (n=255) Left leg deep vein thrombosis (n=66) Odds ratio (95% CI) p value Left leg oedema Yes 102 (40%) 35 (53%) 1·92 (1·07–3·43) 0·027 No 153 (60%) 31 (47%) 1 (ref) .. Left leg pain Yes 43 (17%) 16 (24%) 0·98 (0·48–2·01) 0·96 No 212 (83%) 50 (76%) 1 (ref) .. Table 4 Patient-reported right leg venous thromboembolism signs and symptoms by deep vein thrombosis on admission No right deep vein thrombosis (n=260) Right leg deep vein thrombosis (n=61) Odds ratio (95% CI) p value Right leg oedema Yes 95 (37%) 32 (52%) 1·17 (10·67–2·07) 0·57 No 165 (63%) 29 (48%) 1 (ref) .. Right leg pain Yes 48 (18%) 13 (21%) 0·72 (0·33–1·53) 0·39 No 212 (82%) 48 (79%) 1 (ref) ..

No left leg deep vein thrombosis (n=255) Left leg deep vein thrombosis (n=66) Odds ratio (95% CI) p value Left leg oedema Yes 102 (40%) 35 (53%) 1·92 (1·07–3·43) 0·027 No 153 (60%) 31 (47%) 1 (ref) .. Left leg pain Yes 43 (17%) 16 (24%) 0·98 (0·48–2·01) 0·96 No 212 (83%) 50 (76%) 1 (ref) .. Table 4 Patient-reported right leg venous thromboembolism signs and symptoms by deep vein thrombosis on admission No right deep vein thrombosis (n=260) Right leg deep vein thrombosis (n=61) Odds ratio (95% CI) p value Right leg oedema Yes 95 (37%) 32 (52%) 1·17 (10·67–2·07) 0·57 No 165 (63%) 29 (48%) 1 (ref) .. Right leg pain Yes 48 (18%) 13 (21%) 0·72 (0·33–1·53) 0·39 No 212 (82%) 48 (79%) 1 (ref) .. At week 1, there were 188 remaining participants of whom seven (4%) reported new lower limb oedema, two (1%) reported new breathlessness, and three (2%) reported new chest pain. Data were missing for 15 (7%). At week 2 there were 80 remaining participants of whom one (1%) reported new lower limb pain, and none reported new breathlessness or new chest pain. Data were missing for four (5%). At week 3 (n=31) one had new leg oedema and one had new breathlessness. In those with an evaluable scan at week 1 (n=96) and week 2 (n=35), there was no difference in new venous thromboembolism-attributable symptoms between those with and those without a deep vein thrombosis (at week 1, two [2%] oedema, two [2%] leg pain, zero breathlessness, and one [1%] chest pain in patients with deep vein thrombosis vs one [1%] oedema, three [5%] leg pain, zero breathlessness, and two [2%] chest pain in patients without deep vein thrombosis; at week 2, one [3%] oedema, zero leg pain, zero breathlessness, and zero chest pain in patients with deep vein thrombosis vs two [2%] oedema, zero leg pain, zero breathlessness, and zero chest pain in patients without deep vein thrombosis.

reathlessness, and two [2%] chest pain in patients without deep vein thrombosis; at week 2, one [3%] oedema, zero leg pain, zero breathlessness, and zero chest pain in patients with deep vein thrombosis vs two [2%] oedema, zero leg pain, zero breathlessness, and zero chest pain in patients without deep vein thrombosis. There were too many missing dates of SPCU discharge to give an estimate on length of stay. Of 68 early study scans (done between June 2016 and September 2016), 28 (41%) indicated a deep vein thrombosis, 13 (19%) no deep vein thrombosis, 21 (31%) could not be evaluated, and six (9%) were missing. Of 275 later study scans (done between October, 2016, and October, 2017), 64 (23%) indicated a deep vein thrombosis, 168 (61%) no deep vein thrombosis, 31 (11%) could not be evaluated, and 12 (4%) were missing. The difference between early and later study scans for detection of deep vein thrombosis was strongly significant (p<0·001), suggesting a learning curve for scanning. We therefore did a post-hoc sensitivity analysis that excluded the early scans. Of 232 later study scans with a definitive evaluation, 64 (28%, 95% CI 22–34) showed femoral deep vein thrombosis. In a further sensitivity analysis excluding patients with a past history of deep vein thrombosis (n=28), 75 of 244 patients (31%, 95% CI 25–37) had a femoral vein thrombosis. If early scans in this group were excluded, 52 of 209 patients (25%, 19–31) had a femoral vein thrombosis.

Of 68 early study scans (done between June 2016 and September 2016), 28 (41%) indicated a deep vein thrombosis, 13 (19%) no deep vein thrombosis, 21 (31%) could not be evaluated, and six (9%) were missing. Of 275 later study scans (done between October, 2016, and October, 2017), 64 (23%) indicated a deep vein thrombosis, 168 (61%) no deep vein thrombosis, 31 (11%) could not be evaluated, and 12 (4%) were missing. The difference between early and later study scans for detection of deep vein thrombosis was strongly significant (p<0·001), suggesting a learning curve for scanning. We therefore did a post-hoc sensitivity analysis that excluded the early scans. Of 232 later study scans with a definitive evaluation, 64 (28%, 95% CI 22–34) showed femoral deep vein thrombosis. In a further sensitivity analysis excluding patients with a past history of deep vein thrombosis (n=28), 75 of 244 patients (31%, 95% CI 25–37) had a femoral vein thrombosis. If early scans in this group were excluded, 52 of 209 patients (25%, 19–31) had a femoral vein thrombosis. Discussion In this prospective longitudinal observational study, bedside compression ultrasonography identified femoral deep vein thrombosis in about a third of eligible people with advanced cancer admitted to an SPCU, with post-hoc analyses suggesting apparently new diagnoses of deep vein thrombosis (ie, in patients with no history of deep vein thrombosis, using optimised scanning technique) in a quarter of patients. Although iliofemoral deep vein thrombosis indicates a large clot burden, our findings showed no difference in relevant symptoms between participants with or without deep vein thrombosis apart from an association with lower limb oedema. Participants with deep vein thrombosis were more likely to have a history of venous thromboembolism, or to have been bedbound for any reason during the previous 3 months, than were those without deep vein thrombosis. Notably, there was no association between the presence or absence of deep vein thrombosis and thromboprophylaxis use. No statistical association was seen between the presence of deep vein thrombosis and serum albumin concentration, despite previous studies suggesting such an association.19 However, the previous data were recorded in a much healthier population followed up for a mean 723 days. Our findings show no difference in survival between patients with or without deep vein thrombosis. The numbers of participants developing new deep vein thrombosis subsequent to admission was low and conferred no additional symptom burden, although these findings should be treated with caution in view of the small numbers of patients.

ow no difference in survival between patients with or without deep vein thrombosis. The numbers of participants developing new deep vein thrombosis subsequent to admission was low and conferred no additional symptom burden, although these findings should be treated with caution in view of the small numbers of patients. More than half of all screened patients were ineligible (61%), mainly because death was expected within 5 days. Of the 549 eligible patients, 206 (38%) declined participation. Ethical approval did not permit the recording of demographics of non-consenting patients, although trial nurses reported that the population was similar to patients who consented. However, enrolled participants had similar demographics to those reported in the National Council for Palliative Care Minimum Data Set for SPC inpatient units with respect to sex, age, primary diagnosis, and metastatic burden.

ts, although trial nurses reported that the population was similar to patients who consented. However, enrolled participants had similar demographics to those reported in the National Council for Palliative Care Minimum Data Set for SPC inpatient units with respect to sex, age, primary diagnosis, and metastatic burden. Apart from limb oedema, our findings showed no effect of femoral vein thrombosis on experience of venous thromboembolism-related symptoms, and no effect of thromboprophylaxis on deep vein thrombosis risk. Thromboprophylaxis might therefore confer no benefit over analgesia or other appropriate control measures for deep vein thrombosis symptoms in patients with cancer admitted to an SPCU. Recruited participants had a mean AKPS score of 49 (a value of 50 represents the need for considerable assistance and frequent medical care),16 most had metastatic disease and at least one comorbidity, and mean length of survival was only 44 days. These characteristics not only demonstrate a functionally dependent population with advanced stage of illness but also one previously unrepresented in thromboprophylaxis studies, in which life expectancy of less than 3 months was invariably an exclusion criterion.

orbidity, and mean length of survival was only 44 days. These characteristics not only demonstrate a functionally dependent population with advanced stage of illness but also one previously unrepresented in thromboprophylaxis studies, in which life expectancy of less than 3 months was invariably an exclusion criterion. This was a pragmatic multicentre study done in SPCUs across the UK, with broad entry criteria. Clinical studies in this environment are notoriously challenging, yet HIDDen completed target recruitment ahead of time, allowing analysis of adequately powered data. Unlike previous venous thromboembolism studies, which have been difficult to contextualise in advanced cancer, we used pre-agreed outcome measures of relevance to patients and treating clinicians.8 To optimise recruitment, compression ultrasonography was done at the bedside by trained research nurses and independently validated by a consultant radiologist. Every effort was made to ensure the quality of scans and for this reason additional training was provided to the research nurses at month 3 of the study. Specifically, focus was placed on improved compression technique and increased number of images recorded for radiologist review. Furthermore, we did a sensitivity analysis excluding the first 3 months' scans to allow for a learning curve. However, we acknowledge that these results are likely to under-represent the prevalence of venous thromboembolism because the analysis will have omitted distal deep vein thrombosis and pulmonary embolus. In patients with cancer, distal deep vein thrombosis is associated with an increased risk of residual vein thrombosis, an established risk factor for recurrent deep vein thrombosis.20 Furthermore, identification of fresh deep vein thrombosis by compression ultrasonography in the presence of a recent deep vein thrombosis is notoriously challenging and it is for this reason that we omitted patients with history of deep vein thrombosis (proximal or distal) in our additional sensitivity analysis.21 Additionally, 5% of scans were missing. These were unlikely to be missing completely at random because participants, although they consented to participate, might then have declined a scan if they felt less well. This missing group might therefore have included at least some patients more likely to have a deep vein thrombosis because of poorer performance status, and thus the prevalence might be underestimated.

because participants, although they consented to participate, might then have declined a scan if they felt less well. This missing group might therefore have included at least some patients more likely to have a deep vein thrombosis because of poorer performance status, and thus the prevalence might be underestimated. This issue is greater for development of deep vein thrombosis during follow-up, where missing or unevaluable scans are greater. However, data were more complete for symptom report, and there were no between-group differences in new symptoms that could be attributable to thromboembolism. The study sample size was calculated to provide adequate power for estimates of prevalence and not for survival, and the data are observational. A randomised controlled trial of thromboprophylaxis with survival as the primary endpoint, using data from an epidemiological model of venous thromboembolism-related hospital-acquired deaths, would need a sample size of 72 000— unlikely to be feasible or to provide value for information in this patient population.22

l. A randomised controlled trial of thromboprophylaxis with survival as the primary endpoint, using data from an epidemiological model of venous thromboembolism-related hospital-acquired deaths, would need a sample size of 72 000— unlikely to be feasible or to provide value for information in this patient population.22 Our data challenge current recommendations for prevention of venous thromboembolism prevention in advanced cancer.4 Strategies to prevent hospital-acquired thrombosis remain a global health priority and patients with advanced cancer would seem an ideal patient group to target because they are highly thrombotic and 88% will have an average of five hospital admissions in their last year of life.23 A prospective observational study of 22 SPCUs in France identified a 9·8% (95% CI 8·3–11·6) incidence of clinically relevant bleeding in 1199 patients with cancer.24 Multivariate analysis suggested an association between clinically relevant bleeding and pharmacological thromboprophylaxis (HR 1·48, 95% CI 1·02–2·15; p=0·04).24 It would seem expedient to minimise the risk of harm by avoiding the use of thromboprophylaxis unless there will be a clear net benefit. Our data show a high prevalence of femoral deep vein thrombosis at the point of admission to the SPCU, making the issue of thromboprophylaxis a moot point. Since our data suggest these deep vein thromboses confer a minimal symptom burden with no evidence that they shorten life, rethinking the utility of pharmacological thromboprophylaxis in this population would seem reasonable. However, use of thromboprophylaxis is dependent upon place of admission, and is much less likely in SPCU inpatients than in those admitted to an acute hospital setting.

burden with no evidence that they shorten life, rethinking the utility of pharmacological thromboprophylaxis in this population would seem reasonable. However, use of thromboprophylaxis is dependent upon place of admission, and is much less likely in SPCU inpatients than in those admitted to an acute hospital setting. Whether these data can be extrapolated to the acute setting is debatable. The advanced cancer population in this study were defined by SPCU admission. The factors influencing the clinical decision to admit to the SPCU setting are unknown but will be more complex than extent of disease or prognosis, and are likely to include patient preferences, performance status, and clinician judgment regarding potential reversibility of any deterioration. Qualitative research offers some explanation for discrepant practice between hospitals and SPCUs. First, although around half of people admitted to SPCUs have evidence of obstruction to lower limb venous flow measured by light reflection rheography (17% bilateral)18 and 50% have venous thromboembolism at post mortem, palliative care teams perceive that venous thromboembolism is not a common clinical problem.9 Clinicians might misattribute symptoms to other pathologies such as lymphoedema, hypoalbuminaemia, or cellulitis (for deep vein thrombosis), and anaemia, pneumonia, pleural effusion, heart failure, lymphangitis, and lung metastases (for pulmonary embolism). Conversely, perhaps despite a high prevalence, venous thromboembolism-related symptoms contribute a small proportion to the overall symptom burden of people with advanced incurable illness. In our study, a significant predictor for deep vein thrombosis was being bedbound for any reason within the past 12 weeks. This characteristic not only delineates a subpopulation of people with advanced disease with a potentially poorer prognosis, but also suggests that their thrombotic insult might have been experienced before admission as part of an inevitable decline rather than the cause of it. This suggestion raises the question whether thromboprophylaxis should be started before developing advanced disease. However, LMWH prophylaxis in patients with newly diagnosed lung cancer25 or high-dose LMWH prophylaxis in patients with pancreatic cancer26 did not improve survival, although high-dose LMWH prophylaxis prevented fatal pulmonary emboli.

ion whether thromboprophylaxis should be started before developing advanced disease. However, LMWH prophylaxis in patients with newly diagnosed lung cancer25 or high-dose LMWH prophylaxis in patients with pancreatic cancer26 did not improve survival, although high-dose LMWH prophylaxis prevented fatal pulmonary emboli. People with advanced cancer admitted to an acute hospital setting likely represent a broader population with respect to performance status and prognosis; therefore, without a clearer understanding of the demographics, we cannot apply these findings beyond the SPCU setting. However, our study offers new insights into the utility and appropriateness of thromboprophylaxis strategies for patients with cancer nearing the end of life. The numbers of participants receiving anticoagulation for previously diagnosed deep vein thrombosis or pulmonary embolism were too small to draw any conclusions about any benefit from secondary prevention of venous thromboembolism for symptom control or survival. These novel data show that approximately a third of patients admitted to SPCUs with advanced cancer and who were not expected to die within 5 days had a femoral deep vein thrombosis. Deep vein thrombosis was not associated with thromboprophylaxis, survival, or symptoms other than leg oedema. Findings are consistent with venous thromboembolism being a manifestation of advanced disease rather than a cause of premature death. Thromboprophylaxis for SPCU inpatients with poor performance status seems to be of little benefit.

as not associated with thromboprophylaxis, survival, or symptoms other than leg oedema. Findings are consistent with venous thromboembolism being a manifestation of advanced disease rather than a cause of premature death. Thromboprophylaxis for SPCU inpatients with poor performance status seems to be of little benefit. Data sharing Data can be accessed by contacting the corresponding author. Supplementary Material Supplementary appendix Acknowledgments We acknowledge the skills and hard work of the research nurses without whom we would not have recruited so well and to time: June Bowes, Rebecca Cloudsdale, Alison Dick, Stacey McKinven, and Liz Reed. We would also like to thank our patient and public representative group ably led by Kathy Seddon. This paper presents independent research funded by the National Institute for Health Research (NIHR) under its Research for Patient Benefit (RfPB) Programme (grant reference number PB-PG-0614-34007). The views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR or the Department of Health and Social Care. AN and SIRN's posts are funded by Marie Curie Cancer Care core grant funding (grant reference MCCC-FCO-17-C).

Programme (grant reference number PB-PG-0614-34007). The views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR or the Department of Health and Social Care. AN and SIRN's posts are funded by Marie Curie Cancer Care core grant funding (grant reference MCCC-FCO-17-C). Contributors CW, MJJ, SIRN, and MW contributed to concept of the study. CW, MJJ, SIRN, MW, AN, EN, and VLA contributed to design of the study. FS, VLA, MJJ, CW, SIRN, JM, JD, BL, and EN were responsible for data collection and management. EN and MW were responsible for radiological review and training. AN and SIRN were the patient and public involvement leads. FS and VLA did the data analysis. CW wrote the first draft of the report. CW, SIRN, MJJ, VLA, and FS participated in critical revision of the report. All authors contributed to and approved the final report. Declaration of interests MW designed and teaches the Focussed Abdominal Ultrasound in Palliative Care training programme, running since 2007, which trained the research nurses for this study. SIRN has received speaker's bureau fees from Pfizer, Daiichi Sankyo, and Bayer, and advisory board fees from Daiichi Sankyo. All other authors declare no competing interests.

Introduction Wiskott-Aldrich syndrome is a rare, X-linked, primary immunodeficiency characterised by microthrombocytopenia, recurrent infections, eczema, and increased risk for autoimmunity and lymphoid malignant diseases.1, 2 The disease is due to mutations in the WAS gene, which encodes the Wiskott-Aldrich syndrome protein (referred to as WASP)—an intracellular key regulator of actin polymerisation.2, 3 WASP-deficient immune cells have compromised immunological synapsis formation, cell migration, and cytotoxicity.1 Survival of patients with Wiskott-Aldrich syndrome is dependent on the severity of the disease. Patients with classic severe phenotype (Zhu clinical score ≥3) have an approximate survival of 15 years with supportive treatment only.4, 5 Haemopoietic stem/progenitor cell (HSPC) transplantation from an HLA-identical sibling donor is the treatment of choice for patients with Wiskott-Aldrich syndrome, but such a donor is not always available.6, 7, 8, 9, 10 HSPC transplantation from an HLA-matched unrelated donor can also be curative but can be hampered by development of graft-versus-host disease, graft rejection, or autoimmune complications if complete chimerism is not achieved.8 The best outcome for unrelated HSPC transplantation occurs when the recipient is younger than 5 years of age at the time of transplant.9, 10

ed donor can also be curative but can be hampered by development of graft-versus-host disease, graft rejection, or autoimmune complications if complete chimerism is not achieved.8 The best outcome for unrelated HSPC transplantation occurs when the recipient is younger than 5 years of age at the time of transplant.9, 10 An alternative potentially curative option for patients with Wiskott-Aldrich syndrome is gene therapy, consisting of a reduced intensity conditioning regimen followed by infusion of ex-vivo genetically corrected autologous HSPCs. Ex-vivo gene therapy has several potential advantages, including absence of graft-versus-host disease and decreased toxicity due to the reduced intensity conditioning regimen. In a previous trial using a γ-retroviral vector under the control of a strong viral promoter, gene therapy was feasible and resulted in functional correction of blood cell defects.11 However, insertions of the γ-retroviral vector were clustered in proto-oncogenes; integration-driven overexpression of these genes triggered the development of severe side-effects such as leukaemia and myelodysplasia in most patients.11 Research in context Evidence before this study

An alternative potentially curative option for patients with Wiskott-Aldrich syndrome is gene therapy, consisting of a reduced intensity conditioning regimen followed by infusion of ex-vivo genetically corrected autologous HSPCs. Ex-vivo gene therapy has several potential advantages, including absence of graft-versus-host disease and decreased toxicity due to the reduced intensity conditioning regimen. In a previous trial using a γ-retroviral vector under the control of a strong viral promoter, gene therapy was feasible and resulted in functional correction of blood cell defects.11 However, insertions of the γ-retroviral vector were clustered in proto-oncogenes; integration-driven overexpression of these genes triggered the development of severe side-effects such as leukaemia and myelodysplasia in most patients.11 Research in context Evidence before this study We searched PubMed from database inception to Nov 25, 2018, for relevant studies of treatments for Wiskott-Aldrich syndrome, including acronyms, synonyms, and closely related words for the terms “Wiskott-Aldrich syndrome”, “gene therapy”, and “haematopoietic stem cell transplantation”. We did not restrict our search by study design or language. We identified further studies by searching relevant websites and the reference lists of reports identified by our search. The only curative treatment for Wiskott-Aldrich syndrome is allogeneic haemopoietic stem/progenitor cell (HSPC) transplantation with a suitable HLA-matched donor. However, many patients do not have such a donor available, and treatment can be hampered by development of graft-versus-host disease, graft rejection, and autoimmune complications if complete chimerism is not achieved. Proof-of-principle for an autologous approach with gene therapy came from a previous trial using a γ-retroviral vector carrying a functional WAS gene under the control of a strong viral promoter. However, integration-driven overexpression of proto-oncogenes triggered development of severe side-effects such as leukaemia and myelodysplasia in most patients. To address the safety issues with γ-retroviral vectors for Wiskott-Aldrich syndrome, a self-inactivating lentiviral vector was developed, encoding for human Wiskott-Aldrich syndrome protein (WASP) under the control of a 1·6 kb reconstituted WAS gene promoter. In this construct the transgene is expressed in the context of the proper endogenous regulatory elements, rather than a viral promoter. A phase 1/2 clinical study on this lentiviral vector-derived gene therapy combined with a reduced-intensity conditioning regimen was initiated in patients with Wiskott-Aldrich syndrome in 2010, and preliminary results for the first three patients treated (follow-up of 20–32 months) were reported in 2013. An additional study, using the same lentiviral vector construct but manufactured by a different process reported treatment of seven patients with Wiskott-Aldrich syndrome, with follow-up of 7–42 months.

2010, and preliminary results for the first three patients treated (follow-up of 20–32 months) were reported in 2013. An additional study, using the same lentiviral vector construct but manufactured by a different process reported treatment of seven patients with Wiskott-Aldrich syndrome, with follow-up of 7–42 months. Added value of this study We report findings of an interim analysis of the phase 1/2 clinical study that was initiated in 2010. Eight patients with Wiskott-Aldrich syndrome are included in the analysis of safety and efficacy data, with median follow-up of 3·6 years (range 0·5–5·6), the longest follow-up published to date. After a reduced-intensity conditioning regimen and subsequent infusion of transduced CD34+ cells, all eight patients showed good haematological reconstitution and sustained engraftment of transduced cells. WASP expression was increased in peripheral blood myeloid and lymphoid lineages and in platelets. Significant improvement of immune function was seen by normalisation of in-vitro T-cell function, pronounced reduction of severe infections, and successful discontinuation of immunoglobulin supplementation followed by response to vaccination. Platelet counts increased substantially, resulting in platelet transfusion independence and absence of severe bleeding events. Implications of all the available evidence

We report findings of an interim analysis of the phase 1/2 clinical study that was initiated in 2010. Eight patients with Wiskott-Aldrich syndrome are included in the analysis of safety and efficacy data, with median follow-up of 3·6 years (range 0·5–5·6), the longest follow-up published to date. After a reduced-intensity conditioning regimen and subsequent infusion of transduced CD34+ cells, all eight patients showed good haematological reconstitution and sustained engraftment of transduced cells. WASP expression was increased in peripheral blood myeloid and lymphoid lineages and in platelets. Significant improvement of immune function was seen by normalisation of in-vitro T-cell function, pronounced reduction of severe infections, and successful discontinuation of immunoglobulin supplementation followed by response to vaccination. Platelet counts increased substantially, resulting in platelet transfusion independence and absence of severe bleeding events. Implications of all the available evidence Lentiviral vector-mediated HSPC gene therapy is an alternative, potentially curative, treatment for patients with Wiskott-Aldrich syndrome that could be administered to a wider range of patients than allogeneic HSPC transplantation and might result in fewer complications. This possibility will be confirmed by longer follow-up of patients treated with gene therapy.

s an alternative, potentially curative, treatment for patients with Wiskott-Aldrich syndrome that could be administered to a wider range of patients than allogeneic HSPC transplantation and might result in fewer complications. This possibility will be confirmed by longer follow-up of patients treated with gene therapy. To address the safety issues with γ-retroviral vectors for Wiskott-Aldrich syndrome, we developed a self-inactivating lentiviral vector encoding human WASP under the control of a 1·6 kb reconstituted WAS gene promoter.12 The use of this endogenous promoter ensures that the transgene is expressed in the context of the proper endogenous regulatory elements.3 Its moderate enhancer activity combined with the self-inactivating long-terminal repeat design and absence of preference to integrate in or near oncogenes reduces the risk of insertional mutagenesis, as shown by in-vitro transformation assays13 and preclinical in-vivo studies in WASP-deficient mice.14, 15

ry elements.3 Its moderate enhancer activity combined with the self-inactivating long-terminal repeat design and absence of preference to integrate in or near oncogenes reduces the risk of insertional mutagenesis, as shown by in-vitro transformation assays13 and preclinical in-vivo studies in WASP-deficient mice.14, 15 In a phase 1/2 clinical study, autologous CD34+ cells genetically modified with a lentiviral vector encoding for human WAS cDNA12, 14 were re-infused to patients with severe Wiskott-Aldrich syndrome after a reduced-intensity conditioning regimen.16 Preliminary data from the initial follow-up (20–32 months) of the first three patients in the study have been reported.16 Here, we report data from an interim analysis that was planned in the study protocol to be done after the first six patients treated had completed at least 3 years of follow-up. This interim analysis was agreed with the European Medicines Agency (EMA) and its paediatric committee (PDCO) as adequate to support a marketing authorisation application. Further analyses are planned when all patients in the study have completed 3 years of follow-up and again when all patients have completed at least 8 years of follow-up.

was agreed with the European Medicines Agency (EMA) and its paediatric committee (PDCO) as adequate to support a marketing authorisation application. Further analyses are planned when all patients in the study have completed 3 years of follow-up and again when all patients have completed at least 8 years of follow-up. Methods Study design and participants We undertook an open-label, non-randomised, phase 1/2 clinical study at the Pediatric Clinical Research Unit and Pediatric Immunohematology and Bone Marrow Transplantation Unit of the San Raffaele Scientific Institute (Milan, Italy). The drug product (autologous CD34+ cells genetically modified with a lentiviral vector encoding for human WAS cDNA), trial design, and procedures have been described previously (appendix pp 1–3).16 Patients were eligible for the study if they had Wiskott-Aldrich syndrome defined by a WAS genetic mutation, with absent WASP expression, severe WAS mutation,2 or severe clinical phenotype (Zhu clinical score ≥3).3 We included patients if they had no HLA-identical sibling donor or, for children younger than 5 years of age, no suitable 10/10 matched unrelated donor or 6/6 unrelated cord blood donor.9 We excluded patients who either had HIV infection, neoplasia, cytogenetic alterations typical of myelodisplastic syndrome or acute myeloid leukaemia, or end-organ functions or any other severe disease which, in the judgment of the investigator, would have been inappropriate for entry into this study. We also excluded patients who had undergone an allogeneic HSPC transplant in the previous 6 months or earlier and had evidence of residual cells of donor origin. The conditions required by the study protocol for enrolment of patients have been assessed and fulfilment of the inclusion and exclusion criteria have been assessed in the screening phase, during which time eligible patients underwent clinical, laboratory, and instrumental pretreatment work-up, including additional disease-specific assessments. We obtained a bone marrow aspirate to assess morphology, cellularity, CD34+ cell content, and clonogenic activity.

eria have been assessed in the screening phase, during which time eligible patients underwent clinical, laboratory, and instrumental pretreatment work-up, including additional disease-specific assessments. We obtained a bone marrow aspirate to assess morphology, cellularity, CD34+ cell content, and clonogenic activity. The protocol and other trial-related materials were approved by the independent ethics committee of the San Raffaele Scientific Institute and the Italian regulatory authority (Agenzia Italiana del Farmaco [AIFA]). Written informed consent was provided by parents or legal representatives of patients before initiation of study-specific procedures. For the assessment of WASP expression and T lymphocyte proliferative responses, we obtained peripheral blood samples from healthy controls, in accordance with the Declaration of Helsinki. Informed consent was approved by the institutional ethics committee of the San Raffaele Hospital in 2009 (Tiget Periblood protocol). Procedures We obtained a backup of autologous HSPCs and cryopreserved them approximately 5–10 weeks before the planned day of gene therapy, either by bone marrow harvest or by leukapheresis, after administration of granulocyte colony-stimulating factor (in most cases, lenograstim), with or without plerixafor. If the HSPC source was mobilised peripheral blood, we also gathered cells for future transduction at this time and cells were cryopreserved until the start of gene transduction on day −3. If the HSPC source was bone marrow, we obtained cells from iliac crests under general anaesthesia on day −3.

or without plerixafor. If the HSPC source was mobilised peripheral blood, we also gathered cells for future transduction at this time and cells were cryopreserved until the start of gene transduction on day −3. If the HSPC source was bone marrow, we obtained cells from iliac crests under general anaesthesia on day −3. We administered intravenous CD20 monoclonal antibody (rituximab 375 mg/m2) on day −22. The reduced-intensity conditioning regimen comprised eight doses of intravenous busulfan administered every 6 h from days −3 to −1 (plus an optional ninth dose to reach the target cumulative busulfan area under the curve [AUC], if needed) and two doses of intravenous fludarabine (30 mg/m2 per day) on days −3 and −2. We monitored busulfan pharmacokinetics and adjusted the dose to avoid excessive toxicity or insufficient exposure.17 Initially, the target busulfan AUC range was 4500–6000 ng × h / mL per dose, equivalent to a cumulative target of 36 000–48 000 ng × h / mL. We amended the cumulative busulfan AUC target to 48 000 (±10%) ng × h / mL after treatment of six patients. The actual doses and AUC of busulfan received by each patient are summarised in the appendix (p 13). One intravenous infusion of autologous CD34+ cells obtained from bone marrow harvest or mobilised peripheral blood stem cells and transduced with the WAS lentiviral vector12, 14 was given on day 1. The planned target dose was 5–10 × 106 CD34+ cells per kg, with an acceptable range of 2–20 × 106 cells per kg. CD34+ cell manipulation and transduction procedures were done at MolMed SpA (Milan, Italy), as described previously.16 Patients were in hospital for a median of 52 days (range 36–82), during which time they underwent chemotherapy, gene therapy, and short-term follow-up, then they were followed up as outpatients unless invasive procedures were needed or complications arose. Clinical examinations were scheduled during screening, at baseline (timing of baseline varied by patient and was from the end of screening to the day before harvest of peripheral blood cells or the day before rituximab administration [which was on day −22]), day −14, day 1 (when gene therapy was initiated), day 7, day 14, day 21, day 30, day 60, day 90, day 180, and year 1, year 1·5, year 2, year 2·5, year 3, and annually to year 8.

m the end of screening to the day before harvest of peripheral blood cells or the day before rituximab administration [which was on day −22]), day −14, day 1 (when gene therapy was initiated), day 7, day 14, day 21, day 30, day 60, day 90, day 180, and year 1, year 1·5, year 2, year 2·5, year 3, and annually to year 8. After treatment, patients' assessment also included routine laboratory tests, microbiological tests, diagnostic imaging, immune profiling, and specific immunological tests, according to the time schedule defined in the study protocol. All adverse events—regardless of seriousness or relation to the drug product—have been recorded in trial case report forms, and the investigators have kept detailed records of all reported adverse events. We used Common Terminology Criteria for Adverse Events for grading of adverse events, or clinical judgment if this grading was not applicable. We assessed quality of life with a questionnaire (appendix pp 1, 2) given to patients' parents, which was composed of questions about lifestyle and life in a protected environment (eg, hygiene restrictions, food, contact with other children), attendance at school (according to age), social ability with peers, and practice of sport activities (according to age). The questionnaire was given to parents before treatment and then every year at the follow-up visit (starting in year 1 after gene therapy).

, hygiene restrictions, food, contact with other children), attendance at school (according to age), social ability with peers, and practice of sport activities (according to age). The questionnaire was given to parents before treatment and then every year at the follow-up visit (starting in year 1 after gene therapy). Outcomes The primary study objectives were to assess the safety of drug product administration after a reduced-intensity conditioning regimen, the long-term engraftment of WASP-expressing gene-transduced cells, and the efficacy of gene therapy in terms of improvement in immune function and thrombocytopenia. Secondary study objectives were to assess the efficacy of gene therapy in improving the patient's clinical condition with respect to severe infections, bleeding episodes, autoimmunity measures, and eczema. Details of safety and efficacy endpoints are in the appendix (p 1). The primary safety endpoints were safety of the conditioning regimen and short-term and long-term safety of lentiviral gene transfer into HSPCs. The primary efficacy endpoints were overall survival, sustained engraftment of genetically corrected HSPCs, expression of vector-derived WASP, improved T-cell function, antigen-specific responses to vaccinations, and improved platelet count and mean platelet volume normalisation. The endpoints for this study are the same as those defined after initial follow-up of the first three patients treated, reported previously.16

PCs, expression of vector-derived WASP, improved T-cell function, antigen-specific responses to vaccinations, and improved platelet count and mean platelet volume normalisation. The endpoints for this study are the same as those defined after initial follow-up of the first three patients treated, reported previously.16 Statistical analysis The sample size was calculated based on the low prevalence of Wiskott-Aldrich syndrome and the degree of novelty of the proposed experimental approach. Thus, we planned to treat eight patients over an estimated 5-year recruitment period. Planned analyses are presented for the intention-to-treat population, defined as all patients treated with lentiviral gene therapy in the study, and consider data available within the clinical database. Data are summarised using descriptive statistics. Adverse events, serious adverse events, infections, bleeding events, number of platelet transfusions, days in hospital, and anti-infective treatments are summarised as rates (number of events or treatments per patient-year of observation [PYO]). We calculated 95% CIs for crude rates based on the Poisson exact method. We used Medidata Rave 2017.2.3 software to collect data in this study (via electronic case report forms). For analysis of descriptive statistics we used SAS version 9.4, and for data analysis and graphical representation of data we used Microsoft Excel 365 and Graph Pad Prism version 5 for Mac OS X. This trial is registered with ClinicalTrials.gov (number NCT01515462) and EudraCT (number 2009-017346-32).

Statistical analysis The sample size was calculated based on the low prevalence of Wiskott-Aldrich syndrome and the degree of novelty of the proposed experimental approach. Thus, we planned to treat eight patients over an estimated 5-year recruitment period. Planned analyses are presented for the intention-to-treat population, defined as all patients treated with lentiviral gene therapy in the study, and consider data available within the clinical database. Data are summarised using descriptive statistics. Adverse events, serious adverse events, infections, bleeding events, number of platelet transfusions, days in hospital, and anti-infective treatments are summarised as rates (number of events or treatments per patient-year of observation [PYO]). We calculated 95% CIs for crude rates based on the Poisson exact method. We used Medidata Rave 2017.2.3 software to collect data in this study (via electronic case report forms). For analysis of descriptive statistics we used SAS version 9.4, and for data analysis and graphical representation of data we used Microsoft Excel 365 and Graph Pad Prism version 5 for Mac OS X. This trial is registered with ClinicalTrials.gov (number NCT01515462) and EudraCT (number 2009-017346-32). Role of the funding source The funder had no role in study design but did contribute to protocol amendments. The funder had a role in data collection, data analysis, data interpretation, and writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

under had no role in study design but did contribute to protocol amendments. The funder had a role in data collection, data analysis, data interpretation, and writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication. Results The clinical protocol was approved by the Italian regulatory authority (AIFA) on March 15, 2010. Between April 20, 2010, and Feb 26, 2015, nine patients (all male) were enrolled to the study, of whom eight received gene therapy. The fifth patient enrolled was withdrawn during screening because he had a revertant mutation in more than 5% of lymphoid cells, which was an exclusion criterion at that time (figure 1). The protocol was subsequently amended in April, 2014, to allow treatment of subsequent patients with revertant mutations (approved by the San Raffaele Scientific Institute ethics committee on April 14, 2014, and by AIFA on April 17, 2014).Figure 1 Study design PBSC=peripheral blood stem cell. At the time of this planned interim analysis (data cutoff April 29, 2016), median follow-up was 3·6 years (range 0·5–5·6). Six patients had at least 3 years of follow-up; of the other two patients, one had 1·5 years of follow-up and the other had 6 months of follow-up (appendix p 14). All eight patients were alive and well at the time of this interim analysis.

is (data cutoff April 29, 2016), median follow-up was 3·6 years (range 0·5–5·6). Six patients had at least 3 years of follow-up; of the other two patients, one had 1·5 years of follow-up and the other had 6 months of follow-up (appendix p 14). All eight patients were alive and well at the time of this interim analysis. The median age of patients on the day of gene therapy was 2·2 years (range 1·1–12·4). Baseline characteristics for each patient are summarised in table 1. The observed WAS gene mutations were nonsense, deletions, missense, and a novel complete inversion of exons 1–7 that included part of the promoter region.18 All eight patients had severe clinical disease before gene therapy (Zhu clinical score range 3–5A). Before gene therapy, five patients had less than 5% of peripheral blood lymphocytes expressing WASP and three patients had residual WASP expression (20–36% of peripheral blood lymphocytes). All patients (n=8) had recurrent infections, eczema, thrombocytopenia, bleeding, and immune disorders (eg, allergies and autoimmune or autoinflammatory manifestations; table 1).Table 1 Baseline clinical characteristics of patients with Wiskott-Aldrich syndrome and drug product characteristics

pheral blood lymphocytes). All patients (n=8) had recurrent infections, eczema, thrombocytopenia, bleeding, and immune disorders (eg, allergies and autoimmune or autoinflammatory manifestations; table 1).Table 1 Baseline clinical characteristics of patients with Wiskott-Aldrich syndrome and drug product characteristics Patient 1 Patient 2 Patient 3 Patient 4 Patient 6 Patient 7 Patient 8 Patient 9 Infections Recurrent ENT infections, bronchiolitis, frequent viral infections (CMV, VZV, HSV, EBV) Severe infections (pneumonia, colitis, arthritis, cellulitis, CVC-related), chronic CMV infection, URTI, UTI Pneumonia with respiratory distress (Pneumonia jiroveci plus CMV), chronic CMV infection, URTI, otitis media Neonatal sepsis, candida infection, chronic CMV infection, viral enteritis Pneumonia, URTI, bacterial conjunctivitis, bacterial UTI, bacterial and viral gastroenteritis, folliculitis, candida infection Recurrent respiratory-tract infections, otitis media, bacterial gastrointestinal infection, colitis or gastroenteritis Recurrent respiratory-tract and skin infections, pneumonia, severe eye infections with residual visual impairment Severe episodes of bacterial and viral enteritis, pneumonia, VZV infection Bleeding events Skin petechiae Skin petechiae, gastrointestinal bleeding, conjunctival bleeding Skin petechiae, gastrointestinal bleeding, epistaxis Skin petechiae, gastrointestinal bleeding, epistaxis, signs of CNS microhaemorrhages, CVC-related bleeding Skin purpura or petechiae, gastrointestinal bleeding, mucosal bleeding, epistaxis, conjunctival bleeding Skin purpura or petechiae, gastrointestinal bleeding, mucosal bleeding Skin petechiae, gastrointestinal bleeding, epistaxis, eye bleeding, mucosal bleeding Gastrointestinal bleeding, haematuria, epistaxis, conjunctival bleeding Eczema score* 3 (moderate) 2 (mild) 4 (severe) 3 (moderate) 2 (mild) 2 (mild) 2 (mild) 2 (mild) Other Developmental disorder, allergy Failure to thrive, elevated inflammatory indexes or vasculitis, hepatosplenomegaly, allergy Gastro-oesophageal reflux or food aversion (fed through a nasogastric tube), food or drug allergy, mild developmental delay Severe refractory autoimmune thrombocytopenia, food or drug allergy with anaphylaxis Food allergy, hepatomegaly, splenomegaly, inflammatory lymphadenopathy, eosinophilia Suspected food allergy Food allergy Recurrent arthritis and vasculitis, Henoch-Schonlein purpura with nephritic-nephrotic syndrome, panuveitis with visual impairment, severe Crohn-like

bocytopenia, food or drug allergy with anaphylaxis Food allergy, hepatomegaly, splenomegaly, inflammatory lymphadenopathy, eosinophilia Suspected food allergy Food allergy Recurrent arthritis and vasculitis, Henoch-Schonlein purpura with nephritic-nephrotic syndrome, panuveitis with visual impairment, severe Crohn-like enterocolitis, perianal fistulae and abscesses, pyoderma gangrenosum WAS gene mutation (rs number) Exon10, 995C→T (Arg321X) in cDNA(rs782802310)† 1337–1338 + 9del in cDNA(rs number awaited) Exon1, 37C→T (Arg13X) in cDNA(rs 193922415) Exon1, 91G→A in cDNA(rs782730988) Exon 10, 1595del, proximal breakpoint (5247_6842del) in genomic DNA(rs number awaited) Exon12, 1509A→T in cDNA(rs1289921805) 735-2A→G in cDNA(rs number awaited) inv(X)(5721;11840)18 in genomic DNA (rs number awaited) Type of mutation Nonsense Deletion Nonsense Missense Deletion Nonstop or readthrough Splice site Inversion Peripheral blood lymphocytes expressing WASP (%)‡ <5% <5% <5% <5% 35·6%§ <5% 20·8% (revertant cell population) 3·4% Zhu clinical score 3 4 4 5 4 3 4 5A Age on day of gene therapy (years) 5·9 1·6 1·1 2·4 1·9 1·9 11·1 12·4 Source of transduced CD34+ cells Bone marrow, mobilised peripheral blood¶‖ Bone marrow Bone marrow Bone marrow Bone marrow Bone marrow Mobilised peripheral blood¶ Mobilised peripheral blood** Total cell dose CD34+ (×106 per kg) 3·7, 5·3‖ 14·1 10·2 10·3 7·8 7·8 7·0 16·8 Transduction efficiency (%) 92%, 88%‖ 97% 100% 94% 93% 91% 96% 88% Vector copy number per genome 1·9, 1·4‖ 2·4 2·8 2·3 2·3 4·3 3·2 3·0 Patients 1, 2, 3, and 9 have been described previously.16, 18 Patient 5 was enrolled but subsequently withdrawn during screening.

(×106 per kg) 3·7, 5·3‖ 14·1 10·2 10·3 7·8 7·8 7·0 16·8 Transduction efficiency (%) 92%, 88%‖ 97% 100% 94% 93% 91% 96% 88% Vector copy number per genome 1·9, 1·4‖ 2·4 2·8 2·3 2·3 4·3 3·2 3·0 Patients 1, 2, 3, and 9 have been described previously.16, 18 Patient 5 was enrolled but subsequently withdrawn during screening. ENT=ear, nose, and throat. CMV=cytomegalovirus. VZV=varicella zoster virus. HSV=herpes simplex virus. EBV=Epstein-Barr virus. CVC=central venous cathether. URTI=upper respiratory-tract infection. UTI=urinary-tract infection. WASP=Wiskott-Aldrich syndrome protein. * Eczema scores were 1 (absent), 2 (mild), 3 (moderate), and 4 (severe); see appendix pp 1–3. † Patient 1 also had a variant of unknown significance (non-pathogenic mutation in exon 6, 572C→A [His180Asn]). ‡ Measured by fluorescence-activated cell sorting. § Undetectable by Western blot. ¶ Obtained after administration of lenograstim. ‖ Patient 1 received both bone marrow and mobilised peripheral blood; transduction was done separately, and drug product characteristics are reported for both. ** Obtained after administration of lenograstim and plerixafor.

‡ Measured by fluorescence-activated cell sorting. § Undetectable by Western blot. ¶ Obtained after administration of lenograstim. ‖ Patient 1 received both bone marrow and mobilised peripheral blood; transduction was done separately, and drug product characteristics are reported for both. ** Obtained after administration of lenograstim and plerixafor. The CD34+ cell source was bone marrow for five patients, mobilised peripheral blood for two patients, and a combination of both for one patient. Mobilised peripheral blood stem cells were obtained after administration of either lenograstim (n=2) or lenograstim and plerixafor (n=1). The dose of drug product administered was 7·0–16·8 × 106 CD34+ cells per kg. CD34+ cells sourced from bone marrow had a transduction efficiency of 91–100%, with an average vector copy number per genome of 1·9–4·3. CD34+ cells sourced from peripheral blood had a transduction efficiency of 88–96% with an average vector copy number per genome of 1·4–3·2 (table 1). The reduced-intensity conditioning regimen was not associated with any unexpected toxic effects during the first 100 days after gene therapy. All eight patients had severe neutropenia, with an absolute neutrophil count of 0·2 × 109 cells per L or lower (appendix p 13). All eight patients had adverse events during follow-up but none of the events was judged by the investigator to be related to the drug product. No patient had severe mucositis.

r gene therapy. All eight patients had severe neutropenia, with an absolute neutrophil count of 0·2 × 109 cells per L or lower (appendix p 13). All eight patients had adverse events during follow-up but none of the events was judged by the investigator to be related to the drug product. No patient had severe mucositis. Engraftment was successful in all eight patients (absolute neutrophil count ≥0·5 × 109 cells per L within 60 days; appendix pp 4, 13) without the need for backup autologous cell administration. Time to neutrophil engraftment ranged from 18 days to 59 days. One patient (patient 4) needed administration of granulocyte colony-stimulating factor (on day 42 and day 44) for delayed neutrophil recovery.

5 × 109 cells per L within 60 days; appendix pp 4, 13) without the need for backup autologous cell administration. Time to neutrophil engraftment ranged from 18 days to 59 days. One patient (patient 4) needed administration of granulocyte colony-stimulating factor (on day 42 and day 44) for delayed neutrophil recovery. The incidence of adverse events and serious adverse events was highest in the first 6 months of follow-up then declined and reached a plateau from 6 months after gene therapy onwards (table 2). The rate of serious adverse events per PYO declined from 4·8 (95% CI 2·9–7·4) during the first 6 months of follow-up after gene therapy to 0·6 (0·2–1·6) between 1 year and 2 years of follow-up and to 0·2 (0·0–0·9) between 2 years and 3 years of follow-up. Likewise, the rate of adverse events per PYO declined from 65·5 (95% CI 57·8–73·9) during the first 6 months of follow-up after gene therapy to 22·1 (17·5–27·6) between 6 months and 12 months of follow-up and to 15·8 (12·8–19·3) between 2 years and 3 years of follow-up. No adverse event or serious adverse event was considered to be related to drug product by the investigator. Rates of adverse events commonly associated with Wiskott-Aldrich syndrome (eg, bleeding events, eczema, and infections) declined after gene therapy. 27 serious adverse events in six patients were reported after gene therapy, of which 23 (85%) were of infectious origin, including pyrexia (five events in three patients), device-related infections including one case of sepsis (four events in three patients), and gastroenteritis of which one case was due to rotavirus (three events in two patients; table 2).Table 2 Serious adverse events after gene therapy

(85%) were of infectious origin, including pyrexia (five events in three patients), device-related infections including one case of sepsis (four events in three patients), and gastroenteritis of which one case was due to rotavirus (three events in two patients; table 2).Table 2 Serious adverse events after gene therapy 0–6 months after gene therapy (n=8) 6–12 months after gene therapy (n=8) 1–2 years after gene therapy (n=7) 2–3 years after gene therapy (n=6) ≥3 years after gene therapy (n=5)* Total after gene therapy (n=8) Patients Events Patients Events Patients Events Patients Events Patients Events Patients Events Any serious adverse event 6 19 1 1 3 4 1 1 1 2 6 27 Infection-related event Pyrexia 2 2 1 1 1 2 0 0 0 0 3 5 Device-related infection 2 2 0 0 1 1 0 0 0 0 2 3 Acute respiratory distress syndrome† 1 1 0 0 0 0 0 0 0 0 1 1 Aspergillus infection 1 1 0 0 0 0 0 0 0 0 1 1 Bacterial sepsis† 1 1 0 0 0 0 0 0 0 0 1 1 Device-related sepsis 1 1 0 0 0 0 0 0 0 0 1 1 Disseminated intravascular coagulation† 1 1 0 0 0 0 0 0 0 0 1 1 Influenza 1 1 0 0 0 0 0 0 0 0 1 1 Interstitial lung disease 1 1 0 0 0 0 0 0 0 0 1 1 Gastroenteritis 1 2 0 0 0 0 0 0 0 0 1 2 Pneumonia aspiration 1 1 0 0 0 0 0 0 0 0 1 1 Viral infection 1 1 0 0 0 0 0 0 0 0 1 1 Gastroenteritis rotavirus 0 0 0 0 1 1 0 0 0 0 1 1 Tooth abscess 0 0 0 0 0 0 1 1 0 0 1 1 Cellulitis 0 0 0 0 0 0 0 0 1 1 1 1 Pneumonia bacterial 0 0 0 0 0 0 0 0 1 1 1 1 Other type of event Electrolyte imbalance 1 1 0 0 0 0 0 0 0 0 1 1 Food allergy 1 1 0 0 0 0 0 0 0 0 1 1 Irritability 1 1 0 0 0 0 0 0 0 0 1 1 Neutropenia 1 1 0 0 0 0 0 0 0 0 1 1 MedDRA (version 19.0) preferred terms are used. MedDRA=Medical Dictionary for Regulatory Activities.

Pneumonia bacterial 0 0 0 0 0 0 0 0 1 1 1 1 Other type of event Electrolyte imbalance 1 1 0 0 0 0 0 0 0 0 1 1 Food allergy 1 1 0 0 0 0 0 0 0 0 1 1 Irritability 1 1 0 0 0 0 0 0 0 0 1 1 Neutropenia 1 1 0 0 0 0 0 0 0 0 1 1 MedDRA (version 19.0) preferred terms are used. MedDRA=Medical Dictionary for Regulatory Activities. * Range 3·0–5·6 years. † These three events were linked, occurring in patient 2, and were considered to be part of the same condition. Transient decreases in haematological indices were noted after the conditioning phase in all eight patients. Results of serum chemistry tests did not show any consistent patterns of clinical concern (data not shown).No adverse reactions to the drug product were reported within 48 h of infusion. No adverse findings were reported from replication competent lentivirus assessments and no evidence was seen of abnormal clonal proliferation or leukaemia development after gene therapy. No antibody against WASP was detected in any patient after gene therapy.

s to the drug product were reported within 48 h of infusion. No adverse findings were reported from replication competent lentivirus assessments and no evidence was seen of abnormal clonal proliferation or leukaemia development after gene therapy. No antibody against WASP was detected in any patient after gene therapy. Data for the first six patients with at least 3 years of follow-up were analysed, and 184 745 unique insertion sites were identified from bone marrow CD34+ cells, bone marrow precursors, and peripheral blood mature cells from lymphoid and myeloid lineages at different timepoints after gene therapy. The diversity of the clonal repertoire of engineered cells was maintained at a high complexity level starting from around 1 year after gene therapy without major inbalances up to and including the latest follow-up analysed (appendix p 5). The common insertion sites detected in all six patients reflect the classic insertional pattern of lentiviral vectors, including PACS1 and KDM2A previously reported in metachromatic leukodystrophy,19 adrenoleukodystrophy, and β thalassemia20 gene therapy trials and not associated with clonal expansion. No enrichment for genomic areas associated with insertional mutagenesis events was detected when comparing data at 12–18 months and 30–36 months after gene therapy (appendix p 6).

d in metachromatic leukodystrophy,19 adrenoleukodystrophy, and β thalassemia20 gene therapy trials and not associated with clonal expansion. No enrichment for genomic areas associated with insertional mutagenesis events was detected when comparing data at 12–18 months and 30–36 months after gene therapy (appendix p 6). All seven patients who were followed up for at least 1 year after gene therapy (figure 1) had adequate engraftment of genetically corrected colony-forming cells in bone marrow (figure 2A). Engraftment was sustained over time in both bone marrow and peripheral blood cell lineages (figures 2B, 2C). Gene-corrected granulocytes (CD15+), megakaryocytic precursors (CD61+), erythroid cells (glycophorin A-positive [GlyA+]), B cells (CD19+), and natural killer (NK) cells (CD56+) were detectable from 1 month after gene therapy. T-cell engraftment was slower compared with other lineages, as expected by the required time for T-cell development through the thymus, with transduced T cells appearing from 3 months after gene therapy. Transduced cell engraftment was higher in lymphoid lineages (CD3+, CD4+, CD8+, CD19+, and CD56+) compared with myeloid lineages (CD15+, CD61+, and GlyA+) because of their known selective advantage.Figure 2 Multilineage engraftment of genetically corrected haemopoietic stem cells and peripheral blood cells in eight patients with Wiskott-Aldrich syndrome treated with lentiviral vector gene therapy

CD19+, and CD56+) compared with myeloid lineages (CD15+, CD61+, and GlyA+) because of their known selective advantage.Figure 2 Multilineage engraftment of genetically corrected haemopoietic stem cells and peripheral blood cells in eight patients with Wiskott-Aldrich syndrome treated with lentiviral vector gene therapy (A) In-vivo engraftment of transduced progenitor cells in bone marrow (LV-positive colonies). (B) Median VCN per cell in bone marrow cell lineages. (C) Median VCN per cell in peripheral blood cell lineages. No vector was detected in bone marrow total cells analysed before gene therapy. LV=lentiviral vector. VCN=vector copy number. GlyA=glycophorin A. NK=natural killer.

lls in bone marrow (LV-positive colonies). (B) Median VCN per cell in bone marrow cell lineages. (C) Median VCN per cell in peripheral blood cell lineages. No vector was detected in bone marrow total cells analysed before gene therapy. LV=lentiviral vector. VCN=vector copy number. GlyA=glycophorin A. NK=natural killer. Lymphocyte counts decreased after conditioning, as expected, then increased progressively. At the latest follow-up visit, T-cell counts (both CD4+ helper and CD8+ cytotoxic T lymphocytes), B-cell counts, and NK cell counts were normal for age in all eight patients,21 apart from one patient's (patient 7) total and CD8+ T-cell counts, which were just below the normal range (appendix pp 15–20). Naive CD4+ T cells were normal in four of eight patients at the last follow-up visit, similar to pretreatment levels (appendix p 21).22 The fraction of lymphocytes expressing WASP was substantially improved in all eight patients, with a ten times increase in the median value compared with baseline seen at 3 months after gene therapy (range 11–72%) and a 20 times increase recorded consistently at 12 months after gene therapy (figure 3A). The fraction of WASP-positive lymphocytes increased from a median of 3·9% (range 1·8–35·6) before gene therapy to 66·7% (55·7–98·6) at 12 months after gene therapy. A selective advantage was noted in lymphoid lineages, particularly T cells, which showed a higher level of expression compared with myeloid lineages (appendix pp 7, 8). The T-cell proliferative response to anti-immobilised CD3 (anti-CD3i) was severely compromised in all eight patients before gene therapy and steadily improved after gene therapy. In the six patients with 3 years of follow-up, the proliferative response at 3 years was comparable with the response in T cells from healthy controls (figure 3B). The proliferative response to phytohaemagglutinin was normal in seven of the eight patients and returned to normal levels in all seven patients with follow-up at least 1 year after gene therapy, including the patient with abnormal levels before gene therapy.Figure 3 Immune reconstitution and clinical benefit

igure 3B). The proliferative response to phytohaemagglutinin was normal in seven of the eight patients and returned to normal levels in all seven patients with follow-up at least 1 year after gene therapy, including the patient with abnormal levels before gene therapy.Figure 3 Immune reconstitution and clinical benefit Gene therapy was administered on day 1 (month 0). (A) Percentage of peripheral blood lymphocytes expressing WASP, as measured by flow cytometry. (B) T-cell proliferation response (expressed as SI) after incubation with increasing concentrations of anti-CD3i antigen, by observation period. Boxes represent upper and lower quartiles (outliers excluded). Horizontal line within the box is the median. Whiskers represent most extreme points ≤1·5 × IQR. Control data are from 12 healthy children. (C) Rate of severe infections (events per PYO) for each observation period. Error bars represent the 95% CI. (D) Number of days in hospital (days per PYO) for each observation period. Error bars represent the 95% CI. (E) Eczema scores are 1 (absent), 2 (mild), 3 (moderate), and 4 (severe). Six patients were followed up for at least 3 years after gene therapy. WASP=Wiskott-Aldrich syndrome protein. SI=stimulation index. Anti-CD3i=anti-immobilised CD3. PYO=patient-year of observation.

ervation period. Error bars represent the 95% CI. (E) Eczema scores are 1 (absent), 2 (mild), 3 (moderate), and 4 (severe). Six patients were followed up for at least 3 years after gene therapy. WASP=Wiskott-Aldrich syndrome protein. SI=stimulation index. Anti-CD3i=anti-immobilised CD3. PYO=patient-year of observation. The improvement in immune function was accompanied by clinical benefit. The rate of severe infections increased in the first 6 months after gene therapy, in accordance with the effects of the conditioning regimen, but declined strikingly thereafter to approximately an order of magnitude lower than the rate in the year before treatment (figure 3C). Severe infections fell from 2·38 (95% CI 1·44–3·72) events per PYO in the year before gene therapy to 0·31 (0·04–1·11) events per PYO in the second year after gene therapy and 0·17 (0·00–0·93) events per PYO in the third year after gene therapy. No reactivation of cytomegalovirus was detected after 6 months of follow-up in three patients who had chronic cytomegalovirus infection before treatment. The rate of anti-infective drug use 2–3 years after gene therapy was approximately four times lower than the rate in the year before treatment (11·3 treatments per PYO vs 42·0 treatments per PYO). The number of days spent in hospital also declined substantially from 6 months after gene therapy onwards (figure 3D). Eczema scores improved in all six patients who reached at least 3 years of follow-up, and eczema had completely resolved in five patients by 3 years (figure 3E).

ts per PYO vs 42·0 treatments per PYO). The number of days spent in hospital also declined substantially from 6 months after gene therapy onwards (figure 3D). Eczema scores improved in all six patients who reached at least 3 years of follow-up, and eczema had completely resolved in five patients by 3 years (figure 3E). At the time of this interim analysis, intravenous immunoglobulin had been discontinued in seven patients with follow-up longer than 1 year (median time to discontinuation after gene therapy, 1·7 years [IQR 0·9–3·0]). Serum immunoglobulin levels before and after gene therapy are reported in the appendix (p 22). Immunological status was considered to be sufficiently improved to allow the start of childhood vaccinations in five patients. A protective antibody response was shown in all five patients to tetanus toxoid, diphtheria, Haemophilus influenzae type B, and hepatitis B. Four of five patients also had a protective response to Bordetella pertussis. Protective levels of IgG antibodies against Streptococcus pneumoniae were detected in one of five assessable patients after polysaccharide pneumococcal vaccine and in three of four assessable patients after pneumococcal conjugated vaccine (appendix pp 23, 24).

nts also had a protective response to Bordetella pertussis. Protective levels of IgG antibodies against Streptococcus pneumoniae were detected in one of five assessable patients after polysaccharide pneumococcal vaccine and in three of four assessable patients after pneumococcal conjugated vaccine (appendix pp 23, 24). The proportion of platelets expressing WASP, as measured by flow cytometry, progressively improved over time in all eight patients after gene therapy (figure 4A; appendix pp 9, 10). WASP-positive platelets increased from 19·1% (range 4·1–31·0) before gene therapy to 76·6% (53·1–98·4) at 12 months after gene therapy. The ratio of mean fluorescence intensity in platelets of patients normalised to those of healthy controls increased from a median of 0·18 (0·01–0·31) before gene therapy to 0·43 (0·20–1·03) at 1 year after gene therapy. A substantial improvement in platelet count was also seen (figure 4B). Before gene therapy, platelet counts were lower than 20 × 109 per L in seven of eight patients. At the last follow-up visit, the platelet count had increased to 20–50 × 109 per L in one patient, 50–100 × 109 per L in five patients, and more than 100 × 109 per L in two patients, which resulted in independence from platelet transfusions and absence of severe bleeding events, although platelets counts remained below the lower limit of the normal range in most patients. Mean platelet volume, which is characteristically reduced in patients with Wiskott-Aldrich syndrome, was within the normal range (7·4–10·9 fL) in seven assessable patients at 1 year after gene therapy and remained at this level except for a transient dip for one patient (patient 6) at year 2. All eight patients needed platelet transfusions before gene therapy (2·5 transfusions per PYO in the year before enrolment), falling to less than 0·3 transfusions per PYO for observation periods beyond 1 year, when transfusions were given only as a precautionary measure for surgery or trauma. The transfusion rate was higher in the preparatory period before gene therapy (42·9 transfusions per PYO, equal to 3·6 transfusions per person-month of observation), mainly because of standard use as bleeding prophylaxis for invasive procedures such as bone marrow aspiration or harvest, central venous cathether positioning, and peripheral blood stem cell leukoapheresis.

eriod before gene therapy (42·9 transfusions per PYO, equal to 3·6 transfusions per person-month of observation), mainly because of standard use as bleeding prophylaxis for invasive procedures such as bone marrow aspiration or harvest, central venous cathether positioning, and peripheral blood stem cell leukoapheresis. Two patients (patients 8 and 9) who received mobilised peripheral blood-derived drug product had a faster increase in platelet count after gene therapy compared with those who received bone marrow-derived drug product (appendix pp 9, 10). None of the treated patients required splenectomy.Figure 4 Platelets and bleeding events Gene therapy was administered on day 1 (month 0). (A) Percentage of platelets expressing WASP, as measured by flow cytometry. (B) Mean (SD) platelet count (× 109 cells per L). SD is not plotted for follow-up at 60 months because data were available for only two patients. (C) Moderate or severe bleeding events (per PYO) for each observation period. Error bars represent the 95% CI. (D) Distribution of all bleeding events for each observation period, by body site. Total number of events for each observation period is shown in the centre of each chart. WASP=Wiskott-Aldrich syndrome protein. PYO=patient-year of observation.

ents (per PYO) for each observation period. Error bars represent the 95% CI. (D) Distribution of all bleeding events for each observation period, by body site. Total number of events for each observation period is shown in the centre of each chart. WASP=Wiskott-Aldrich syndrome protein. PYO=patient-year of observation. The improvement in platelet count was accompanied by a reduction in the rate and severity of bleeding events (figure 4C). No severe bleeding events were reported after gene therapy, and a pronounced reduction in moderate bleeding events was also seen after gene therapy. Bleeding events after gene therapy were predominantly related to the skin or mucosa and none resulted in hospitalisation (figure 4D).

erity of bleeding events (figure 4C). No severe bleeding events were reported after gene therapy, and a pronounced reduction in moderate bleeding events was also seen after gene therapy. Bleeding events after gene therapy were predominantly related to the skin or mucosa and none resulted in hospitalisation (figure 4D). Patient 4 had severe refractory autoimmune thrombocytopenia23 before gene therapy and was treated both before and after gene therapy with immunosuppressive drugs (high-dose intravenous immunoglobulin, steroids, and rituximab) and with thrombopoietin receptor agonists for 8 months after gene therapy. This patient became independent of platelet transfusions after gene therapy, with only one transfusion in the 1–2 year observation period and one transfusion in the 4–5 year observation period; these transfusions were precautionary because of a surgical procedure and head trauma, respectively. Patient 2 developed severe autoimmune thrombocytopenia approximately 3 months after gene therapy, which resolved within 6 months. Patient 9 had a prolonged history of inflammatory and autoimmune manifestations (Crohn-like enterocolitis, pyoderma gangrenosum, and arthritis) requiring multiple chronic treatments before gene therapy.18 However, disease symptoms and signs fully resolved after gene therapy (appendix p 11). A panel of 15 serum autoantibodies were tested before and after gene therapy. The incidence of autoimmune antibodies was highest at baseline (ten positive reports in four of eight patients) but declined by year 1 (two positive reports in two of eight patients) and fell further at year 3 (three positive reports in two of six patients; data not shown). Positive autoimmunity test results were intermittent in most patients and were not associated with any clinical manifestations, except for autoimmune thrombocytopenia in patient 2.

(two positive reports in two of eight patients) and fell further at year 3 (three positive reports in two of six patients; data not shown). Positive autoimmunity test results were intermittent in most patients and were not associated with any clinical manifestations, except for autoimmune thrombocytopenia in patient 2. Four patients were living in a protected environment before entering the study but were able to enter the community after gene therapy. The clinical improvements enabled patients to attend school (according to age), have normal social interactions, and to take part in sporting activities. Discussion This interim analysis of data from a study of eight patients with severe Wiskott-Aldrich syndrome, six of whom have been followed up for more than 3 years, suggests that this disease can be successfully treated by one infusion of autologous CD34+ HSPCs transduced with a lentiviral vector encoding for a functional WAS gene, preceded by a reduced-intensity conditioning regimen. Treatment outcome was favourable for all eight patients, and all patients were alive and well at the last follow-up visit.

e can be successfully treated by one infusion of autologous CD34+ HSPCs transduced with a lentiviral vector encoding for a functional WAS gene, preceded by a reduced-intensity conditioning regimen. Treatment outcome was favourable for all eight patients, and all patients were alive and well at the last follow-up visit. Preparatory conditioning was tolerated as anticipated in all eight patients. Expected chemotherapy-related severe neutropenia was transient, and all eight patients showed good haemopoietic recovery thereafter, with only one patient needing administration of granulocyte colony-stimulating factor. The infusion of autologous lentiviral vector-transduced HSPCs was well tolerated, with no adverse reactions. No immune response against the vector or WASP was detected. Sustained multilineage engraftment of genetically corrected cells was recorded in peripheral blood and bone marrow, attributable to persistent engraftment of HSPCs in the bone marrow over time. Of note, haemopoiesis is driven by distinct subsets of HSPCs, which contribute differently in early and late phases of reconstitution.24 Expression of vector-derived WASP in most lymphoid cells and platelets led to improvement of immune function and platelet counts. This expression resulted in clinical benefit for patients, who showed a progressive amelioration of their clinical status, with protection from severe infections and bleeding events.

Preparatory conditioning was tolerated as anticipated in all eight patients. Expected chemotherapy-related severe neutropenia was transient, and all eight patients showed good haemopoietic recovery thereafter, with only one patient needing administration of granulocyte colony-stimulating factor. The infusion of autologous lentiviral vector-transduced HSPCs was well tolerated, with no adverse reactions. No immune response against the vector or WASP was detected. Sustained multilineage engraftment of genetically corrected cells was recorded in peripheral blood and bone marrow, attributable to persistent engraftment of HSPCs in the bone marrow over time. Of note, haemopoiesis is driven by distinct subsets of HSPCs, which contribute differently in early and late phases of reconstitution.24 Expression of vector-derived WASP in most lymphoid cells and platelets led to improvement of immune function and platelet counts. This expression resulted in clinical benefit for patients, who showed a progressive amelioration of their clinical status, with protection from severe infections and bleeding events. It should be considered that this study has a small sample size and is based at one centre, is single arm, and is open label, using the patient's prestudy status as the control for the assessment of efficacy of gene therapy. This study design is justified by the rarity and severity of the disease and type of intervention. The study design was agreed with the EMA and its PDCO and takes EU guidelines into consideration.

arm, and is open label, using the patient's prestudy status as the control for the assessment of efficacy of gene therapy. This study design is justified by the rarity and severity of the disease and type of intervention. The study design was agreed with the EMA and its PDCO and takes EU guidelines into consideration. With respect to safety, adverse events and serious adverse events were recorded most frequently during the first 6 months after gene therapy. These events were mainly of infectious origin. Two serious adverse events that were related to infections—both within the first 6 months after gene therapy—were life-threatening, but these events resolved with minor sequelae in one patient.16 Infections are not unexpected during ongoing immune reconstitution because of the time needed for engraftment, similar to patients who undergo traditional allogeneic bone marrow transplantation.9, 10 No graft failure was reported after gene therapy; graft-versus-host disease was not reported either, a finding that was expected because of the autologous nature of the procedure. No organ deterioration related to disease progression or complications or to treatment-related toxicity was seen. Importantly, malignant diseases typically reported in patients with Wiskott-Aldrich syndrome have not been detected in any of the eight patients treated to date. Intervention-free survival (ie, free from second HSPC transplant, splenectomy, or chronic immunosuppression) is 100% as of the data cutoff.

toxicity was seen. Importantly, malignant diseases typically reported in patients with Wiskott-Aldrich syndrome have not been detected in any of the eight patients treated to date. Intervention-free survival (ie, free from second HSPC transplant, splenectomy, or chronic immunosuppression) is 100% as of the data cutoff. Autoimmunity is a key risk factor for a poor outcome in patients with Wiskott-Aldrich syndrome.5, 18 Despite the presence of circulating autoantibodies, clinical autoimmune manifestations reported after gene therapy were restricted to one case of transient autoimmune thrombocytopenia occurring in the early follow-up phase and one case already present before treatment. No autoimmune endocrinopathies were seen after gene therapy. Importantly, the pre-existing severe refractory autoimmune thrombocytopenia in one patient and severe autoimmune and autoinflammatory clinical manifestations in another resolved completely after treatment. Eczema was cleared in seven of eight patients; in the one patient in whom eczema did not resolve, eczema was mild and intermittent by 3-year follow-up. These results indicate that the immune dysregulation that predisposes patients with Wiskott-Aldrich syndrome to develop autoimmune manifestations could be ameliorated by insertion of the correct WAS gene and establishment of functional T cells and B cells. These data accord with the finding that lentiviral gene therapy corrected the alterations of both central and peripheral B-cell tolerance checkpoints and ameliorated B-cell development and functionality in patients with Wiskott-Aldrich syndrome.25, 26 Of note, the engraftment and expansion of gene-corrected B cells expressing WASP could have been favoured by pretreatment rituximab-mediated depletion of B cells, particularly autoreactive B cells, highlighting the key role of this drug in the conditioning of patients with Wiskott-Aldrich syndrome before gene therapy.

Of note, the engraftment and expansion of gene-corrected B cells expressing WASP could have been favoured by pretreatment rituximab-mediated depletion of B cells, particularly autoreactive B cells, highlighting the key role of this drug in the conditioning of patients with Wiskott-Aldrich syndrome before gene therapy. Validated quality-of-life general assessments might not adequately assess the key disease-related effects of Wiskott-Aldrich syndrome and, hence, risk underestimating the impaired quality of life of a patient with this disease. In this study, we used a simple questionnaire (appendix pp 1, 2) to assess the effect of disease burden on patients' quality of life before and after treatment, focusing in particular on life restrictions, social interactions, attendance to educational services and participation in sporting activities. Moreover, we also assessed the frequency of hospitalisation before and after gene therapy, as a measure of medical care requirement affecting patients' daily life. Overall, gene therapy resulted in a substantial improvement in quality of life for patients and their families, with patients socialising in their community with peers, attending school (according to age), starting to participate in sporting activities, and progressively reducing their time spent in hospital and their chronic medical care needs. Further work would be needed to develop quality-of-life assessments specific to Wiskott-Aldrich syndrome for implementation in a future clinical trial.

school (according to age), starting to participate in sporting activities, and progressively reducing their time spent in hospital and their chronic medical care needs. Further work would be needed to develop quality-of-life assessments specific to Wiskott-Aldrich syndrome for implementation in a future clinical trial. By comparison with a British and French gene therapy clinical study in patients with Wiskott-Aldrich syndrome that used a lentiviral vector structure,27 our results are similar with respect to improvement in immune functions and resolution of eczema and autoimmunity, but they differ with respect to the degree of HSPC engraftment, immune reconstitution, and platelet count increase. Specifically, three of four patients in the British and French trial who did not have splenectomy had platelet counts lower than 20 × 109 cells per L after gene therapy (median follow-up after gene therapy, 2 years [range 9–30 months]), compared with our study in which none of the eight patients had splenectomy and all eight patients had platelet counts greater than 20 × 109 cells per L after gene therapy at the last follow-up visit. Immune reconstitution was suboptimum in the British and French trial because only two of six patients with follow-up longer than 1 year stopped immunoglobulin replacement therapy and CD8+ T cells remained low in most patients. This finding contrasts with our results showing normalisation of lymphocyte counts in seven of eight patients, allowing the suspension of antimicrobial prophylaxis in all eight patients after gene therapy and discontinuation of intravenous immunoglobulin in all seven patients with follow-up longer than 1 year, followed by vaccine administration in five patients with a protective antibody response for most vaccines. Differences in disease severity, age at treatment, study conduct, and preconditioning might have contributed to the diverse outcome between the two studies. Furthermore, the specific vector manufacture and transduction processes could account for the differences seen in the correction level in vitro and in vivo.

cines. Differences in disease severity, age at treatment, study conduct, and preconditioning might have contributed to the diverse outcome between the two studies. Furthermore, the specific vector manufacture and transduction processes could account for the differences seen in the correction level in vitro and in vivo. Although the CD34+ cell dose was similar (median 9·6 × 106 cells per kg in our study vs 7·3 × 106 cells per kg in the British and French trial), median vector copy number per genome in transduced CD34+ cells was 2·4 in our study and 1·3 in the British and French trial.27 Moreover, the degree of in-vivo gene marking in the myeloid lineage in the British and French trial was lower compared with our trial (median vector copy number per genome 0·1 in the British and French trial vs 0·8 in our study at the last available follow-up visit). This difference was seen despite us using a less intense dose of chemotherapy in our study.

rking in the myeloid lineage in the British and French trial was lower compared with our trial (median vector copy number per genome 0·1 in the British and French trial vs 0·8 in our study at the last available follow-up visit). This difference was seen despite us using a less intense dose of chemotherapy in our study. Insertional mutagenesis and consequent leukaemia is considered a known risk related to gene therapy using integrating viral vectors. In a German study of gene therapy for Wiskott-Aldrich syndrome using a γ-retroviral vector,11 seven of ten treated patients developed leukaemia, which was attributed to the type of vector used in combination with a strong promoter. Leukaemias were diagnosed 1–5 years after gene therapy, with four of seven cases diagnosed within 3 years. In our study that used a lentiviral vector, no clonal expansion or leukaemia development have been recorded to date. The lack of a transcriptionally active long-terminal repeat of the self-inactivating lentiviral backbone, combined with a moderately active internal promoter to drive transgene expression, such as the WASP endogenous promoter, is a major safety advantage of the vector compared with the γ-retroviral vector used in the German Wiskott-Aldrich syndrome study.11 The insertion site analyses we did in our current trial show that the vector insertions are polyclonal, and there is no evidence of any aberrant selection of clones carrying insertion sites in the proximity of known oncogenes.

or compared with the γ-retroviral vector used in the German Wiskott-Aldrich syndrome study.11 The insertion site analyses we did in our current trial show that the vector insertions are polyclonal, and there is no evidence of any aberrant selection of clones carrying insertion sites in the proximity of known oncogenes. In the British and French clinical study,27 no leukaemia events were reported in patients followed up for 7–42 months. To date, more than 100 patients with various inherited diseases have been treated using lentiviral vectors and no causally related adverse events have been reported.28

or compared with the γ-retroviral vector used in the German Wiskott-Aldrich syndrome study.11 The insertion site analyses we did in our current trial show that the vector insertions are polyclonal, and there is no evidence of any aberrant selection of clones carrying insertion sites in the proximity of known oncogenes. In the British and French clinical study,27 no leukaemia events were reported in patients followed up for 7–42 months. To date, more than 100 patients with various inherited diseases have been treated using lentiviral vectors and no causally related adverse events have been reported.28 The current accepted curative treatment for Wiskott-Aldrich syndrome is allogeneic HSPC transplantation when an HLA-identical donor is available. Historically, transplants from alternative donors were characterised by increased risk of toxicity and mortality, and larger cohorts using current alternative donor transplant protocols have not been published (eg, NCT02064933 by the Primary Immune Deficiency Treatment Consortium). Findings of two large retrospective series of patients who received HSPC transplants—an international study of 194 patients by Moratto and colleagues9 and a single-centre US study of 47 patients by Shin and colleagues10—showed overall survival of 84% and 81%, respectively, with most deaths occurring within the first 3–6 months after transplantation and severe infection being the main cause of death. In the international study,9 a risk factor for determining outcome was age at the time of transplant, with patients older than 5 years and an unrelated donor having a poorer outcome. In the US study,10 older children were most likely to die: seven of nine patients who died were older than 2 years at the time of HSPC transplantation.

tional study,9 a risk factor for determining outcome was age at the time of transplant, with patients older than 5 years and an unrelated donor having a poorer outcome. In the US study,10 older children were most likely to die: seven of nine patients who died were older than 2 years at the time of HSPC transplantation. In patients with Wiskott-Aldrich syndrome who underwent allogeneic HSPC transplantation, complications were most common within the first year after the transplant (46%), with the main complications being infections needing hospitalisation (55 [28%] of 194 patients in the international study),9 autoimmune manifestations (27 [14%] of 194 patients in the international study,9 and 17 [55%] of 31 patients receiving a transplant after 2001 in the US study),10 severe (grade >2) acute graft-versus-host disease (22 [11%] of 194 patients in the international study,9 and between 15% and 44% of patients in the US study),10 and graft failure or rejection (13 [7%] of 194 patients in the international study,9 and three [6%] of 47 patients in the US study).10 Autoimmune cytopenias resolved in most patients in the international study, with median time to recovery of 14 months after HSPC transplantation.9 The most common cytopenia was thrombocytopenia, typically occurring within the first 6 months after HSPC transplantation in the international study.9 A recent improvement in outcome of HSPC transplantation in patients with Wiskott-Aldrich syndrome29 has been obtained owing to advances in graft manipulation techniques and optimisation of conditioning regimens.30

nia, typically occurring within the first 6 months after HSPC transplantation in the international study.9 A recent improvement in outcome of HSPC transplantation in patients with Wiskott-Aldrich syndrome29 has been obtained owing to advances in graft manipulation techniques and optimisation of conditioning regimens.30 With the limitations of a single-centre study, our results are suggestive of a more benign safety profile, with fewer complications after gene therapy compared with those previously reported after HSPC transplantation.9, 10 No deaths were reported in our study, with a median age at treatment of 2·2 years (range 1·1–12·4). Gene therapy is, therefore, a potentially suitable alternative to HSPC transplantation because of the absence of graft-versus-host disease, since gene therapy is an autologous procedure and is preceded by use of a reduced-intensity conditioning regimen. This strategy opens the possibility to treat adult patients with Wiskott-Aldrich syndrome with chronic complications, for whom allogeneic transplantation would be associated with high risks; the report of successful lentiviral gene therapy in a 30-year-old patient with severe Wiskott-Aldrich syndrome manifestations supports this idea.31

s the possibility to treat adult patients with Wiskott-Aldrich syndrome with chronic complications, for whom allogeneic transplantation would be associated with high risks; the report of successful lentiviral gene therapy in a 30-year-old patient with severe Wiskott-Aldrich syndrome manifestations supports this idea.31 Successful HSPC transplantation has been proven to correct thrombocytopenia in most patients. In the international study by Moratto and colleagues,9 HSPC transplantation resulted in a significant increase of mean platelet count at the last follow-up visit; however, in 36 (24%) of 152 patients, the platelet count did not return to normal and, in 14 (9%) patients, severe thrombocytopenia was persistent (platelet count 10–42 × 109 cells per L). Analysis of variables in the study by Moratto and colleagues9 affecting the degree of thrombocytopenia correction after allogeneic HSPC transplantation showed a significant effect of myeloid chimerism on the reconstitution of platelet counts in patients with follow-up longer than 1 year. Although gene therapy in our study did not return platelet counts to normal, platelet levels increased significantly and platelets were of a typical size, reducing the risk of bleeding and related disorders, and no patients needed to be admitted to hospital because of bleeding events after treatment. It is noteworthy that the patients in our study had platelet counts after treatment in the range of those measured in patients with mixed donor–host chimerism after successful allogeneic bone marrow transplant (range 30 000–150 000 cells per μL).9 This observation suggests that, in gene therapy, a minimum level of engraftment could be needed to correct thrombocytopenia with a significant effect on reconstitution of platelet number. We noted some correlation between median platelet counts and the proportion of transduced clonogenic progenitors in bone marrow at 1 year after gene therapy (appendix p 12).

herapy, a minimum level of engraftment could be needed to correct thrombocytopenia with a significant effect on reconstitution of platelet number. We noted some correlation between median platelet counts and the proportion of transduced clonogenic progenitors in bone marrow at 1 year after gene therapy (appendix p 12). Because bone marrow contains a mixed population of transduced and non-transduced HSPCs and, hence, megakaryocytes both positive and negative for WASP, and since the percentage of WASP-positive platelets in peripheral blood was considerably higher than the percentage of lentiviral vector-transduced HSPCs in bone marrow (median 76·6% [figure 4A] vs 42·1% [figure 2A] at 1 year of follow-up), it is possible that a selective advantage of platelets expressing WASP exists and favours their survival in the periphery or their release from bone marrow into peripheral blood. Correction of thrombocytopenia is of paramount importance for cure of Wiskott-Aldrich syndrome, particularly for patients with milder disease in whom thrombocytopenia is the prominent feature. Therefore, efforts to improve correction of platelet counts should be made in the future. Further optimisation of the conditioning regimen could increase lentiviral vector-corrected HSPC engraftment and lead to enhanced platelet recovery. Improvement of the vector construct might be also considered but should be weighted with the potential risks of using a stronger viral promoter.32

d be made in the future. Further optimisation of the conditioning regimen could increase lentiviral vector-corrected HSPC engraftment and lead to enhanced platelet recovery. Improvement of the vector construct might be also considered but should be weighted with the potential risks of using a stronger viral promoter.32 Differences in stem-cell source, cell dose, purification, manipulation, and conditioning might account for the longer duration of neutropenia noted in our study compared with allogeneic transplantation.29, 30 CD34+ cells from bone marrow and mobilised peripheral blood have similar capacity to re-establish long-term haemopoiesis after allogeneic HSPC transplantation, but the use of mobilised peripheral blood is associated with a reduction in time to engraftment, probably attributable to differences in composition of progenitor cells.33 Studies in a larger cohort of patients are needed to establish if ex-vivo gene-corrected CD34+ cells from mobilised peripheral blood and bone marrow differ in their biological properties and capacity to restore haemopoiesis after gene therapy for Wiskott-Aldrich syndrome.

fferences in composition of progenitor cells.33 Studies in a larger cohort of patients are needed to establish if ex-vivo gene-corrected CD34+ cells from mobilised peripheral blood and bone marrow differ in their biological properties and capacity to restore haemopoiesis after gene therapy for Wiskott-Aldrich syndrome. In conclusion, data from the interim analysis of our study suggest that patients with severe Wiskott-Aldrich syndrome can be treated with gene therapy to enable restoration of immune function, substantially increased platelet count, and reduction in autoimmunity; patients had strikingly reduced incidence of disease-specific clinical events without complications related to the drug product. Furthermore, these improvements were sustained or increased over the duration of follow-up, leading to substantial improvements in quality of life for patients. Longer follow-up through a dedicated registry, along with results in a larger cohort of patients with Wiskott-Aldrich syndrome, will provide further evidence of the long-term safety and efficacy of this treatment. Data sharing For further information, please contact the corresponding author. Supplementary Material Supplementary appendix

In conclusion, data from the interim analysis of our study suggest that patients with severe Wiskott-Aldrich syndrome can be treated with gene therapy to enable restoration of immune function, substantially increased platelet count, and reduction in autoimmunity; patients had strikingly reduced incidence of disease-specific clinical events without complications related to the drug product. Furthermore, these improvements were sustained or increased over the duration of follow-up, leading to substantial improvements in quality of life for patients. Longer follow-up through a dedicated registry, along with results in a larger cohort of patients with Wiskott-Aldrich syndrome, will provide further evidence of the long-term safety and efficacy of this treatment. Data sharing For further information, please contact the corresponding author. Supplementary Material Supplementary appendix Acknowledgments The San Raffaele Telethon Institute for Gene Therapy (SR-Tiget) is a joint venture between the Italian Telethon Foundation and Ospedale San Raffaele (OSR). The TIGET-WAS gene therapy clinical trial (NCT01515462) was originally sponsored by the Italian Telethon Foundation and promoted by SR-Tiget. Wiskott-Aldrich syndrome gene therapy was subsequently in-licensed to GlaxoSmithKline in 2014, and GlaxoSmithKline became the financial sponsor of the trial. The Italian Telethon Foundation and OSR are entitled to receive milestone payments and royalties on commercialisation of such therapies. GlaxoSmithKline divested this gene therapy before submission of this report to Orchard Therapeutics (April, 2018). We thank all medical and nursing personnel of the Paediatric Immunohaematology and Haematology and Bone Marrow Transplant Unit of San Raffaele Hospital (Milan, Italy); Stefano Zancan and all personnel of the SR-Tiget Clinical Trial Office; local referring doctors who helped with patients' management; and all eight patients who participated in this study and their families. We acknowledge the Italian Telethon Foundation for continuous support and guidance. We thank Michela Gabaldo (SR-Tiget) for support to projects; and Giuliana Tomaselli, Luisella Meroni, and Samih El Hossary (SR-Tiget) for support to patients.

nt; and all eight patients who participated in this study and their families. We acknowledge the Italian Telethon Foundation for continuous support and guidance. We thank Michela Gabaldo (SR-Tiget) for support to projects; and Giuliana Tomaselli, Luisella Meroni, and Samih El Hossary (SR-Tiget) for support to patients. Contributors FF and MPC were key study investigators, provided clinical care of patients, and collected, reviewed, analysed, and interpreted data. AA designed and coordinated the study as principal investigator. StGa and MGV participated in study design and contributed to data analysis. StGi, IB, SaSc, and FD designed the molecular and immunological experimental plan, analysed biological patient samples, and contributed to data interpretation. MF, CF, and EA collected and managed data. CD and KvR contributed to protocol amendments and reviewed, verified, and interpreted data. GA oversaw the study and data analyses. EJC-S provided critical analysis and review of the publication. FB, MM, MEB, VC, AAA, RP, DC, RF, PS, and SaGa participated in patients' assessment, treatment and follow up and provided data for the clinical trial. MHA referred and managed patients and provided data for the clinical trial. AB, FAS, LB-R, and SeSc did experimental analyses on patient samples and contributed to data analysis. HMW assessed specific antibody responses to vaccinations. LB did vector insertion analyses and contributed to data interpretation. MC coordinated patient's assessments and managed biological samples. AV contributed to data review and data interpretation. MGR contributed to study design, data review, and data interpretation. FC contributed to study design, treatment of patients, data review, and data interpretation. LN had the idea for the study and contributed to study design and data interpretation.

gical samples. AV contributed to data review and data interpretation. MGR contributed to study design, data review, and data interpretation. FC contributed to study design, treatment of patients, data review, and data interpretation. LN had the idea for the study and contributed to study design and data interpretation. Declaration of interests LN is an inventor on pending and issued patents on lentiviral vector technology. At the time of the interim study reporting, KvR, EJC-S, and GA were employed by and held stock in GlaxoSmithKline. CD was a contractor at GlaxoSmithKline at the time of the interim study reporting and is now employed and holds stock in Orchard Therapeutics. MHA reports grants, consultancy fees, and stock ownership from GlaxoSmithKline, non-financial support, travel support, and honoraria from Octapharma, honoraria from CSL Behring and MSD, and non-financial support from Medac and Shire, outside of the submitted work. LB reports personal fees from GlaxoSmithKline during the conduct of the study, for consulting activity on safety and integration site analysis for the trial object of the study. FF, MPC, StGa, StGi, FD, FB, MM, MEB, VC, AAA, MF, CF, EA, SaSc, IB, SeSc, LB-R, RP, MC, DC, FAS, AB, HMW, RF, PS, SaGa, AV, MGV, MGR, FC, and AA declare no competing interests.

White C, Noble SIR, Watson M, et al. Prevalence, symptom burden, and natural history of deep vein thrombosis in people with advanced cancer in specialist palliative care units (HIDDen): a prospective longitudinal observational study. Lancet Haematol 2019; 6: e79–88—In figure 1, the number of ineligible patients should have been 825 (59%), among which 28 non-cancer; the number of eligible patients should have been 565, of whom 359 were recruited and 16 excluded because they were repeat admissions. The second sentence of the results section should have been “16 participants were admitted more than once during the study period”. These corrections have been made to the online version as of May 21, 2019.

Introduction INTERVAL was the first randomised trial, to the best of our knowledge, to evaluate the efficiency and safety of varying the frequency of whole blood donation.1, 2, 3 We randomly assigned over 45 000 blood donors recruited across England, UK, to different inter-donation intervals (8, 10, and 12 weeks for men; and 12, 14, and 16 weeks for women) over a period of 2 years with more intensive reminders than standard for NHS Blood and Transplant (NHSBT). During that time, there was a substantial increase in the amount of blood collected by reducing the inter-donation intervals combined with intensive reminders to follow up missed appointments without detectable effects on overall quality of life, physical activity, or cognitive function of the donors.1, 4 Research in context Evidence before this study

Introduction INTERVAL was the first randomised trial, to the best of our knowledge, to evaluate the efficiency and safety of varying the frequency of whole blood donation.1, 2, 3 We randomly assigned over 45 000 blood donors recruited across England, UK, to different inter-donation intervals (8, 10, and 12 weeks for men; and 12, 14, and 16 weeks for women) over a period of 2 years with more intensive reminders than standard for NHS Blood and Transplant (NHSBT). During that time, there was a substantial increase in the amount of blood collected by reducing the inter-donation intervals combined with intensive reminders to follow up missed appointments without detectable effects on overall quality of life, physical activity, or cognitive function of the donors.1, 4 Research in context Evidence before this study We searched for randomised trials published in English from database inception to March 1, 2019, investigating the effect of intensive approaches to help whole blood donors keep appointments, or of varying the inter-donation interval. We searched PubMed, Scientific Citation Index Expanded, and Embase using relevant terms: “blood donation intervals”, “blood donation frequency”, “blood supply”, “donor health”, “appointments”, and “reminders”. Regarding trials of approaches to remind donors to keep appointments, we could not identify any previous relevant studies. Regarding trials of varying the inter-donation interval, we identified only the INTERVAL trial, a trial of 45 263 donors that showed that, over a two-year period, inter-donation intervals for whole blood donation can be safely reduced to meet blood shortages. However, longer-term data are needed to inform policy more appropriately.

als of varying the inter-donation interval, we identified only the INTERVAL trial, a trial of 45 263 donors that showed that, over a two-year period, inter-donation intervals for whole blood donation can be safely reduced to meet blood shortages. However, longer-term data are needed to inform policy more appropriately. Added value of this study As probably the first randomised trial of the effects of giving blood donors intensive reminders to help keep their appointments, the present study should provide unique insight into this question. Regarding the longer-term effects of varying the inter-donation interval, the present study extended the original INTERVAL trial beyond its initial 2-year period for up to a further 2-year period, recording a set of comprehensive outcomes relating to blood donation, clinical safety, and biochemistry. Implications of all the available evidence Our results give policy makers in the UK two additional evidence-based options to meet blood supply needs, that is, the use of frequent reminders to help donors keep appointments and shorter inter-donation intervals than are now standard. Our data also quantify the extent of iron depletion within 4 years of repeated donation, thus informing safety guidelines. Finally, our results suggest a need to review the screening method used in the UK to test individuals' eligibility to donate.

ppointments and shorter inter-donation intervals than are now standard. Our data also quantify the extent of iron depletion within 4 years of repeated donation, thus informing safety guidelines. Finally, our results suggest a need to review the screening method used in the UK to test individuals' eligibility to donate. These results suggested that, over a duration of about 2 years, blood collection services could safely use shorter donation intervals to meet shortages, such as during periods of high demand.5 However, the INTERVAL trial showed that increased donation frequency resulted in a greater number of deferrals (temporary suspension of donors from giving blood) because of low haemoglobin, lower average haemoglobin and ferritin concentrations, and more self-reported symptoms (more self-reported symptoms were seen especially among men).1 Hence, it is important to assess the acceptability and sustainability of varying the frequency of whole blood donation for periods longer than 2 years. We extended the INTERVAL trial for up to a further 2 years to compare the longer-term effects of donating whole blood using standard inter-donation intervals in the UK with shorter inter-donation intervals used in other countries.6, 7, 8 During the extension study, we also compared the use of more intensive reminders to keep blood donation appointments versus the routine reminders used by the NHSBT blood service in England.

hole blood using standard inter-donation intervals in the UK with shorter inter-donation intervals used in other countries.6, 7, 8 During the extension study, we also compared the use of more intensive reminders to keep blood donation appointments versus the routine reminders used by the NHSBT blood service in England. Methods Study design and participants INTERVAL was a parallel group, pragmatic, randomised trial.1, 2, 3 Full details of the INTERVAL trial have been published previously.1, 2, 3 In brief, eligible donors were aged 18 years or older, fulfilled routine criteria for donation, had an email address and access to the internet to respond to web-based questionnaires, and were willing to be randomly assigned to any of the trial's intervention groups at one of the 25 static donor centres of NHSBT, the sole blood provider to the NHS in England, UK.

or older, fulfilled routine criteria for donation, had an email address and access to the internet to respond to web-based questionnaires, and were willing to be randomly assigned to any of the trial's intervention groups at one of the 25 static donor centres of NHSBT, the sole blood provider to the NHS in England, UK. In the main trial, men were randomly assigned to 12-week (standard), 10-week, or 8-week inter-donation intervals, and women to 16-week (standard), 14-week, or 12-week intervals. Randomisation of donors to sex-specific intervention groups in the ratio of 1:1:1 was done at the coordinating centre using a minimisation algorithm to ensure key characteristics (age, weight, and numbers of new vs existing donors) were balanced across trial groups at baseline. Because of the nature of the intervention, it was not possible to mask participants to their allocated inter-donation interval intervention group. During the main trial, donors were followed up for a period of 2 years after randomisation. Routine NHSBT blood donation procedures, including eligibility screening with the copper sulphate test, were adopted because of the pragmatic trial design.

ticipants to their allocated inter-donation interval intervention group. During the main trial, donors were followed up for a period of 2 years after randomisation. Routine NHSBT blood donation procedures, including eligibility screening with the copper sulphate test, were adopted because of the pragmatic trial design. In the extension study reported here, donors nearing completion of their 2-year participation in the main trial were invited by email to continue donating blood at their allocated inter-donation intervals beyond the 2-year period initially agreed (appendix p 22–24). Participants were assigned to active (ie, more intensive) or routine reminders for donation appointments. The active reminder system (as used in the main trial) consisted of a uniform three-step reminder process of email, text message, and telephone call to encourage donation attendance, with a particular focus on donors missing appointments. The routine reminders followed the standard NHSBT protocol, which was less intense (appendix p 22–24). Donors aged 20 years or older were eligible immediately after completion of their 2-year participation in the main trial, provided they could contribute at least 6 months of follow-up before the end of the main trial follow-up study period (ie, June 16, 2016). Participants gave electronic informed consent. The National Research Ethics Service approved (11/EE/0538) this study.

ter completion of their 2-year participation in the main trial, provided they could contribute at least 6 months of follow-up before the end of the main trial follow-up study period (ie, June 16, 2016). Participants gave electronic informed consent. The National Research Ethics Service approved (11/EE/0538) this study. Randomisation and masking Participants were block-randomised within each of the main trial groups (inter-donation interval, men: 12, 10, and 8 weeks; women: 16, 14, and 12 weeks) to active (ie, more intensive) or routine reminders for donation appointments (figure 1). Simple 1:1 randomisation was done by the trial's senior data manager (MW) at the coordinating centre using computer-generated random sequences in block sizes of six or eight within the main trial groups. As was the case in the initial trial period, it was not possible to mask participants in the extended study to their allocated inter-donation interval group because of the nature of the intervention. Participants were not informed of their randomly allocated group in the extension study, although individuals returning to routine reminders might have noticed the change. Donors who did not consent to participate in the extension study returned to NHSBT's standard inter-donation intervals (12 weeks for men, 16 weeks for women) and routine appointment reminders. For these participants, consent given at the beginning of the main trial allowed retrieval of anonymised data for blood donations from NHSBT's national database. During the extension study, only the trial's senior data manager (MW) and study coordinator (CM) knew the allocations to active versus routine reminders for purposes of coordination. Laboratory technicians were unaware of the groups to which participants had been randomised.Figure 1 Trial profile

T's national database. During the extension study, only the trial's senior data manager (MW) and study coordinator (CM) knew the allocations to active versus routine reminders for purposes of coordination. Laboratory technicians were unaware of the groups to which participants had been randomised.Figure 1 Trial profile CONSORT flowchart showing recruitment, participation, and completeness of main outcomes in the extension study. *Participants who were randomised but later withdrew consent for any further use of their data. †Due to staggered roll-out of the main 2-year trial, only participants expected to attend at least two more sessions were considered eligible for invitation to the extension study. ‡Participants not consenting to the extension study reverted to routine NHS Blood and Transplant reminders (men every 12 weeks, women every 16 weeks). §Number for whom a physical component score could be calculated at the end of the extension trial. ¶Number who provided a research blood sample at the end of the extension trial from which haemoglobin and ferritin were measured. ||Number who responded to at least one question in any of the 6-monthly questionnaires administered during their participation in the extension study.

ated at the end of the extension trial. ¶Number who provided a research blood sample at the end of the extension trial from which haemoglobin and ferritin were measured. ||Number who responded to at least one question in any of the 6-monthly questionnaires administered during their participation in the extension study. Procedures The extension study used the same procedures as in the main trial.1, 2, 3 These included online administration of 6-monthly questionnaires to monitor donor safety characteristics, and a final questionnaire and collection of a non-fasting research blood sample at the end of the study. These blood samples were transported to a central laboratory for a full blood count analysis (Sysmex XN-2000 haematology analyser, UK BioCentre, Stockport, UK). Ferritin concentrations were measured in stored serum samples with an immunoturbidimetric assay (Roche/Hitachi chemistry analyser, Stichting Huisartsen Laboratorium, Etten-Leur, Netherlands). As with the main trial, at each attendance, donors underwent routine screening for eligibility to donate blood, including pin-prick haemoglobin screening via a gravimetric method (copper sulphate test), followed by the spectrophotometric HemoCue test (HemoCue AB, Ängelholm, Sweden) with venous blood for individuals who did not pass the copper sulphate test (minimum thresholds to donate in England, UK, are 135 g/L for men and 125 g/L for women).9

ck haemoglobin screening via a gravimetric method (copper sulphate test), followed by the spectrophotometric HemoCue test (HemoCue AB, Ängelholm, Sweden) with venous blood for individuals who did not pass the copper sulphate test (minimum thresholds to donate in England, UK, are 135 g/L for men and 125 g/L for women).9 Outcomes The primary outcome was the number of whole blood donations made during the extension study expressed as units per year, with standard practice being to donate 1 unit of blood per session (full donation unit 470 mL). The primary outcome was assessed in 20 757 randomly assigned participants, by intention-to-treat. Secondary outcomes related to safety were deferrals of donors (ie, temporary rejection) for low haemoglobin and other factors, haemoglobin and ferritin concentrations, quality of life (using physical and mental wellbeing scores from the Short Form Health Survey, version 210), self-reported symptoms potentially related to blood donation (fainting or feeling faint, tiredness, breathlessness, palpitations, dizziness, chest pain, restless legs, reported low iron concentrations, use of iron supplements, pica), cost-effectiveness of reducing donation intervals (not reported here), and other blood cell-related measures at the end of the extension study reported as secondary exploratory outcomes.

breathlessness, palpitations, dizziness, chest pain, restless legs, reported low iron concentrations, use of iron supplements, pica), cost-effectiveness of reducing donation intervals (not reported here), and other blood cell-related measures at the end of the extension study reported as secondary exploratory outcomes. Statistical analysis The statistical analysis followed a prespecified plan for the extension study. The sample size calculation was done for the original trial.1 Data for men and women were analysed separately by the intention-to-treat principle according to their randomised groups. For prespecified subgroup analyses, ferritin values were log transformed and presented as geometric means and used to classify donors as iron depleted (<15 μg/L) according to WHO criteria.11 For all other outcomes, we present means and percentages without adjustment. Analysis of outcomes by active versus routine reminders involved simple differences between groups. For analysis of outcomes by main trial inter-donation interval groups, linear trend was assessed statistically; any non-linearity was identified only graphically. To inform generalisability, we assessed differences in baseline characteristics and outcomes at the end of the main trial, first between individuals who participated in the extension study versus individuals who did not, and second across the main trial randomised inter-donation groups in individuals who took part in the extension study. We compared groups by calculating p values for differences or linear trend using Poisson regression models for rates, normal regression models for continuous outcomes, and logistic regression models for binary outcomes. To minimise potential bias, we adjusted for centre, baseline age, weight, and new donor status, and other covariates when relevant. Precision of estimates were displayed as 95% CIs. For outcomes derived from multiple donation sessions attended, or multiple questionnaires answered by each participant, the 95% CIs were based on robust standard error estimates to avoid optimism in the level precision. Because of the number of statistical tests done, we used the following guidelines for considering whether the results provided strong evidence: p<0·005 for the analyses of whole blood donation rates (ie, the primary outcome), and p<0·0001 for other tests. Analyses were done with Stata, version 13.

the level precision. Because of the number of statistical tests done, we used the following guidelines for considering whether the results provided strong evidence: p<0·005 for the analyses of whole blood donation rates (ie, the primary outcome), and p<0·0001 for other tests. Analyses were done with Stata, version 13. There were no protocol amendments or deviations from the trial protocol. This trial is registered with ISRCTN, number ISRCTN24760606, and has completed. An independent data monitoring committee periodically reviewed summaries of the trial data for safety purposes. Role of the funding source The academic investigators and representatives of NHSBT, a funder of the trial, participated in the study design and oversight. The investigators at the trial's academic coordinating centre had sole access to the trial database, and had final responsibility for data collection, data integrity, data analysis and interpretation, as well as manuscript drafting and the decision to submit the manuscript for publication. All authors gave approval to submit for publication.

e trial's academic coordinating centre had sole access to the trial database, and had final responsibility for data collection, data integrity, data analysis and interpretation, as well as manuscript drafting and the decision to submit the manuscript for publication. All authors gave approval to submit for publication. Results Between Oct 19, 2014, and May 3, 2016, of 45 042 participants who completed the main trial, 38 035 (84·4%; 18 754 men, 19 281 women) were invited to participate in the extension study. Of those invited, 20 757 (54·6%; 10 843 men, 9914 women) consented and were randomly assigned to active versus routine appointment reminders (figure 1, appendix p 12). The percentage of participants invited and those consenting were similar across the main trial's sex-specific randomised inter-donation interval groups (figure 1). Median follow-up during the extension study was 1·1 years (IQR 0·7–1·3).

assigned to active versus routine appointment reminders (figure 1, appendix p 12). The percentage of participants invited and those consenting were similar across the main trial's sex-specific randomised inter-donation interval groups (figure 1). Median follow-up during the extension study was 1·1 years (IQR 0·7–1·3). Participants who consented to the extension study differed from participants who did not in several characteristics recorded at the beginning and during the main trial (appendix pp 3–4). Compared with participants who did not take part, participants were older (by a mean of 7·4 years [95% CI 7·1–7·6]), more committed and adherent within the main trial (donating 79% [95% CI 77–82] more blood), had fewer deferrals, and had a lower frequency of self-reported symptoms (appendix p 3–4). Donation rates in donors who did not take part in the extension study (ie, individuals reverting to standard inter-donation intervals at the end of the main trial) were lower than in individuals who participated (figure 2).Figure 2 Whole blood donation rate during the main trial and in the extension study by sex and inter-donation intervals

rs who did not take part in the extension study (ie, individuals reverting to standard inter-donation intervals at the end of the main trial) were lower than in individuals who participated (figure 2).Figure 2 Whole blood donation rate during the main trial and in the extension study by sex and inter-donation intervals All participants in the main trial were allocated to active reminders. Participants not included in the extension study automatically reverted to standard inter-donation intervals (12 weeks for men, 16 weeks for women) at their completion of the main trial, with anonymised lookup of blood donation information from NHS Blood and Transplant records made possible by consent given at the beginning of the main trial. The blood donation rates for these participants during the period of the extension study are shown according to the original randomised groups, purely for comparison purposes, even though they had all reverted to the standard inter-donation intervals. Error bars denote 95% CI. *Allocated to routine reminders in the extension study. †Allocated to active reminders in the extension study.

the extension study are shown according to the original randomised groups, purely for comparison purposes, even though they had all reverted to the standard inter-donation intervals. Error bars denote 95% CI. *Allocated to routine reminders in the extension study. †Allocated to active reminders in the extension study. Information on the primary outcome was available for 20 717 (99·8%) of 20 757 participants (figure 1). Baseline characteristics of participants were well balanced across the randomised active versus routine reminders trial groups (appendix p 8). Mean whole blood donation rates for active versus routine reminders in men were 3·50 (95% CI 3·45–3·54) versus 3·39 (3·34–3·44) units per year, or a mean difference of 0·11 (95% CI 0·04–0·17; p=0·00028) units per year (figure 2; appendix pp 9, 13). Corresponding results in women were 2·33 (95% CI 2·30–2·37) versus 2·28 (2·24–2·31) units per year, or mean difference of 0·06 units per year (95% CI 0·01–0·11; p=0·0094). No significant differences were observed between the active and routine reminder groups in outcomes related to safety (appendix p 9). The effect of active reminders on blood donation rates did not vary according to inter-donation intervals (figure 2; interaction test p=0·86 in men and p=0·55 in women).

0·11; p=0·0094). No significant differences were observed between the active and routine reminder groups in outcomes related to safety (appendix p 9). The effect of active reminders on blood donation rates did not vary according to inter-donation intervals (figure 2; interaction test p=0·86 in men and p=0·55 in women). From the 20 757 participants, availability of secondary outcomes assessed at the end of the extension study varied: 20 717 (99·8%) for deferrals per donation session attended; 18 638 (89·8%) for self-reported symptoms; 16 388 (79·0%) for physical wellbeing score; 15 572 (75·0%) for haemoglobin and other blood cell measures; and 13 681 (65·9%) for ferritin concentration (figure 1). Availability of these outcomes was broadly similar between randomised groups (appendix p 12). In the participants included in the extension study, the effects of shorter inter-donation intervals during the first 2 years were consistent with the main trial findings, including lower concentrations of haemoglobin and ferritin (appendix p 5). Exploratory analyses showed that shorter inter-donation intervals also led to lower concentrations of other commonly assessed haematological variables at the end of the main trial (appendix p 6). In this subset of participants, however, there was no evidence of the effects of shorter inter-donation intervals on self-reported symptoms (eg, tiredness, feeling faint, dizziness, breathlessness), although there was a higher reported frequency of doctor-diagnosed low iron concentrations and prescription of iron supplements in men (both p<0·0001; appendix p 7).

re was no evidence of the effects of shorter inter-donation intervals on self-reported symptoms (eg, tiredness, feeling faint, dizziness, breathlessness), although there was a higher reported frequency of doctor-diagnosed low iron concentrations and prescription of iron supplements in men (both p<0·0001; appendix p 7). Baseline characteristics of participants were broadly similar across the inter-donation interval groups (appendix p 5). Donors continuing to donate at shorter inter-donation intervals gave more blood during the extension study than individuals continuing on the longer intervals (men an extra 0·23 units per year [95% CI 0·21–0·25], women an extra 0·14 units per year [0·12–0·15], per week shorter interval based on linear trend, both p<0·0001; figure 3, table 1). There were no clear differences across trial groups in physical and mental wellbeing scores or reported frequency of self-reported symptoms other than a higher reported frequency of doctor diagnosed low iron concentrations and prescription of iron supplements in men (table 1; appendix pp 10, 14). Similarly, there were no differences across trial groups in the frequency of serious adverse events (eg, heart failure, myocardial infarction, stroke, falls, or transport accidents; table 2). However, donors allocated to shorter inter-donation intervals had more deferrals for low haemoglobin (odds ratio per week shorter inter-donation interval 1·19 [95% CI 1·15–1·22] in men and 1·10 [1·06–1·14] in women), and had lower mean haemoglobin (difference per week shorter inter-donation interval −0·84 g/L [95% CI −0·99 to −0·70] in men and −0·45 g/L [–0·59 to −0·31] in women) and ferritin concentrations (percentage difference per week shorter inter-donation interval −6·5% [95% CI −7·6 to −5·5] in men and −5·3% [–6·5 to −4·2] in women) at the end of the extension study (all p<0·0001; table 1, figure 4; appendix p 15). Shorter inter-donation intervals also led to lower concentrations of other commonly assessed haematological variables at the end of the extension study (appendix p 11).

·5% [95% CI −7·6 to −5·5] in men and −5·3% [–6·5 to −4·2] in women) at the end of the extension study (all p<0·0001; table 1, figure 4; appendix p 15). Shorter inter-donation intervals also led to lower concentrations of other commonly assessed haematological variables at the end of the extension study (appendix p 11). The proportion of individuals donating blood with haemoglobin concentrations less than the minimum regulatory threshold and individuals with ferritin less than 15 μg/L was higher in donors allocated to shorter intervals than in individuals allocated to the standard donation intervals (appendix p 16).Figure 3 Whole blood donation rates during the extension study, the main trial period, and in the previous 2 years by sex and inter-donation intervals The p values compare across inter-donation intervals and are adjusted for baseline characteristics (centre, age, weight, new donor status). Minimum inter-donation intervals allowed before the trial were 12 weeks for men and 16 weeks for women. Error bars denote 95% CI. Table 1 Outcomes during the extension study by sex and inter-donation groups

The p values compare across inter-donation intervals and are adjusted for baseline characteristics (centre, age, weight, new donor status). Minimum inter-donation intervals allowed before the trial were 12 weeks for men and 16 weeks for women. Error bars denote 95% CI. Table 1 Outcomes during the extension study by sex and inter-donation groups Men Women 8 weeks 10 weeks 12 weeks p value* 12 weeks 14 weeks 16 weeks p value* Participants† 3554 (33%) 3695 (34%) 3594 (33%) ·· 3369 (34%) 3309 (33%) 3236 (33%) ·· Follow-up time, years (median, IQR) 1·2 (0·8–1·3) 1·2 (0·8–1·3) 1·1 (0·7–1·3) ·· 1·1 (0·6–1·3) 1·1 (0·6–1·3) 1·0 (0·6–1·2) ·· Whole blood donation rate (times per year) 3·90 (3·83–3·96) 3·44 (3·38–3·49) 2·97 (2·93–3·02) <0·0001 2·57 (2·53–2·62) 2·29 (2·25–2·33) 2·02 (1·99–2·06) <0·0001 Deferral for low haemoglobin‡ 5·94% (5·56–6·33) 4·43% (4·07–4·79) 3·04% (2·71–3·38) <0·0001 6·22% (5·70–6·74) 5·19% (4·68–5·70) 4·42% (3·92–4·93) <0·0001 Deferral for other reasons‡ 3·32% (3·04–3·60) 3·71% (3·39–4·04) 3·76% (3·40–4·12) 0·064 4·25% (3·84–4·67) 5·18% (4·68–5·68) 5·20% (4·66–5·75) 0·003 Fainting at donation session‡ 0·15% (0·09–0·21) 0·18% (0·11–0·26) 0·17% (0·09–0·24) 0·61 0·52% (0·37–0·67) 0·45% (0·30–0·60) 0·48% (0·30–0·66) 0·68 SF-36 physical wellbeing score 56·5 (56·3–56·7) 56·6 (56·4–56·7) 56·4 (56·3–56·6) 0·94 56·6 (56·4–56·8) 56·4 (56·2–56·7) 56·3 (56·1–56·5) 0·11 SF-36 mental wellbeing score 54·3 (54·0–54·5) 54·2 (54·0–54·4) 54·1 (53·8–54·3) 0·63 53·3 (53·0–53·6) 53·2 (52·9–53·5) 53·0 (52·7–53·2) 0·077 Haemoglobin (g/L) 140·8 (140·3–141·2) 142·7 (142·3–143·1) 144·2 (143·8–144·6) <0·0001 130·0 (129·6–130·4) 130·8 (130·4–131·2) 131·8 (131·4–132·2) <0·0001 Haemoglobin <135 g/L (men) or <125 g/L (women)§ 22·29% (20·38–24·19) 16·26% (14·72–17·81) 14·01% (12·58–15·43) <0·0001 21·80% (19·87–23·73) 18·81% (17·00–20·62) 16·81% (15·14–18·49) <0·0001 Ferritin (μg/L)¶ 26·3 (25·5–27·2) 30·3 (29·4–31·3) 34·5 (33·5–35·6) <0·0001 22·6 (21·8–23·4) 25·5 (24·7–26·4) 28·2 (27·2–29·1) <0·0001 Ferritin <15 μg/L§ 21·19% (19·20–23·18) 16·41% (14·75–18·08) 11·87% (10·46–13·28) <0·0001 25·00% (22·85–27·15) 20·04% (18·07–22·01) 18·46% (16·62–20·30) <0·0001 Serious adverse events‖ 2·35% (1·83–2·88) 2·75% (2·19–3·31) 2·88% (2·30–3·45) 0·25 2·73% (2·15–3·32) 3·48% (2·82–4·15) 3·21% (2·57–3·86) 0·30 Data are mean or percentage (95% CI) unless otherwise stated. SF-36=36-item short-form health survey.

·0001 25·00% (22·85–27·15) 20·04% (18·07–22·01) 18·46% (16·62–20·30) <0·0001 Serious adverse events‖ 2·35% (1·83–2·88) 2·75% (2·19–3·31) 2·88% (2·30–3·45) 0·25 2·73% (2·15–3·32) 3·48% (2·82–4·15) 3·21% (2·57–3·86) 0·30 Data are mean or percentage (95% CI) unless otherwise stated. SF-36=36-item short-form health survey. * p values are for linear trend across groups, from analyses adjusted for baseline characteristics (centre, age, weight, new donor status) and value of the outcome at baseline (when available). † Additional missing data during the extension study were: <0·2% for blood donation, deferrals, or fainting; 21·0% for SF-36 Physical/Mental wellbeing scores, 25·0% for haemoglobin, 34·1% for ferritin. Higher SF-36 scores indicate better physical or mental wellbeing (0–100 scale range). ‡ Deferral or fainting rate per donation session attended during the extension study. § Among individuals donating blood at end of the extension study. ¶ Values are geometric means. ‖ Percentage of participants reporting any serious adverse events during the extension study, in any of the 6-monthly questionnaires, including doctor-confirmed heart failure, heart attack, angina, stroke, or transient ischaemic attack; and hospital visit for falls or transport accidents. Table 2 Adverse events during the extension study by inter-donation interval groups

‖ Percentage of participants reporting any serious adverse events during the extension study, in any of the 6-monthly questionnaires, including doctor-confirmed heart failure, heart attack, angina, stroke, or transient ischaemic attack; and hospital visit for falls or transport accidents. Table 2 Adverse events during the extension study by inter-donation interval groups Grade* Overall Men (n=10 843 in extension) Women (n=9914 in extension) N n (%) 8 weeks 10 weeks 12 weeks 12 weeks 14 weeks 16 weeks Any self-reported serious adverse events† ·· 18 550 536 (2·9%) 75 (2·4%) 91 (2·8%) 93 (2·9%) 82 (2·7%) 102 (3·5%) 93 (3·2%) Doctor diagnosed heart problems 3 18 528 69 (0·4%) 14 (0·4%) 20 (0·6%) 17 (0·5%) 4 (0·1%) 5 (0·2%) 9 (0·3%) Doctor diagnosed heart failure 3 18 528 18 (0·1%) 5 (0·2%) 4 (0·1%) 5 (0·2%) 1 (0·0%) 1 (0·0%) 2 (0·1%) Doctor diagnosed heart attack 3 18 526 20 (0·1%) 7 (0·2%) 5 (0·2%) 4 (0·1%) 1 (0·0%) 0 (0·0%) 3 (0·1%) Doctor diagnosed angina 3 18 526 24 (0·1%) 6 (0·2%) 6 (0·2%) 5 (0·2%) 1 (0·0%) 4 (0·1%) 2 (0·1%) Doctor diagnosed stroke 3 18 527 17 (0·1%) 4 (0·1%) 7 (0·2%) 1 (0·0%) 2 (0·1%) 1 (0·0%) 2 (0·1%) Doctor diagnosed transient ischaemic attack 3 18 527 21 (0·1%) 5 (0·2%) 5 (0·2%) 5 (0·2%) 2 (0·1%) 1 (0·0%) 3 (0·1%) Visit to hospital for a fall 3 18 533 337 (1·8%) 39 (1·2%) 39 (1·2%) 47 (1·5%) 64 (2·1%) 81 (2·8%) 67 (2·3%) Visit to hospital for transport accident 3 18 516 150 (0·8%) 26 (0·8%) 36 (1·1%) 36 (1·1%) 17 (0·6%) 17 (0·6%) 18 (0·6%) Any symptom self-reported 1–2 18 554 9732 (52·5%) 1581 (49·6%) 1556 (47·1%) 1476 (45·6%) 1764 (58·8%) 1699 (58·0%) 1656 (57·2%) Fainting or feeling faint 1–2 18 534 2085 (11·3%) 325 (10·2%) 302 (9·2%) 269 (8·3%) 424 (14·1%) 390 (13·3%) 375 (13·0%) More tired than usual 1–2 18 537 5198 (28·0%) 864 (27·1%) 823 (24·9%) 800 (24·8%) 947 (31·6%) 881 (30·1%) 883 (30·5%) Palpitations 1–2 18 502 2217 (12·0%) 271 (8·5%) 286 (8·7%) 261 (8·1%) 498 (16·6%) 435 (14·9%) 466 (16·2%) Dizziness 1–2 18 533 3197 (17·3%) 457 (14·4%) 456 (13·8%) 419 (13·0%) 657 (21·9%) 617 (21·1%) 591 (20·4%) Restless legs syndrome 1–2 18 464 4158 (22·5%) 642 (20·2%) 642 (19·5%) 613 (19·0%) 776 (26·0%) 753 (25·8%) 732 (25·4%) Data presented are n (%) unless otherwise stated. Adverse events listed in this table were ascertained only through self-report questionnaires and mapped to Common Terminology Criteria for Adverse Events grading using heuristic criteria. For adverse events of grade 1–2, only those occurring in 10% or more of patients are reported.

ed are n (%) unless otherwise stated. Adverse events listed in this table were ascertained only through self-report questionnaires and mapped to Common Terminology Criteria for Adverse Events grading using heuristic criteria. For adverse events of grade 1–2, only those occurring in 10% or more of patients are reported. * Grading with reference to Common Terminology Criteria for Adverse Events version 5.0. † Number and percentage of participants reporting any serious adverse events during the extension study in any of the 6-monthly questionnaires, including doctor-confirmed heart failure, heart attack, angina, stroke, or transient ischaemic attack; or hospital visit for falls or transport accidents. Study participants could contribute to more than one outcome in this table. Figure 4 Haemoglobin (A) and ferritin (B) concentrations at the end of the extension study, end of the main trial period, and at baseline by sex and inter-donation intervals Analysis is restricted to participants in the extension study. The p values assess trends across inter-donation intervals, adjusted for baseline characteristics (centre, age, weight, new donor status, and haemoglobin [A] or loge ferritin [B]). Error bars denote 95% CI.

Figure 4 Haemoglobin (A) and ferritin (B) concentrations at the end of the extension study, end of the main trial period, and at baseline by sex and inter-donation intervals Analysis is restricted to participants in the extension study. The p values assess trends across inter-donation intervals, adjusted for baseline characteristics (centre, age, weight, new donor status, and haemoglobin [A] or loge ferritin [B]). Error bars denote 95% CI. During the extension study, blood donation rates in each trial group were 14·6% (95% CI 13·1–14·2) lower than during the main trial (figure 3). In comparison with the main trial, frequency of self-reported symptoms and rates of deferral for low haemoglobin increased further (appendix pp 14–15), while mean haemoglobin concentrations decreased further (figure 4A), especially in men. By contrast, mean ferritin concentrations increased somewhat, especially in women (figure 4B). Corresponding changes in other haematological variables (appendix p 17) showed similar results to haemoglobin for some traits (eg, lower mean corpuscular haemoglobin and mean corpuscular haemoglobin concentration) and similar to ferritin for other traits (eg, higher mean haematocrit, mean corpuscular volume, and reticulocyte haemoglobin equivalent). There was no evidence that laboratory machine drift or technical errors, as judged by evaluation of internal quality control samples, could explain differences in the above-mentioned variables.

to ferritin for other traits (eg, higher mean haematocrit, mean corpuscular volume, and reticulocyte haemoglobin equivalent). There was no evidence that laboratory machine drift or technical errors, as judged by evaluation of internal quality control samples, could explain differences in the above-mentioned variables. The proportion of donors who reported that their doctor had prescribed iron supplements increased through the duration of the INTERVAL trial up to 4·0% (95% CI 3·7–4·4) by the end of the extension study, and together with individuals reporting the use of over-the-counter iron supplements, comprised 1396 (16%) of 8594 men and 1732 (22%) of 7803 women by the end of the extension study, with higher proportions in donors allocated to shorter intervals (appendix p 18). In post-hoc analyses, which stratified comparisons according to patterns of reported use of any iron supplements during the main trial or extension study, the decrease in mean haemoglobin concentrations was larger (appendix p 19) and the increase in mean ferritin concentrations no longer apparent (appendix p 20) among the participants who did not report using iron supplements throughout the trial. Similarly, stratified post-hoc results for reticulocyte haemoglobin concentration (appendix p 21) showed increased concentrations during the extension study even in the subgroup of participants who were iron supplement-naive.

x p 20) among the participants who did not report using iron supplements throughout the trial. Similarly, stratified post-hoc results for reticulocyte haemoglobin concentration (appendix p 21) showed increased concentrations during the extension study even in the subgroup of participants who were iron supplement-naive. Discussion This trial extended the intervention and follow-up periods of INTERVAL, a randomised trial of varying inter-donation intervals in whole blood donors. We also did a randomised comparison of different approaches to remind donors to keep blood donation appointments. This extension study's main result was that, over a period of 2–4 years, shorter inter-donation intervals and more intensive reminders resulted in more blood being collected.

ntervals in whole blood donors. We also did a randomised comparison of different approaches to remind donors to keep blood donation appointments. This extension study's main result was that, over a period of 2–4 years, shorter inter-donation intervals and more intensive reminders resulted in more blood being collected. Our trial was notable because it quantified key measures that blood services aim to balance in maintaining the blood supply while safeguarding the health of donors. Our extension study showed that, beyond a 2-year period, each week reduction in time between donations led to an increase of 0·23 units in the amount of blood collected per year in men and of 0·14 units in women compared with the donation intervals currently used in the UK (ie, 12 weeks in men and 16 weeks in women).6 With regard to use of more intensive approaches to remind donors of appointments, our study showed a mean increase of 0·11 units of blood per year in men and of 0·06 units of blood per year in women. These modest increases due to additional reminders could potentially translate to collection of an approximate extra 75 000 units of blood from a donor base of 900 000 with about 47% of men and 53% of women (ie, the approximate size of the current donor base in England, UK). If more intensive reminders (eg, a telephone call when an appointment is missed) could be done at little additional cost, then the gain in the amount of blood collected could be worthwhile, at least for priority blood groups.12, 13, 14 The cost implications of a range of alternative policies to encourage blood donation, partly based on the INTERVAL trial, have been published elsewhere.15

is missed) could be done at little additional cost, then the gain in the amount of blood collected could be worthwhile, at least for priority blood groups.12, 13, 14 The cost implications of a range of alternative policies to encourage blood donation, partly based on the INTERVAL trial, have been published elsewhere.15 Regarding safety, the trial showed that reducing inter-donation intervals during the extension study did not have major adverse effects on self-reported mental and physical wellbeing, specific symptoms potentially related to blood donation, or in other major adverse events we recorded. These results extend those from the main trial showing that reducing inter-donation intervals did not result in major adverse events or impaired wellbeing.1, 8 However, when compared with the initial 2 years of the trial, the proportion of donors reporting specific symptoms increased during the extension study, suggesting a potential cumulative effect over a longer period of time.

ucing inter-donation intervals did not result in major adverse events or impaired wellbeing.1, 8 However, when compared with the initial 2 years of the trial, the proportion of donors reporting specific symptoms increased during the extension study, suggesting a potential cumulative effect over a longer period of time. Use of shorter donation intervals during the extension study also resulted in changes in biomarkers of iron homoeostasis, resulting in more deferrals for low haemoglobin, decreased mean haemoglobin and serum ferritin concentrations, and changes in other red blood cell parameters suggesting lower iron availability and lower incorporation into red blood cells.16 As observed for the main trial, there were modest absolute decreases in mean haemoglobin concentrations and other red blood cell parameters at the end of the extension study. By contrast, proportional reductions were larger for serum ferritin, with up to 21% of men and 25% of women with serum ferritin concentrations less than 15 g/L at the end of the extension study. These results are consistent with previous observational studies, suggesting that shorter inter-donation intervals are associated with sustained and progressively lower iron availability.17, 18 However, although shorter donation intervals resulted in further decreases in haemoglobin levels in the extension study, serum ferritin concentrations actually increased somewhat (in parallel with increases in the haemoglobin concentration of reticulocytes). Exploratory analyses suggest that this result could be explained by the higher proportion of donors who reported using iron supplements,19, 20 as by the end of the extension study 16% of men and 22% of women had either been prescribed iron supplements, or reported taking over-the-counter iron supplements.21, 22

cytes). Exploratory analyses suggest that this result could be explained by the higher proportion of donors who reported using iron supplements,19, 20 as by the end of the extension study 16% of men and 22% of women had either been prescribed iron supplements, or reported taking over-the-counter iron supplements.21, 22 Our findings could have several potential implications for blood donation practice and policy. First, our results provide evidence for the long-term safety of more frequent donation and give policy makers in the UK the option to allow more frequent collection from donors than is now standard.6 Nevertheless, total reliance on this strategy might make a blood service overly dependent on a subgroup of donors who are the most resilient to iron depletion, either biologically or through iron supplementation.23 Another option would be to use shorter inter-donation intervals only for more resilient donors, if such donors could be identified in advance by demographic, haematological, or genetic characteristics.1 Second, our data provide convincing evidence of the cumulative effect on haemoglobin concentrations and iron stores of donating blood frequently, which should inform safety guidelines for blood services that allow more frequent donation than in the UK (eg, USA, France, and Germany). Our results support the recent changes in the Canadian Blood Services that have increased the minimum inter-donation interval in women to reduce iron deficiency and deferrals for low haemoglobin.24

form safety guidelines for blood services that allow more frequent donation than in the UK (eg, USA, France, and Germany). Our results support the recent changes in the Canadian Blood Services that have increased the minimum inter-donation interval in women to reduce iron deficiency and deferrals for low haemoglobin.24 Third, given the decrease in haemoglobin concentrations we observed over a longer period, it is essential for blood services to protect the health of donors by adopting appropriate screening methods to test donors' eligibility to donate whole blood.25 To evaluate the relative merits of different screening methods in the context of NHSBT, the COMPARE study (ISRCTN90871183) aims to provide a systematic, within-person comparison of different methods to measure haemoglobin concentrations in whole blood donors to inform approaches for routine eligibility checks in England, UK. Furthermore, other blood services have implemented or are evaluating additional approaches to detect iron deficiency, such as ferritin monitoring in selected blood donors.22, 26 Fourth, our findings underscore the potential benefits of effective communication with blood donors to encourage attendance, especially in an appointments-based system such as used by NHSBT in England, UK.

Third, given the decrease in haemoglobin concentrations we observed over a longer period, it is essential for blood services to protect the health of donors by adopting appropriate screening methods to test donors' eligibility to donate whole blood.25 To evaluate the relative merits of different screening methods in the context of NHSBT, the COMPARE study (ISRCTN90871183) aims to provide a systematic, within-person comparison of different methods to measure haemoglobin concentrations in whole blood donors to inform approaches for routine eligibility checks in England, UK. Furthermore, other blood services have implemented or are evaluating additional approaches to detect iron deficiency, such as ferritin monitoring in selected blood donors.22, 26 Fourth, our findings underscore the potential benefits of effective communication with blood donors to encourage attendance, especially in an appointments-based system such as used by NHSBT in England, UK. Our study had strengths. Because we evaluated the long-term safety and efficiency of frequent donation beyond a 2-year period in a randomised study, our trial provides more reliable insights than do observational studies, which are susceptible to confounding. The trial recorded a comprehensive set of outcomes relating to blood donation, clinical safety, and biochemistry, and provided almost complete outcome data for amount of blood collected and deferrals because of low haemoglobin.

more reliable insights than do observational studies, which are susceptible to confounding. The trial recorded a comprehensive set of outcomes relating to blood donation, clinical safety, and biochemistry, and provided almost complete outcome data for amount of blood collected and deferrals because of low haemoglobin. The study also had limitations. Continuation into the extension study was accepted by 55% of those invited, and therefore analyses are less powerful than in the main trial. Although the participants in the main trial were broadly representative of the national donor population in England, UK, individuals in the extension study were an older and more committed subset of blood donors; they had also had fewer deferrals for low haemoglobin and reported fewer symptoms. Hence, caution is needed in extrapolating the findings to the general population of blood donors. For example, more intensive reminders could yield even more blood donations in less selected groups than our enthusiastic donors who decided to enrol in the extension study (who tend to miss few opportunities to give blood anyway).27, 28

is needed in extrapolating the findings to the general population of blood donors. For example, more intensive reminders could yield even more blood donations in less selected groups than our enthusiastic donors who decided to enrol in the extension study (who tend to miss few opportunities to give blood anyway).27, 28 During the extension study, half of the participants were switched from active to routine reminders, a switch which could explain a small part of why blood donation during the extension study decreased by about 15% compared with the initial 2 years of the trial. However, drivers of the decreased donation rate between the main trial and the extension study could not be established given the study design.29 Although participants were not informed of their randomly allocated group in the extension study, individuals returning to routine reminders might have noticed the change and potentially be influenced by the active reminders from the main trial. The study relied on self-reported information for some outcomes (eg, symptoms), which might be susceptible to reporting biases and incompleteness (ie, missing data). We did not have accurate information from the 6-monthly questionnaires about the timing of reported iron supplement use, and therefore could not distinguish whether it might be related to previous deferral or subsequent donation.

ms), which might be susceptible to reporting biases and incompleteness (ie, missing data). We did not have accurate information from the 6-monthly questionnaires about the timing of reported iron supplement use, and therefore could not distinguish whether it might be related to previous deferral or subsequent donation. In summary, during a period of 2–4 years, collection of substantially more blood without a detectable effect on donors' mental and physical wellbeing was achieved through more frequent donation than is standard practice in the UK and more intensive reminders to keep blood donation appointments. However, compared with the initial 2 years of the trial, extension of this approach resulted in further lowering of haemoglobin concentrations, more deferrals, and higher rates of self-reported symptoms.

quent donation than is standard practice in the UK and more intensive reminders to keep blood donation appointments. However, compared with the initial 2 years of the trial, extension of this approach resulted in further lowering of haemoglobin concentrations, more deferrals, and higher rates of self-reported symptoms. Data sharing The INTERVAL Study Group has previously published its trial protocol, statistical analysis plan, informed consent form, and other relevant study documents. Bona fide scientists can seek access to relevant de-identified individual participant data (and a copy of the trial's data dictionary) by applying to the INTERVAL Data Access Committee after print publication of the current manuscript at the following: helpdesk@intervalstudy.org.uk. The INTERVAL Data Access Committee review (supplemented, when required, by expertise from additional scientists external to the Committee) applications according to usual academic criteria of scientific validity and feasibility. Following approval by the INTERVAL Data Access Committee, a material transfer or research collaboration agreement will be agreed and signed with the applicants. Applicants might be requested to provide reimbursement of data management or preparation costs, as the INTERVAL trial is no longer in receipt of funding. Applicants will be required to provide updates to the INTERVAL Data Access Committee on their use of the INTERVAL trial data, including provision of copies of any publications. Applicants will be required to adhere in publications with the INTERVAL trial's policy for acknowledgment of the trial's funders, stakeholders, and scientific or technical contributors.

ates to the INTERVAL Data Access Committee on their use of the INTERVAL trial data, including provision of copies of any publications. Applicants will be required to adhere in publications with the INTERVAL trial's policy for acknowledgment of the trial's funders, stakeholders, and scientific or technical contributors. Supplementary Material Supplementary appendix

ates to the INTERVAL Data Access Committee on their use of the INTERVAL trial data, including provision of copies of any publications. Applicants will be required to adhere in publications with the INTERVAL trial's policy for acknowledgment of the trial's funders, stakeholders, and scientific or technical contributors. Supplementary Material Supplementary appendix Acknowledgments Participants in the INTERVAL randomised controlled trial were recruited with the active collaboration of NHS Blood and Transplant (NHSBT) England, which has supported field work and other elements of the trial. DNA extraction and genotyping was co-funded by the National Institute for Health Research (NIHR), the NIHR BioResource, and the NIHR Cambridge Biomedical Research Centre at the Cambridge University Hospitals NHS Foundation Trust. The academic coordinating centre for INTERVAL was supported by core funding from: NIHR Blood and Transplant Research Unit in Donor Health and Genomics (NIHR BTRU-2014-10024), UK Medical Research Council (G0800270, MR/L003120/1), British Heart Foundation (SP/09/002, RG/13/13/30194) and the NIHR Cambridge Biomedical Research Centre at the Cambridge University Hospitals NHS Foundation Trust. Investigators at the University of Oxford, Oxford, UK, have been supported by the Research and Development Programme of NHSBT, the NHSBT Howard Ostin Trust Fund, and the NIHR Oxford Biomedical Research Centre through the programme grant NIHR-RP-PG-0310-1004. We thank the blood donors who participated in the trial and NHSBT's operational staff. We are indebted to NHSBT's leadership team (individual acknowledgments are provided in the main trial publication). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care.

onors who participated in the trial and NHSBT's operational staff. We are indebted to NHSBT's leadership team (individual acknowledgments are provided in the main trial publication). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care. Contributors All authors contributed to data collection, study design, data analysis, interpretation, and drafting of this paper. The INTERVAL Trial Group members Writing Committee: Stephen Kaptoge PhD*, University of Cambridge; Emanuele Di Angelantonio FRCP*, University of Cambridge and NHSBT; Carmel Moore PhD, University of Cambridge; Matthew Walker PhD, University of Cambridge; Jane Armitage FRCP, University of Oxford; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; David J Roberts FRCPath†, University of Oxford and NHSBT; John Danesh FMedSci†, University of Cambridge; Simon G Thompson FMedSci†, University of Cambridge.

of Cambridge; Matthew Walker PhD, University of Cambridge; Jane Armitage FRCP, University of Oxford; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; David J Roberts FRCPath†, University of Oxford and NHSBT; John Danesh FMedSci†, University of Cambridge; Simon G Thompson FMedSci†, University of Cambridge. Trial Steering Committee: Jane Armitage FRCP (independent chair), University of Oxford; John Danesh FMedSci, University of Cambridge; Emanuele Di Angelantonio FRCP, University of Cambridge and NHSBT; Jenny Donovan FMedSci (independent member), University of Bristol; Ian Ford FRSE (independent member), University of Glasgow; Rachel Henry, University of Cambridge; Beverley J Hunt FRCPath (independent member), King's College, London; Bridget le Huray (lay member); Susan Mehenny, NHSBT; Gail Miflin FRCPath, NHSBT; Carmel Moore PhD, University of Cambridge; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; Jane Green, NHSBT; David J Roberts FRCPath, University of Oxford and NHSBT; Mike Stredder, NHSBT; Simon G Thompson FMedSci, University of Cambridge; Matthew Walker PhD, University of Cambridge; Nicholas A Watkins PhD, NHSBT. Previous members: Alan McDermott, NHSBT; Clive Ronaldson, NHSBT; Claire Thomson MSc, University of Cambridge; Zoe Tolkien PhD, University of Cambridge; Lorna Williamson FRCP, NHSBT.

Trial Steering Committee: Jane Armitage FRCP (independent chair), University of Oxford; John Danesh FMedSci, University of Cambridge; Emanuele Di Angelantonio FRCP, University of Cambridge and NHSBT; Jenny Donovan FMedSci (independent member), University of Bristol; Ian Ford FRSE (independent member), University of Glasgow; Rachel Henry, University of Cambridge; Beverley J Hunt FRCPath (independent member), King's College, London; Bridget le Huray (lay member); Susan Mehenny, NHSBT; Gail Miflin FRCPath, NHSBT; Carmel Moore PhD, University of Cambridge; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; Jane Green, NHSBT; David J Roberts FRCPath, University of Oxford and NHSBT; Mike Stredder, NHSBT; Simon G Thompson FMedSci, University of Cambridge; Matthew Walker PhD, University of Cambridge; Nicholas A Watkins PhD, NHSBT. Previous members: Alan McDermott, NHSBT; Clive Ronaldson, NHSBT; Claire Thomson MSc, University of Cambridge; Zoe Tolkien PhD, University of Cambridge; Lorna Williamson FRCP, NHSBT. Trial Management Committee: David Allen DPhil, University of Oxford and NHSBT; John Danesh FMedSci, University of Cambridge; Emanuele Di Angelantonio FRCP, University of Cambridge and NHSBT; Rachel Henry, University of Cambridge; Susan Mehenny, NHSBT; Carmel Moore PhD, University of Cambridge; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; David J Roberts FRCPath, University of Oxford and NHSBT; Jennifer Sambrook PhD, University of Cambridge; Matthew Walker PhD, University of Cambridge.

SBT; Rachel Henry, University of Cambridge; Susan Mehenny, NHSBT; Carmel Moore PhD, University of Cambridge; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; David J Roberts FRCPath, University of Oxford and NHSBT; Jennifer Sambrook PhD, University of Cambridge; Matthew Walker PhD, University of Cambridge. Previous members: Tracey Hammerton, NHSBT; Claire Thomson MSc, University of Cambridge; Zoe Tolkien PhD, University of Cambridge. Haematology Review Group: David Allen DPhil, University of Oxford and NHSBT; David Bruce FRCPath, Oxford University Hospitals NHS Foundation Trust; Fizzah Choudry MRCP, University of Cambridge; Emanuele Di Angelantonio FRCP, University of Cambridge and NHSBT; Cedric Ghevaert FRCPath, University of Cambridge; Kirstie Johnston, NHSBT; Anne Kelly PhD, University of Cambridge; Andrew King FRCPath, Weatherall Institute of Molecular Medicine, University of Oxford; Susan Mehenny, NHSBT; Gail Miflin FRCPath, NHSBT; Alfred Mo MRCGP, NHSBT; Carmel Moore PhD, University of Cambridge; Willem H Ouwehand (co-chair) FMedSci, University of Cambridge and NHSBT; Lizanne Page LMSSA, NHSBT; Penny Richardson, NHSBT; David J Roberts (co-chair) FRCPath, University of Oxford and NHSBT; Jennifer Sambrook PhD, University of Cambridge; Peter Senior, NHSBT; Yagnesh Umrania PhD, University of Cambridge; Matthew Walker PhD, University of Cambridge; Henna Wong FRCPath, Oxford University Hospitals NHS Foundation Trust. Restless legs syndrome adviser: Brendan Burchell PhD, University of Cambridge.

Haematology Review Group: David Allen DPhil, University of Oxford and NHSBT; David Bruce FRCPath, Oxford University Hospitals NHS Foundation Trust; Fizzah Choudry MRCP, University of Cambridge; Emanuele Di Angelantonio FRCP, University of Cambridge and NHSBT; Cedric Ghevaert FRCPath, University of Cambridge; Kirstie Johnston, NHSBT; Anne Kelly PhD, University of Cambridge; Andrew King FRCPath, Weatherall Institute of Molecular Medicine, University of Oxford; Susan Mehenny, NHSBT; Gail Miflin FRCPath, NHSBT; Alfred Mo MRCGP, NHSBT; Carmel Moore PhD, University of Cambridge; Willem H Ouwehand (co-chair) FMedSci, University of Cambridge and NHSBT; Lizanne Page LMSSA, NHSBT; Penny Richardson, NHSBT; David J Roberts (co-chair) FRCPath, University of Oxford and NHSBT; Jennifer Sambrook PhD, University of Cambridge; Peter Senior, NHSBT; Yagnesh Umrania PhD, University of Cambridge; Matthew Walker PhD, University of Cambridge; Henna Wong FRCPath, Oxford University Hospitals NHS Foundation Trust. Restless legs syndrome adviser: Brendan Burchell PhD, University of Cambridge. Cognitive function assessment adviser: John Gallacher FFPH, University of Oxford and University of Cardiff. Independent Data Monitoring Committee: Stephen Kaptoge PhD (independent statistician), University of Cambridge; Gavin Murphy FRCS, University of Leicester, Adrian C Newland FRCPath (chair), Barts and The London NHS Trust, Queen Mary's School of Medicine and Dentistry; Keith Wheatley DPhil, University of Birmingham.

Cognitive function assessment adviser: John Gallacher FFPH, University of Oxford and University of Cardiff. Independent Data Monitoring Committee: Stephen Kaptoge PhD (independent statistician), University of Cambridge; Gavin Murphy FRCS, University of Leicester, Adrian C Newland FRCPath (chair), Barts and The London NHS Trust, Queen Mary's School of Medicine and Dentistry; Keith Wheatley DPhil, University of Birmingham. Previous members: Michael Greaves FRCPath (chair), University of Aberdeen; Marc Turner FRCPath, Scottish National Blood Transfusion Service. NHSBT Donation Centres (managers): Birmingham (Tahir Aziz and Richard Brain); Bradford (Christine Davies and Ruth Turner); Brentwood (Paula Wakeman); Bristol (Alison Dent); Cambridge (Alan Wakeman); Edgware (Ben Anthony, Desmond Bland, and Will Parrondo); Gloucester (Helen Vincent); Lancaster (Candy Weatherill); Leeds CBTU (Andrea Forsyth); Leeds City (Carol Butterfield); Leicester (Tracey Wright and Karen Ellis); Liverpool (Kristie Johnston and Pat Poynton); Luton (Carolyn Brooks and Emma Martin); Manchester Norfolk House (Lara Littler and Lindsay Williams); Manchester Plymouth Grove (Donna Blair and Karen Ackerley); Newcastle (Lynn Woods); Nottingham (Sophie Stanley and Gemma Walsh); Oxford (Gayle Franklin and Cheryl Howath); Poole (Sarah Sharpe); Plymouth (Deborah Smith); Sheffield (Lauren Botham); Southampton (Caroline Williams and Claire Alexander); Stoke (Gareth Sowerbutts and Diane Furnival); Tooting (Michael Thake); West End (Shilpa Patel, Carolyn Roost, and Sandra Sowerby).

and Gemma Walsh); Oxford (Gayle Franklin and Cheryl Howath); Poole (Sarah Sharpe); Plymouth (Deborah Smith); Sheffield (Lauren Botham); Southampton (Caroline Williams and Claire Alexander); Stoke (Gareth Sowerbutts and Diane Furnival); Tooting (Michael Thake); West End (Shilpa Patel, Carolyn Roost, and Sandra Sowerby). NHSBT collaborators: Mary Joy Appleton, Eileen Bays, Geoff Bowyer, Steven Clarkson, Stuart Halson, Kate Holmes, Gareth Humphreys, Kristie Johnston, Lee Parvin-Cooper, Jason Towler. NHSBT INTERVAL Study Administration Team: Joanne Addy, Patricia Barrass, Louise Stennett. INTERVAL Helpdesk: Susan Burton, Hannah Dingwall, Rachel Henry (previous members Victoria Clarke, Maria Potton, Claire Thomson). Trial Data Management Team: Thomas Bolton, Michael Daynes, Stuart Halson, Sarah Spackman, Michael Walker (previous members Abudu Momodu). UK Biocentre: James Fenton, Adam King, Omer Muhammad, Nicholas Oates, Tim Peakman, Christine Ryan, Kristian Spreckley, Craig Stubbins, Joanna Williams (previous members James Brennan, Cedric Mochon, Samantha Taylor, Kimberly Warren). Trial statisticians: Stephen Kaptoge PhD, University of Cambridge; Simon G Thompson FMedSci, University of Cambridge. Co-investigators: Emanuele Di Angelantonio FRCP, University of Cambridge and NHSBT; Carmel Moore PhD, University of Cambridge; Jonathan Mant FRCPE, University of Cambridge; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; Simon G Thompson FMedSci, University of Cambridge.

Trial statisticians: Stephen Kaptoge PhD, University of Cambridge; Simon G Thompson FMedSci, University of Cambridge. Co-investigators: Emanuele Di Angelantonio FRCP, University of Cambridge and NHSBT; Carmel Moore PhD, University of Cambridge; Jonathan Mant FRCPE, University of Cambridge; Willem H Ouwehand FMedSci, University of Cambridge and NHSBT; Simon G Thompson FMedSci, University of Cambridge. Chief Investigators: John Danesh FMedSci, University of Cambridge; David J Roberts FRCPath, University of Oxford and NHSBT. *Joint first authors. †Joint last authors. Declaration of interests SK received grants from the UK Medical Research Council, British Heart Foundation, National Institute for Health Research, and NHS Blood and Transplant (NHSBT) during the conduct of the study. EDA received research funding from the UK Medical Research Council, British Heart Foundation, National Institute for Health Research, and NHSBT during the conduct of the study. SGT received grants from the UK Medical Research Council, British Heart Foundation, and National Institute for Health Research during the conduct of the study. JD has received research funding from the UK Medical Research Council, British Heart Foundation, National Institute for Health Research, NHSBT, European Research Council, Merck Sharpe & Dohme UK, Novartis, Pfizer, AstraZeneca, and Wellcome Trust during the conduct of the study. All other members of the writing committee declare no competing interests.

Introduction Although exposure to high-dose ionising radiation is rare outside of radiotherapy, repeated or protracted low-dose exposure has become increasingly common over the past 25 years.1 Occupational and environmental sources of radiation exposure are important; however, the largest contributor to this trend is medical radiation exposure. In 1982, the average yearly dose of ionising radiation from medical exposures was about 0·5 mGy per person in the USA; by 2006, it had increased to 3·0 mGy.2 A similar pattern exists in other high-income countries: use of diagnostic procedures involving radiation in the UK more than doubled over that period3 and more than tripled in Australia.4 Because ionising radiation is a carcinogen,5 its use in medical practice must be balanced against the risks associated with patient exposure.6 The primary basis for estimating cancer risks from ionising radiation exposures are epidemiological studies of Japanese survivors of the atomic bombings of Hiroshima and Nagasaki in August, 1945.7 Within a few years of the bombings there was evidence of an excess of leukaemia, predominantly myeloid subtypes, among the survivors.8, 9, 10, 11, 12 These findings helped to establish that ionising radiation causes leukaemia.13 However, this evidence mostly relates to acute high-dose exposure. The risks associated with protracted or repeated low-dose exposures are more relevant to the public and health practitioners.

d subtypes, among the survivors.8, 9, 10, 11, 12 These findings helped to establish that ionising radiation causes leukaemia.13 However, this evidence mostly relates to acute high-dose exposure. The risks associated with protracted or repeated low-dose exposures are more relevant to the public and health practitioners. The International Nuclear WORKers Study (INWORKS) was done to strengthen the scientific basis for protecting people from low-dose protracted or intermittent radiation exposure. It included workers from France,14 the UK,15 and the USA16 who have been monitored for external exposure to radiation with personal dosimeters and followed up for up to 60 years after exposure. Here, we report data for leukaemia, lymphoma, and multiple myeloma mortality among participants of INWORKS. Research in context Evidence before this study Ionising radiation causes leukaemia. The primary quantitative basis for radiation protection standards comes from studies of populations exposed to acute, high doses of ionising radiation. Although previous studies of nuclear workers addressed leukaemia radiogenicity, questions remain about the size of the risk from protracted radiation exposure in occupational settings. Added value of this study

Ionising radiation causes leukaemia. The primary quantitative basis for radiation protection standards comes from studies of populations exposed to acute, high doses of ionising radiation. Although previous studies of nuclear workers addressed leukaemia radiogenicity, questions remain about the size of the risk from protracted radiation exposure in occupational settings. Added value of this study We report a positive dose–response relationship between cumulative, external, protracted, low-dose exposure to ionising radiation, and subsequent death caused by leukeamia (excluding chronic lymphocytic leukaemia). The risk coefficient per unit dose was consistent with those derived from analyses of other populations exposed to higher radiation doses and dose rates. Implications of all the available evidence The present study provides strong evidence of a positive association between radiation exposure and leukaemia even for low-dose exposure. This finding shows the importance of adherence to the basic principles of radiation protection—to optimise protection to reduce exposures as much as reasonably achievable and—in the case of patient exposure—to justify that the exposure does more good than harm.

diation exposure and leukaemia even for low-dose exposure. This finding shows the importance of adherence to the basic principles of radiation protection—to optimise protection to reduce exposures as much as reasonably achievable and—in the case of patient exposure—to justify that the exposure does more good than harm. Methods Study design and participants The INWORKS cohort consists of nuclear workers from three of the major partners included in the previously published 15-country study of cancer among workers in the nuclear industry:17 France,14 the UK,15 and the USA.16 Less than 20% of deaths from leukaemia were contributed by the other 12 countries.18 These cohorts have been updated since the 15-country study. INWORKS includes fewer partners than the earlier 15-country study because of the limited resources and the consequent need for efficiency in project coordination.

d the USA.16 Less than 20% of deaths from leukaemia were contributed by the other 12 countries.18 These cohorts have been updated since the 15-country study. INWORKS includes fewer partners than the earlier 15-country study because of the limited resources and the consequent need for efficiency in project coordination. The study includes workers employed by the French Atomic Energy Commission, AREVA Nuclear Cycle, and Electricité de France, workers employed by the British Atomic Weapons Establishment, British Nuclear Fuels, the UK Atomic Energy Authority, British Energy Generation, the UK Ministry of Defence, and other organisations providing data to the National Registry for Radiation Workers, and workers employed by the US Department of Energy's Hanford Site, Savannah River Site, Oak Ridge National Laboratory, Idaho National Laboratory, and the Portsmouth Naval Shipyard. Workers who were employed in the nuclear industry for less than 1 year were excluded. In France, workers were given the opportunity to refuse participation, which is required by the French Data Protection Authority; however, none did. In the USA, worker information was taken from existing records, with no direct contact with any participants; because there is minimal risk to participants, the National Institute for Occupational Safety and Health institutional review board waived requirements for informed consent. UK workers can refuse to participate in the National Registry for Radiation Workers and associated studies; less than 1% did.

th any participants; because there is minimal risk to participants, the National Institute for Occupational Safety and Health institutional review board waived requirements for informed consent. UK workers can refuse to participate in the National Registry for Radiation Workers and associated studies; less than 1% did. Procedures Participants were followed up for a total of 8·22 million person-years to ascertain vital status up to 2004 in France, 2001 in the UK, and 2005 in the USA. Underlying cause of death was abstracted from death certificates and generally coded according to the revision of the International Classification of Diseases (ICD) in effect at the time of death. We assessed leukaemia other than chronic lymphocytic leukaemia (CLL; ICD9 codes 204–208 excluding 204.1 and 204.9), acute myeloid leukaemia (ICD9 codes 205.0, 206.0, 207.0, and 207.2), chronic myeloid leukaemia (ICD9 code 205.1), acute lymphoblastic leukaemia (ICD9 code 204.0), and CLL (ICD9 code 204.1). We assessed lymphoma deaths separately for non-Hodgkin lymphoma (ICD9 codes 200, 202, 273.3), Hodgkin's lymphoma (ICD9 code 201), and multiple myeloma (ICD9 code 203). The appendix (p 2) shows an exhaustive list of ICD codes.

(ICD9 code 205.1), acute lymphoblastic leukaemia (ICD9 code 204.0), and CLL (ICD9 code 204.1). We assessed lymphoma deaths separately for non-Hodgkin lymphoma (ICD9 codes 200, 202, 273.3), Hodgkin's lymphoma (ICD9 code 201), and multiple myeloma (ICD9 code 203). The appendix (p 2) shows an exhaustive list of ICD codes. Data for monitoring exposure to ionising radiation were available from dose registry, government, and company records, providing individual yearly estimates of whole-body exposure to external penetrating radiation (primarily γ rays). Red bone marrow absorbed doses expressed in Gy were derived by dividing recorded external penetrating radiation dose estimates by the appropriate organ dose conversion factor.19, 20 In this report, dose indicates absorbed dose to red bone marrow expressed in Gy. Because most external exposures were to high-energy photons, with a radiation weighting factor of 1·0, absorbed dose in Gy could be expressed in terms of equivalent dose in Sieverts.

s by the appropriate organ dose conversion factor.19, 20 In this report, dose indicates absorbed dose to red bone marrow expressed in Gy. Because most external exposures were to high-energy photons, with a radiation weighting factor of 1·0, absorbed dose in Gy could be expressed in terms of equivalent dose in Sieverts. Statistical analysis Participants entered the study either 1 year after the date of first employment or on the date of first dosimetric monitoring, whichever was later. In France, the national death registry recorded information on individual causes of death only since 1968; therefore, French workers entered follow-up on Jan 1, 1968, or later. Participants remained in the study until the earliest of date of death, date lost to follow-up, or end of follow-up. We estimated relative risk (RR) by a model of the form RR=1 + βd, generally used in studies of radiation effects,21 where d is the dose and β is an estimate of the excess relative risk (ERR; RR – 1) per unit dose; we derived likelihood-based CIs. All models were stratified by country, sex, calendar period (<1946, 1946–50…1996–2000, ≥2001), and age (<35, 35–39…70–74, ≥75); these potential confounders were selected a priori from a set of measured covariates. We also fitted linear-quadratic and pure-quadratic functions of dose and selected a model with Akaike information criterion.22

tified by country, sex, calendar period (<1946, 1946–50…1996–2000, ≥2001), and age (<35, 35–39…70–74, ≥75); these potential confounders were selected a priori from a set of measured covariates. We also fitted linear-quadratic and pure-quadratic functions of dose and selected a model with Akaike information criterion.22 To allow for an induction and latency period between exposure to radiation and death, cumulative doses were lagged by 2 years for analyses of leukaemia mortality and by 10 years for analyses of lymphoma and multiple myeloma. These lag assumptions were chosen a priori. In sensitivity analyses we assessed a 10-year lag for analyses of leukaemia mortality and a 2-year lag for analyses of lymphoma and multiple myeloma, fitted models to restricted ranges of dose, and excluded workers with substantial doses from neutrons (ie, workers with recorded cumulative neutron doses exceeding 10% of the total equivalent dose for external radiation). To provide empirical support for the absence of confounding by socioeconomic status, we report supplementary analyses adjusted for socioeconomic status (based on job title: managers and engineers, administrative staff, skilled workers, unskilled workers, uncertain); and, to address concern about potential confounding by internal contamination, we report analyses adjusted for known or suspected internal radiation exposure. We did the analyses excluding one country at a time to assess the effect of a single country on overall results. Because the objective of most contemporary radiation epidemiological studies is to investigate the potential for an increased cancer risk in relation to radiation exposure, one-sided p values and corresponding 90% CIs are usually presented; we follow that convention here by reporting 90% CIs. All models were fitted with EPICURE software.23

of most contemporary radiation epidemiological studies is to investigate the potential for an increased cancer risk in relation to radiation exposure, one-sided p values and corresponding 90% CIs are usually presented; we follow that convention here by reporting 90% CIs. All models were fitted with EPICURE software.23 Role of the funding source The funders had no role in study design, data analysis, data interpretation, or writing of the report. AREVA and Électricité de France provided historical occupational data and individual monitoring data for part of the French cohort. KL, DBR, and MM had full access to all the data in the study. KL and DBR had final responsibility for the decision to submit for publication. Results We assembled a cohort of 308 297 radiation-monitored workers. Table 1 shows the characteristics of the study population. Mean follow-up was 27 years (SD 12) and nearly 22% of the workers were deceased at the end of follow-up. Mean cumulative dose was 16 mGy. The median was 2·1 mGy (IQR 0·3–11·7), with a tenth percentile of 0·0 mGy and a 90th percentile of 40·8 mGy (appendix p 1). The mean yearly dose was 1·1 mGy (SD 2·6).

udy population. Mean follow-up was 27 years (SD 12) and nearly 22% of the workers were deceased at the end of follow-up. Mean cumulative dose was 16 mGy. The median was 2·1 mGy (IQR 0·3–11·7), with a tenth percentile of 0·0 mGy and a 90th percentile of 40·8 mGy (appendix p 1). The mean yearly dose was 1·1 mGy (SD 2·6). We recorded 531 deaths caused by leukaemia excluding CLL, 814 caused by lymphoma, and 293 caused by multiple myeloma. 281 (53%) of 531 deaths caused by leukaemia excluding CLL occurred in people who had accrued less than 5 mGy (appendix p 3). The RR of death caused by leukaemia excluding CLL by categories of cumulative dose showed a substantial risk for cumulative dose above 200 mGy (appendix p 3). The estimated ERR of mortality caused by leukaemia excluding CLL was 2·96 per Gy (90% CI 1·17–5·21; table 2). The trend in the ERR of leukaemia excluding CLL with dose was well described by a simple linear function of cumulative dose; inclusion of a higher order polynomial function (ie, a linear-quadratic or pure-quadratic function of dose) did not substantially improve the model fit (the Akaike information criterion was lowest for the pure-quadratic model but only differed by 0·3 from that of the linear model; data not shown). The ERR of leukaemia excluding CLL was not attenuated when restricted to doses of less than 300 mGy or less than 100 mGy (figure); however, 90% CIs were much wider when based on data for the restricted dose range.

west for the pure-quadratic model but only differed by 0·3 from that of the linear model; data not shown). The ERR of leukaemia excluding CLL was not attenuated when restricted to doses of less than 300 mGy or less than 100 mGy (figure); however, 90% CIs were much wider when based on data for the restricted dose range. We assessed the associations between cumulative dose and subtypes of leukaemia. We detected positive associations for chronic myeloid leukaemia, acute myeloid leukaemia, and acute lymphoblastic leukaemia; the association was largest for chronic myeloid leukaemia (table 2). Associations also were positive but highly imprecise for Hodgkin's lymphoma, non-Hodgkin lymphoma, and multiple myeloma with CIs that spanned zero (table 2). The association between radiation dose and CLL mortality was negative (table 2).

c leukaemia; the association was largest for chronic myeloid leukaemia (table 2). Associations also were positive but highly imprecise for Hodgkin's lymphoma, non-Hodgkin lymphoma, and multiple myeloma with CIs that spanned zero (table 2). The association between radiation dose and CLL mortality was negative (table 2). Alternative lag assumptions resulted in little change in the ERR per Gy (appendix p 4). When adjusting the ERR model for socioeconomic status, the ERR per Gy was practically unchanged for leukaemia excluding CLL and for chronic myeloid leukaemia (appendix p 5). Similarly, adjustment for internal radiation contamination had little effect (appendix p 5). We assessed the effect of excluding people who had recorded neutron exposures; we showed a positive association for leukaemia excluding CLL (ERR per Gy 4·19, 90% CI 1·42–7·80, 453 deaths) and chronic myeloid leukaemia (ERR per Gy 9·55, 90% CI 2·39–21·7, 79 deaths). To assess whether any single country substantially affected the results, we assessed radiation-mortality associations excluding one country at a time (appendix p 6). The estimated ERR per Gy for leukaemia excluding CLL was 2·95 (90% CI 1·13–5·24) when excluding France, 2·32 (0·03–5·33) when excluding the UK, and 3·68 (1·09–7·29) when excluding the USA (appendix p 6). For multiple myeloma and Hodgkin's lymphoma, the associations could not be estimated when excluding the USA, but the multiple myeloma was positive when excluding the UK (ERR per Gy 3·32 [90% CI 0·27–7·64]).

uding France, 2·32 (0·03–5·33) when excluding the UK, and 3·68 (1·09–7·29) when excluding the USA (appendix p 6). For multiple myeloma and Hodgkin's lymphoma, the associations could not be estimated when excluding the USA, but the multiple myeloma was positive when excluding the UK (ERR per Gy 3·32 [90% CI 0·27–7·64]). Discussion We showed a positive association between cumulative dose of ionising radiation and death caused by leukaemia (excluding CLL) among adults who were typically exposed to low doses. The association was greatest for chronic myeloid leukaemia, with positive but imprecise dose–response for deaths caused by acute myeloid leukaemia, acute lymphoblastic leukaemia, Hodgkin's lymphoma, non-Hodgkin lymphoma, and multiple myeloma. The estimated association between cumulative radiation dose with a 2-year exposure lag assumption and death caused by leukaemia excluding CLL was similar in size and precision to the linear dose–response estimate for male atomic bomb survivors exposed between the ages of 20 and 60 years (ERR at 1 Sv 2·63, 90% CI 1·50–4·27).14 Although based on a substantially lower dose distribution than in analyses of atomic bomb survivors, typically with very low doses accrued over a long period, the similar size of the associations supports contemporary estimates of risk of leukaemia after adult exposure to radiation. This is notable because our estimates were not extrapolated from data for acute exposures.

ion than in analyses of atomic bomb survivors, typically with very low doses accrued over a long period, the similar size of the associations supports contemporary estimates of risk of leukaemia after adult exposure to radiation. This is notable because our estimates were not extrapolated from data for acute exposures. In previous analyses of cancer among workers in 15 countries,18 the association between mortality for leukaemia excluding CLL and cumulative radiation dose with a 2-year exposure lag assumption (ERR per Sv 1·93, 90% CI <0–7·14) was smaller and much less precise than the estimate we obtained in our pooled analysis of three countries. The gain in precision is a result of the larger number of deaths from leukaemia excluding CLL in INWORKS (n=531) compared to the earlier study (n=196), because of longer follow-up (mean follow-up in INWORKS was 27 years vs 13 years in the 15-country study17) and the enlargement of the French, UK, and US cohorts compared with previous analyses.14, 15, 16 Moreover, the 15-country study excluded people with potential exposures from neutron and internal contamination. In our study, we included 127 deaths caused by leukaemia excluding CLL for workers with potential exposure to neutron and internal contamination. Similarly, the risk estimate for non-Hodgkin lymphoma in the INWORKS study was more precise than the estimate reported in the 15-country study,24 again because the present study included more deaths (248 in the 15-country study, 710 in the present study). The CIs do not overlap for estimated associations between radiation dose and death caused by acute and chronic myeloid leukaemia; a formal test of heterogeneity in associations by leukaemia subtype would require a joint modelling approach and was not used here.

s (248 in the 15-country study, 710 in the present study). The CIs do not overlap for estimated associations between radiation dose and death caused by acute and chronic myeloid leukaemia; a formal test of heterogeneity in associations by leukaemia subtype would require a joint modelling approach and was not used here. We did not find any effect of a single country on the estimated association for leukaemia excluding CLL. For multiple myeloma, the association was significantly positive when only the UK data were excluded, suggesting a possible heterogeneity in the risk pattern between the three cohorts. Schubauer-Berigan and colleagues16 reported a significant increased risk of multiple myeloma mortality associated with dose in their analysis of the USA cohort (ERR per 10 mSv 3·9, 90% CI 0·6–9·6), whereas no significant dose-related excess was detected in the third analysis of the UK National Registry for Radiation Workers (although a significant excess risk was recorded in an analysis of incidence).15 Multiple myeloma has a potentially long period of development of up to 20 years. The older age at the end of follow-up in the USA cohort might explain the heterogeneity.

the third analysis of the UK National Registry for Radiation Workers (although a significant excess risk was recorded in an analysis of incidence).15 Multiple myeloma has a potentially long period of development of up to 20 years. The older age at the end of follow-up in the USA cohort might explain the heterogeneity. We tried to reduce uncertainties in dose estimates that could bias dose–response analyses.20 Nevertheless, occupational radiation dose estimates are prone to measurement error; consequently, exposure misclassification is an unavoidable study limitation. Outcome misclassification is also a potential concern in studies that rely on death certificates for classification of leukaemia and lymphoma by subtype. This concern is well known for CLL, for which incidence studies seem more appropriate.25, 26, 27, 28 Poor sensitivity and imperfect specificity of death certificates might reduce statistical precision and induce bias in analyses of subtypes. However, death certificate information remains a valuable resource for this type of cohort investigation.

for CLL, for which incidence studies seem more appropriate.25, 26, 27, 28 Poor sensitivity and imperfect specificity of death certificates might reduce statistical precision and induce bias in analyses of subtypes. However, death certificate information remains a valuable resource for this type of cohort investigation. There are few potential confounders of the associations under study. For example, smoking causes myeloid leukaemia;29, 30 however, the size of this association is relatively small31 and therefore would require large differences in smoking across levels of cumulative dose to cause substantial confounding of the radiation–leukaemia association. Moreover, adjusting risk analyses by socioeconomic status would reduce substantial confounding by smoking.32 Adjustment for socioeconomic status resulted in little change in the risk estimate for leukaemia excluding CLL. Exposure of nuclear workers to other causes of leukaemia such as benzene29, 30 cannot be excluded as a potential source of bias, even though benzene was not widely used in the nuclear industry. In a previous analysis of US nuclear workers, Schubauer-Berigan and coworkers33 reported weak evidence of confounding by benzene exposure when analysing leukaemia risk associated with external radiation exposure. Benzene exposure could not be assessed for the INWORKS study. Internal exposures to radionuclides—notably uranium and plutonium—occurred at the study sites, and we did not evaluate doses from these intakes. However, our sensitivity analyses showed that internal contamination might have little effect on the relation between external radiation exposure and leukaemia risk. These results are consistent with the conclusions of Shilnikova and colleagues,34 who reported no indication of any effect of internal contamination on leukaemia mortality among nuclear workers, whereas the risk of leukaemia was positively associated with external γ-ray exposure.

radiation exposure and leukaemia risk. These results are consistent with the conclusions of Shilnikova and colleagues,34 who reported no indication of any effect of internal contamination on leukaemia mortality among nuclear workers, whereas the risk of leukaemia was positively associated with external γ-ray exposure. Medical workers are also exposed to low doses of external γ-rays or x-rays. No study has provided estimates of leukaemia risk for medical workers because accurate historical dosimetry data are not available for these populations.35 Liu and colleagues36 estimated mortality in a cohort of 90 268 USA radiological technologists. They reported that the leukaemia risk was doubled for technologists who had worked for more than 30 years compared with those who had worked for less than 10 years, but the cohort did not provide any information about doses received by the workers. In summary, this study provides strong evidence of an association between protracted low dose radiation exposure and leukaemia mortality. At present, radiation protection systems are based on a model derived from acute exposures, and assumes that the risk of leukaemia per unit dose progressively diminishes at lower doses and dose rates.37 Our results provide direct estimates of risk per unit of protracted dose in ranges typical of environmental, diagnostic medical, and occupational exposure. Supplementary Material Supplementary appendix Supplementary audio Klervi Leuraud discusses a study reporting on the risk of developing leukaemia and lymphoma in workers monitored for radiation exposure.

In summary, this study provides strong evidence of an association between protracted low dose radiation exposure and leukaemia mortality. At present, radiation protection systems are based on a model derived from acute exposures, and assumes that the risk of leukaemia per unit dose progressively diminishes at lower doses and dose rates.37 Our results provide direct estimates of risk per unit of protracted dose in ranges typical of environmental, diagnostic medical, and occupational exposure. Supplementary Material Supplementary appendix Supplementary audio Klervi Leuraud discusses a study reporting on the risk of developing leukaemia and lymphoma in workers monitored for radiation exposure. Acknowledgments This work was partly funded by the Centers for Disease Control and Prevention (5R03 0H010056-02) and the Ministry of Health, Labour and Welfare of Japan (GA No 2012-02-21-01). The construction of the French cohort was realised by the Institut de Radioprotection et de Sûreté Nucléaire, with partial funding from AREVA and Electricité de France. The Institut de Radioprotection et de Sûreté Nucléaire thanks all people from the French Atomic Energy Commission, AREVA, and Electricité de France who cooperated in the elaboration of the French cohort. For the US contribution, funding was provided by the National Institute for Occupational Safety and Health, by the US Department of Energy through an agreement with the US Department of Health and Human Services, and through a grant received by the University of North Carolina from the National Institute for Occupational Safety and Health (R03 OH-010056). The construction of the UK cohort was undertaken by Public Health England who operate the UK's National Registry for Radiation Workers. Public Health England thank all of the organisations and individuals participating in the National Registry for Radiation Workers for their cooperation, and the National Registry for Radiation Workers steering group for their continued support. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the National Institute for Occupational Safety and Health.

Registry for Radiation Workers for their cooperation, and the National Registry for Radiation Workers steering group for their continued support. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the National Institute for Occupational Safety and Health. Contributors DBR and AK had the idea for the study. DBR, AK, EC, RDD, MG, JAO'H, GBH, RH, KL, DL, MKS-B, and IT-C designed the study. KL and DL worked on provision of the French data, MKS-B and RDD worked on provision of the US data, MG, JAO'H, and RH worked on provision of the UK data. MM managed, processed, and analysed the data. IT-C analysed and assessed dosimetry data. KL did the statistical analysis. KL and DBR wrote the initial draft of the report, which was revised and approved by all authors. Declaration of interests We declare no competing interests. Figure Relative risk of leukaemia excluding chronic lymphocytic leukaemia associated with 2-year lagged cumulative red bone marrow dose The lines are the fitted linear dose–response model and the shading represents the 90% CIs. Table 1 Characteristics of individuals included in INWORKS

Declaration of interests We declare no competing interests. Figure Relative risk of leukaemia excluding chronic lymphocytic leukaemia associated with 2-year lagged cumulative red bone marrow dose The lines are the fitted linear dose–response model and the shading represents the 90% CIs. Table 1 Characteristics of individuals included in INWORKS France USA UK Overall Study period 1968–2004 1944–2005 1946–2001 1944–2005 Number of participants 59 003 101 428 147 866 308 297 Person-years (millions) 1·47 3·34 3·41 8·22 Duration of follow-up (years) Mean (SD) 25 (9) 33 (13) 23 (12) 27 (12) Median (IQR) 23 (18–36) 31 (23–44) 22 (14–32) 26 (18–36) Age at last observation (years) Mean (SD) 56 (13) 65 (13) 54 (15) 58 (15) Median (IQR) 54 (46–66) 66 (55–76) 54 (42–66) 58 (47–70) Sex Male 51 567 (87%) 81 883 (81%) 134 812 (91%) 268 262 (87%) Female 7436 (13%) 19 545 (19%) 13 054 (9%) 40 035 (13%) Vital status on Dec 31, 2005 Alive 52 565 (89%) 65 573 (65%) 118 775 (80%) 236 913 (77%) Died 6310 (11%) 35 015 (35%) 25 307 (17%) 66 632 (22%) Number of deaths from malignant neoplasm of lymphoid and haemopoietic tissues (% of total deaths) 196 (3%) 1031 (3%) 564 (2%) 1791 (3%) Emigrated or lost to follow-up 128 (<1%) 840 (1%) 3784 (3%) 4752 (2%) Cumulative red bone marrow dose (mGy) Mean (range) 11·6 (0·0–415·8) 15·2 (0·0–820·2) 18·2 (0·0–1217·5) 15·9 (0·0–1217·5) Median (IQR) 1·3 (0·0–10·7) 1·9 (0·2–10·6) 2·6 (0·4–12·9) 2·1 (0·3–11·7) Data are n (%) unless stated otherwise. Table 2 ERR per Gy of cumulative red bone marrow dose for causes of death

France USA UK Overall Study period 1968–2004 1944–2005 1946–2001 1944–2005 Number of participants 59 003 101 428 147 866 308 297 Person-years (millions) 1·47 3·34 3·41 8·22 Duration of follow-up (years) Mean (SD) 25 (9) 33 (13) 23 (12) 27 (12) Median (IQR) 23 (18–36) 31 (23–44) 22 (14–32) 26 (18–36) Age at last observation (years) Mean (SD) 56 (13) 65 (13) 54 (15) 58 (15) Median (IQR) 54 (46–66) 66 (55–76) 54 (42–66) 58 (47–70) Sex Male 51 567 (87%) 81 883 (81%) 134 812 (91%) 268 262 (87%) Female 7436 (13%) 19 545 (19%) 13 054 (9%) 40 035 (13%) Vital status on Dec 31, 2005 Alive 52 565 (89%) 65 573 (65%) 118 775 (80%) 236 913 (77%) Died 6310 (11%) 35 015 (35%) 25 307 (17%) 66 632 (22%) Number of deaths from malignant neoplasm of lymphoid and haemopoietic tissues (% of total deaths) 196 (3%) 1031 (3%) 564 (2%) 1791 (3%) Emigrated or lost to follow-up 128 (<1%) 840 (1%) 3784 (3%) 4752 (2%) Cumulative red bone marrow dose (mGy) Mean (range) 11·6 (0·0–415·8) 15·2 (0·0–820·2) 18·2 (0·0–1217·5) 15·9 (0·0–1217·5) Median (IQR) 1·3 (0·0–10·7) 1·9 (0·2–10·6) 2·6 (0·4–12·9) 2·1 (0·3–11·7) Data are n (%) unless stated otherwise. Table 2 ERR per Gy of cumulative red bone marrow dose for causes of death Deaths ERR per Gy 90% CI Leukaemia excluding CLL* 531 2·96 1·17 to 5·21 Chronic myeloid leukaemia* 100 10·45 4·48 to 19·65 Acute myeloid leukaemia* 254 1·29 −0·82 to 4·28 Acute lymphoblastic leukaemia* 30 5·80 NE to 31·57 CLL* 138 −1·06 NE to 1·81 Multiple myeloma† 293 0·84 −0·96 to 3·33 Non-Hodgkin lymphoma† 710 0·47 −0·76 to 2·03 Hodgkin's lymphoma† 104 2·94 NE to 11·49 ERR estimated with a linear model stratified by country, calendar period, sex, and age. NE lower CI bound could not be estimated because it was on the boundary of the parameter space (−1/maximum dose). 14 deaths were assigned ICD9 code 204.9 (lymphoid leukaemia, unspecified) and one death was assigned ICD9 code 202.9 (other and unspecified malignant neoplasms of lymphoid, haemopoietic, and related tissue) were excluded from the cause-specific analyses.

it was on the boundary of the parameter space (−1/maximum dose). 14 deaths were assigned ICD9 code 204.9 (lymphoid leukaemia, unspecified) and one death was assigned ICD9 code 202.9 (other and unspecified malignant neoplasms of lymphoid, haemopoietic, and related tissue) were excluded from the cause-specific analyses. * 2-year lagged cumulative dose. † 10-year lagged cumulative dose. ERR=excess relative risk. CLL=chronic lymphocytic leukaemia. NE=not estimable.

Introduction The two classical BCR-ABL–negative myeloproliferative neoplasms (MPN) essential thrombocythemia and polycythemia vera are clonal hematopoietic neoplasms characterized by an overproduction of mature blood elements, tendencies toward thrombosis and hemorrhage, extramedullary hematopoiesis (mild splenomegaly), and transformation to myelofibrosis (MF)/acute myeloid leukemia (AML) 1–3. The therapeutic approach primarily focuses on controlling blood counts and reducing the risk of thrombosis. For patients judged to be at high-risk of thrombosis, cytoreductive therapy is instituted (typically hydroxyurea). As an alternative to hydroxyurea, recombinant interferon-alpha is frequently suggested given its biologic, anti-proliferative, immunomodulating, and anti-clonal effects in these patients. However, its widespread use has been limited by high rates of discontinuation due to side effects 4–9. Pegylated forms of interferon have a better pharmacologic profile than short-acting interferons, resulting in a more convenient less-frequent schedule of injections, less immunogenicity and possibly less toxicity 10. In particular, several clinical studies of pegylated interferon alfa-2a (PEG-IFN-α-2a), including our own 11,12, have reported promising results in patients with essential thrombocythemia and polycythemia vera. PEG-IFN-α-2a induced complete hematologic response (CHR) in a great majority of patients and complete molecular remission (CMR) in a fraction of patients, with lower toxicity rates than expected. However, all studies had a relatively short follow-up and the durability of responses and long-term safety have not been reported.

N-α-2a induced complete hematologic response (CHR) in a great majority of patients and complete molecular remission (CMR) in a fraction of patients, with lower toxicity rates than expected. However, all studies had a relatively short follow-up and the durability of responses and long-term safety have not been reported. Herein, we present long-term efficacy and safety data from a prospective phase 2 study of PEG-IFN-α-2a in 83 patients with advanced essential thrombocythemia and polycythemia vera after a median follow-up of 7 years, nearly twice as long as in previously reported studies.

N-α-2a induced complete hematologic response (CHR) in a great majority of patients and complete molecular remission (CMR) in a fraction of patients, with lower toxicity rates than expected. However, all studies had a relatively short follow-up and the durability of responses and long-term safety have not been reported. Herein, we present long-term efficacy and safety data from a prospective phase 2 study of PEG-IFN-α-2a in 83 patients with advanced essential thrombocythemia and polycythemia vera after a median follow-up of 7 years, nearly twice as long as in previously reported studies. Methods Patient Selection This is an ad hoc analysis of data from a prospective, open-label, single-center, phase 2 trial of PEG-IFN-α-2a in patients with essential thrombocythemia and polycythemia vera.11,12. Patients older than 18 years with either newly diagnosed or previously treated essential thrombocythemia or polycythemia vera according to the 2005 Polycythemia Vera Study Group criteria were eligble to enroll. Additional inclusion criteria included Eastern Cooperative Group Oncology status ≤ 2; adequate liver (total bilirubin ≤ 2.0 mg/dl) and renal function (serum creatinine ≤ 2.0 mg/dl); and normal cardiac function. Patients must have been off chemotherapy for at least one week prior to enrollment, but could be receiving hydroxyurea or anagrelide for up to 1 month after study entry. A washout period of one month was required for patients treated with continuous or chronic high doses of steroid. Exclusion criteria included pregnant or lactating women; history of another malignancy unless disease free for > 3 years); ischemic retinopathy; severe cardiac disease; history of medically significant psychiatric disease if not controlled, especially endogenous depression; a seizure disorder requiring anticonvulsant therapy; known infection with hepatitis B or C or HIV or other active systemic infection; renal disease requiring hemodialysis; or known autoimmune disease (except rheumatoid arthritis). All patients provided written informed consent. This study was approved by the Institutional Review Board at The University of Texas MD Anderson Cancer Center and conducted in accordance with the Declaration of Helsinki.

n; renal disease requiring hemodialysis; or known autoimmune disease (except rheumatoid arthritis). All patients provided written informed consent. This study was approved by the Institutional Review Board at The University of Texas MD Anderson Cancer Center and conducted in accordance with the Declaration of Helsinki. Study Procedures PEG-IFN-α-2a was administered subcutaneously once weekly. The initial starting dose was 450 μg/week but was decreased in a stepwise manner due to toxicity to a final starting dose of 90 mcg/week: 3 patients were started at a dose of 450 μg/week, 3 at 360 μg/week, 19 at 270 μg/week, 26 at 180 μg/week, and 32 at 90 μg/week. Treatment was continued as long as the patients derived clinical benefit. During the study, the dose was modified based on toxicity or lack of efficacy. Any grade 3 or 4 event required therapy interruption. If the event resolved to grade 0 or 1 therapy could be resumed at a lower dose level. Persistent (≥ 2 weeks) significant grade 2 adverse events required dose reductions or discontinuation. Criteria for discontinuation included clearly documented disease progression (increasing transfusion requirement, splenomegaly, platelet or white blood cell counts, frequency of phlebotomy, or thromboembolic events) or no response within 6 months from the start of therapy despite dose escalation. The study is ongoing but not enrolling new patients.

ncluded clearly documented disease progression (increasing transfusion requirement, splenomegaly, platelet or white blood cell counts, frequency of phlebotomy, or thromboembolic events) or no response within 6 months from the start of therapy despite dose escalation. The study is ongoing but not enrolling new patients. Outcomes The primary endpoint was hematologic response rate, as defined by European LeukemiaNet criteria 13. Complete hematologic response (CHR) was defined as normalization of blood counts (essential thrombocythemia: platelets ≤ 440 x 109/L; PV: hemoglobin < 15.0 g/L without phlebotomy) with complete resolution of palpable splenomegaly/symptoms in the absence of a thrombotic event. A partial hematologic response (PR) required at least a 50% reduction in the platelet count for essential thrombocythemia or a 50% reduction in the rate of phlebotomies or 50% reduction in spleen size by palpation for polycythemia vera. Secondary endpoints were to evaluate the toxicities in these patients as well as the bone marrow morphologic and molecular disease characteristics before and during therapy. Physical exam and blood counts were assessed every 3 months. Bone marrow aspiration and biopsy with quantitation of JAK2V617F and cytogenetics were performed at the start of therapy and every 3 to 6 months thereafter. 11,12 All patients who tested positive for the JAK2V617F mutation were evaluable for a MR if they had 2 or more adequate bone marrow samples while on therapy taken 3 months apart within the first year. Complete molecular remission (CMR; based solely on the assessment of a JAK2V617F allele burden) required undetectable JAK2V617F, while partial and minor molecular remissions (PMR, mMR) required reductions in baseline allele burden of >50% and 20–49%, respectively. All adverse events were graded according to the National Cancer Institute Common Toxicity Criteria for Adverse Events (NCI-CTAE v2.0).

F allele burden) required undetectable JAK2V617F, while partial and minor molecular remissions (PMR, mMR) required reductions in baseline allele burden of >50% and 20–49%, respectively. All adverse events were graded according to the National Cancer Institute Common Toxicity Criteria for Adverse Events (NCI-CTAE v2.0). Statistical Analysis The analysis was based on an intention to treat population. A hematologic response rate of ≥ 35% was considered an indication of efficacy and justification for larger studies. An adverse event rate up to 20% was allowed. Only patients with a detectable JAK2V61F mutation at the start of therapy were evaluable for a molecular response. Responses and clinical data were analyzed using descriptive statistics. The Kaplan-Meier method with log-rank test were used to define and compare the time to leukemia transformation from the date of diagnosis, with p-value < 0.05 as statistically significant. Fisher’s exact test was used to compare responses in different groups for categorical variable. The Mann-Whitney U or Kruskal-Wallis tests were used to compare continuous variables, as indicated. GraphPad Prism and SPSS v.23 were used for all analyses. This study is registered with http://clinicaltrials.gov, number NCT00452023.

Responses and clinical data were analyzed using descriptive statistics. The Kaplan-Meier method with log-rank test were used to define and compare the time to leukemia transformation from the date of diagnosis, with p-value < 0.05 as statistically significant. Fisher’s exact test was used to compare responses in different groups for categorical variable. The Mann-Whitney U or Kruskal-Wallis tests were used to compare continuous variables, as indicated. GraphPad Prism and SPSS v.23 were used for all analyses. This study is registered with http://clinicaltrials.gov, number NCT00452023. Role of the Funding Source This research is supported in part by the MD Anderson Cancer Center Support Grant P30 CA016672 from the National Cancer Institute. Hoffman-LaRoche provided PEG-IFN-α-2a for five years but had no role in designing the study, data collection, analysis or interpretation of the data, or writing the final report. The corresponding author had full access to all of the data and the final responsibility to submit for publication.

ational Cancer Institute. Hoffman-LaRoche provided PEG-IFN-α-2a for five years but had no role in designing the study, data collection, analysis or interpretation of the data, or writing the final report. The corresponding author had full access to all of the data and the final responsibility to submit for publication. Results Patient characteristics/current status Forty-three patients with polycythemia vera and 40 with essential thrombocythemia were enrolled in a phase 2 trial of PEG-IFN-α-2a between May 31, 2005 and October 13, 2009, and 32 (39%) patients are still on study (polycythemia vera n=14, essential thrombocythemia n=18). Table 1 shows baseline demographic and clinical characteristics, which were evenly distributed between the polycythemia vera and essential thrombocythemia groups. The median follow-up time was 83 months (IQR, 69–94 months). Twenty-six (31%) patients were older than 60 years. Sixty-three percent of patients had received some form of therapy (in addition to aspirin) prior to enrollment, including standard IFN-α (n=14) and PEG-IFN-α-2a (n=1). Eleven essential thrombocythemia patients tested positive for CALR (n=8, 10%) and MPL (n=3, 3%), and 9 (11%) were triple negative.

older than 60 years. Sixty-three percent of patients had received some form of therapy (in addition to aspirin) prior to enrollment, including standard IFN-α (n=14) and PEG-IFN-α-2a (n=1). Eleven essential thrombocythemia patients tested positive for CALR (n=8, 10%) and MPL (n=3, 3%), and 9 (11%) were triple negative. Yearly discontinuation rates varied from 5% to 19% (Figure 1), with a median discontinuation rate of 5 patients/year (range, 3–16 patients/year). The median PEG-IFN-α-2a exposure time was 87 months (IQR, 17–85 months). The median follow-up after PEG-IFN-α-2a discontinuation for patients with available follow-up information (n=44) was 31 months (IQR, 18–65), and 12 patients were followed for more than 60 months after discontinuation. Hematologic response All 83 patients were evaluable for a hematologic response, and the overall response rate was 80% (n=66) (Table 2), as previously reported12. The median duration of HR was 66 months (IQR, 35–83), and 26 (39%) were in HR at the time of last follow-up (all but 1 was a CHR). Neither achievement of a HR nor time to response was associated with age, gender, baseline clinical characteristics, splenomegaly, molecular status or JAK2V617F allele burden (Appendix p.1).

he median duration of HR was 66 months (IQR, 35–83), and 26 (39%) were in HR at the time of last follow-up (all but 1 was a CHR). Neither achievement of a HR nor time to response was associated with age, gender, baseline clinical characteristics, splenomegaly, molecular status or JAK2V617F allele burden (Appendix p.1). Overall, 40 patients lost their response. Nineteen after dose reductions or drug holds due to intolerance or toxicity; one when he developed concurrent diffuse large B-cell lymphoma; and 20 due to progressive disease, despite being treated with the highest tolerable dose of PEG-IFN-α-2a. The median response duration among patients who lost their response was 46 months (IQR, 17–68). Among patients who were treated for at least 46 months, the median dose of PEG-IFN-α-2a was similar regardless of whether or not they lost their response (135 mcg/week vs 90 mcg/week, respectively, p=0.44). Remarkably, 7 patients (28%, 4 essential thrombocythemia, 3 polycythemia vera) have sustained their HR after discontinuation of PEG-IFN-α-2a (median time on therapy, 77 months [IWR, 56–98 months]; median response duration off study, 6 months [IQR, 4–34 months]).

onse (135 mcg/week vs 90 mcg/week, respectively, p=0.44). Remarkably, 7 patients (28%, 4 essential thrombocythemia, 3 polycythemia vera) have sustained their HR after discontinuation of PEG-IFN-α-2a (median time on therapy, 77 months [IWR, 56–98 months]; median response duration off study, 6 months [IQR, 4–34 months]). Molecular response Of the 63 (76%) JAK2V617F-positive patients, 55 (87%) were evaluable for a molecular response (MR). Eight patients who discontinued therapy in the first year and did not have 2 or more representative samples were not evaluable for a MR. The JAK2V617F allele burden was reduced in 63% (n=35) of these patients (Table 2), and 10 of them had a CMR as their best response (Appendix p.1). Among all JAK2-positive patients, the median JAK2V617F allele burden was 43% at baseline and 12% at the time of best response (Figure 2A). Among those with CMR, the JAK2V617F allele burden decreased by 35% and remained at that level (Appendix p.2). Patients with PMR or mMR had maximum reductions in allele burden of 23% and 16%, which were not sustained. This reduction was statistically significant only among polycythemia vera patients (P<0.001) and those with a CMR (P<0.001), as previously reported 12 (Figure 2B).

by 35% and remained at that level (Appendix p.2). Patients with PMR or mMR had maximum reductions in allele burden of 23% and 16%, which were not sustained. This reduction was statistically significant only among polycythemia vera patients (P<0.001) and those with a CMR (P<0.001), as previously reported 12 (Figure 2B). Molecular responses were achieved gradually over time, with a median time to response of 24 months, which did not differ by depth of response. Age was the only variable that significantly differed between patients with (median, 45 years) and without (median, 59 years) a MR (Appendix p.3). The response duration was longest among patients with a CMR (Table 2). Twenty-five patients have maintained at least a mMR, whereas 9 patients (6 polycythemia vera, 3 essential thrombocythemia) have lost their response completely. Four patients lost their response after the drug was withheld (median time from discontinuation to relapse 2 years. The other 5 lost their response while on therapy. Only one patient who achieved a CMR has relapsed after being off therapy for 16 months (CMR duration, 66 months). Among the 20 patients who achieved a PMR, 5 have sustained their best PMR, 7 are now in mMR, and 8 lost their response. Three of 5 patients have sustained their mMR.

5 lost their response while on therapy. Only one patient who achieved a CMR has relapsed after being off therapy for 16 months (CMR duration, 66 months). Among the 20 patients who achieved a PMR, 5 have sustained their best PMR, 7 are now in mMR, and 8 lost their response. Three of 5 patients have sustained their mMR. Ninety-four percent (n=33) of molecular responders also achieved a HR. However, a direct correlation between MR and HR was not observed in all cases, confirming the suggestion that HR is not a useful objective response measure in essential thrombocythemia and polycythemia vera.14,15 For example, some patients who did not achieve a MR achieved a HR (n=11) and vice versa (2 patients with mMR never achieved a HR (Appendix p.3).

MR and HR was not observed in all cases, confirming the suggestion that HR is not a useful objective response measure in essential thrombocythemia and polycythemia vera.14,15 For example, some patients who did not achieve a MR achieved a HR (n=11) and vice versa (2 patients with mMR never achieved a HR (Appendix p.3). Disease–associated clinical complications Over the course of the study, there were 8 major vascular thromboembolic (VTE) events, 3 of which were provoked by heart catheterization, elective chest surgery and angiogram. The incidence rate of major unprovoked VTE in enrolled patients during the entire study follow-up was 1.22 per 100 person-years. All 3 provoked VTEs occurred very shortly after starting therapy (median 2 months), and only 1 of these patients had achieved a HR. The other two patients continued therapy after the VTE, and one is still receiving treatment after more than 96 months without experiencing another VTE. Five patients (3 of whom were in CHR) experienced unprovoked VTEs (i.e., there was no discernable cause) after a median time on therapy of 38 months (range, 14–60 months) (Appendix p.3). Two of them were younger than 60 with no history of thrombosis. In addition, 1 patient had a serious unprovoked cerebrovascular hemorrhage while in CHR after 3 years on therapy.

enced unprovoked VTEs (i.e., there was no discernable cause) after a median time on therapy of 38 months (range, 14–60 months) (Appendix p.3). Two of them were younger than 60 with no history of thrombosis. In addition, 1 patient had a serious unprovoked cerebrovascular hemorrhage while in CHR after 3 years on therapy. Control of disease progression Seven (8%) patients had disease progression while on therapy: six had progression to MF and one transformed to AML (Appendix p.4). All seven patients were women (4 essential thrombocythemia/3 polycythemia vera) and the median time to transformation was 40 months (IQR, 18–61 mo). Two of the polycythemia vera patients had splenomegaly and 3 patients had platelet counts > 1000 x 109/L (2 essential thrombocythemia, 1 polycythemia vera) at the start of therapy. All transformations were documented with repeated bone marrow biopsies and met the IWG-MRT diagnostic criteria for post-essential thrombocythemia or post-polycythemia myelofibrosis or had bone marrow blasts > 20% (for acute myeloid leukemia).16 The cumulative incidence of MF progression among all enrolled patients at 5 years was 11%. However, given the small number of patients with progression to MF this number may not accurately reflect the true cumulative incidence. Cytogenetic analysis of BM biopsies at the time of progression showed no clonal evolution. Sequencing of a 48-gene panel in samples from 6 of 7 patients at baseline and time of transformation showed that two of these patients had acquired a mutation (1 DNMT3A, 1 ASXL1) at the time of transformation. The rate of transformation to MF/AML did not differ when compared with a cohort of age- and gender-matched historical patients not treated with PEG-IFN-α-2a (Figure 3). In addition, we observed no significant differences in disease duration, prior therapies, and cytogenetic/molecular features between patients with or without transformation. Five patients had been in HR at the last assessment before transformation (approximately 6 months earlier); one patient who transformed to AML (patient 7, Appendix) had been off therapy for 6 months prior to transformation for an unrelated reason (knee surgery). One patient was diagnosed with CML after transformation to MF. This patient had tested negative for the BCR-ABL transcript at least 3 times before and during the study. The BCR-ABL transcript was only detected after the patient had discontinued therapy due to progression to MF (biopsy confirmed).

reason (knee surgery). One patient was diagnosed with CML after transformation to MF. This patient had tested negative for the BCR-ABL transcript at least 3 times before and during the study. The BCR-ABL transcript was only detected after the patient had discontinued therapy due to progression to MF (biopsy confirmed). Safety and Tolerability The distribution of hematologic and non-hematologic adverse events (AE) is outlined in Table 5. Among all patients, fatigue (75%), muscle pain (52%), nausea/vomiting and diarrhea (44%), and depression (32%) were the most common AEs. Severe hematologic AEs (Grade 3/4 or recurrent events despite dose reductions) occurred in 89% (n=22/25) of patients with polycythemia vera and 83% (n=20/24) of those with essential thrombocythemia. The AE rate decreased with time on therapy, yet did not completely disappear (Figure 4). New G3/4 toxicities unrelated to dose occurred ≥24 months from start of therapy in 10–17% of patients annually. The most common late AEs were fatigue (prevalent in all years), anemia and neutropenia (highest in the 3rd & 6th year), and depression (highest in 4th–6th year). Three patients with essential thrombocythemia experienced 4 episodes of G4 neutropenia, 3 of which occurred on a dose < 90 mcg/week.

17% of patients annually. The most common late AEs were fatigue (prevalent in all years), anemia and neutropenia (highest in the 3rd & 6th year), and depression (highest in 4th–6th year). Three patients with essential thrombocythemia experienced 4 episodes of G4 neutropenia, 3 of which occurred on a dose < 90 mcg/week. Four patients (3 essential thrombocythemia, 1 polycythemia vera) developed autoimmune toxicities after a median time on therapy of 47.5 months (range, 26 – 78 months). All cases were biopsy proven and included hepatitis, central nervous system vasculitis, lupus nephritis and other presentations (Sjogren syndrome, dermatitis and vasculitis). Screening for auto-antibodies was only performed in patients with a clinical presentation suspicious for autoimmune disease (e.g., significant musculoskeletal or skin symptoms or any other atypical presentation). Thyroid function tests were done for all patients. None of the patients had a history of autoimmune disease, which was an exclusion criterion. Autoimmune thyroiditis, the most frequently reported autoimmune side effect of IFN-α 17, was observed in 15 patients (18%), but only 2 were severe (grade 3) enough to warrant treatment discontinuation.

ere done for all patients. None of the patients had a history of autoimmune disease, which was an exclusion criterion. Autoimmune thyroiditis, the most frequently reported autoimmune side effect of IFN-α 17, was observed in 15 patients (18%), but only 2 were severe (grade 3) enough to warrant treatment discontinuation. Overall, the dose had to be adjusted over the course of treatment, either for toxicity or lack of efficacy, in all but 2 patients. Eighteen patients (22%) discontinued therapy due to drug-related toxicities, G1–2 in 8 (10%) or G3–4 in 10 (12%) (Figure 1). Discontinuation rates were not correlated with PEG-IFN-α-2a dosage (Appendix p.5). The median time on therapy for these patients was 11 months (range, 2–60), although 4 patients requested discontinuation after < 6 months. Half of these patients had more than 1 type of toxicity, with the most common being neuropsychiatric (n=5 patients), gastrointestinal (n=4), and hematologic (n=3).

e (Appendix p.5). The median time on therapy for these patients was 11 months (range, 2–60), although 4 patients requested discontinuation after < 6 months. Half of these patients had more than 1 type of toxicity, with the most common being neuropsychiatric (n=5 patients), gastrointestinal (n=4), and hematologic (n=3). An additional 12 patients had therapy held for >6 months due to toxicity, with a median time on hold of 29 months (range, 7–79). The most common toxicities leading to significant treatment interruptions were G3 neutropenia and multiple G2 toxicities, such as anemia, fatigue, musculoskeletal pain, diarrhea, neuropathy, and depression. Ten of 12 patients were rechallenged at a lower dose of PEG-IFN-a-2a, but because of persistent and recurrent toxicities (mostly G3 hematologic) they remained off therapy. The other two are being treated with a very low dose (45 mcg every 4–6 weeks), despite similar, though less severe, (only G1 non-hematologic) side effects (musculoskeletal, gastrointestinal, rash), which they have deemed tolerable. Other reasons for discontinuation included motor vehicle accident (n=1); loss to follow-up (n=3); death (n=3); other malignancy (n=2); and financial (n=7). Three patients died while on study, but none were thought to be related to the study drug. One patient died of central pontine myelinolysis due to rapid correction of G3 hyponatremia, one due to complications from severe aortic stenosis and pulmonary hypertension, and one as a consequence of a motor vehicle accident.

. Three patients died while on study, but none were thought to be related to the study drug. One patient died of central pontine myelinolysis due to rapid correction of G3 hyponatremia, one due to complications from severe aortic stenosis and pulmonary hypertension, and one as a consequence of a motor vehicle accident. Long-term responders Thirty-two (38%; essential thrombocythemia 18, polycythemia vera 14) patients are still enrolled in the study (Appendix p.6), with 24 on active treatment. Nineteen are in HR. Eighteen of 24 (75%) patients are on a dose ≤ 90 mcg per week. Eight patients are having therapy held (7 essential thrombocythemia, 1 polycythemia vera) due to toxicity (n=5) or for financial reasons (n=3). Two patients have been holding the drug for 6 and 6.5 years. While on hold, 3 of 8 patients have lost their response (2 HR, 1 MR). Despite losing their responses, all 3 patients are symptom-free and prefer to stay on the study with active observation.

cythemia vera) due to toxicity (n=5) or for financial reasons (n=3). Two patients have been holding the drug for 6 and 6.5 years. While on hold, 3 of 8 patients have lost their response (2 HR, 1 MR). Despite losing their responses, all 3 patients are symptom-free and prefer to stay on the study with active observation. Discussion Our long-term follow-up of a phase 2 study of PEG-IFN-α-2a in 83 patients with essential thrombocythemia or polycythemia vera shows that hematologic and molecular responses are durable in some patients and provides some important additional observations. First, patients may continue to derive a clinical benefit from PEG-IFN-α-2a (i.e., remain symptom free with no organomegaly or thrombotic event) even after losing their hematologic or molecular response. Second, only CMRs are durable, and in selected cases, can be sustained after discontinuation of therapy. Third, the clinical activity of PEG-IFN-α-2a is not correlated with the JAK2 mutation status. Fourth, toxicities unrelated to dosage may develop and can be treatment limiting in some patients, even after a long exposure to the drug. Lastly, disease-related vascular complications and/or progression to MF can still occur on therapy.

linical activity of PEG-IFN-α-2a is not correlated with the JAK2 mutation status. Fourth, toxicities unrelated to dosage may develop and can be treatment limiting in some patients, even after a long exposure to the drug. Lastly, disease-related vascular complications and/or progression to MF can still occur on therapy. Despite losing a HR, 13 patients continue to have significant clinical benefit and have elected to stay on study. We also found that neither molecular status nor achievement of a MR is a prerequisite for obtaining a HR or clinical benefit. For example, 11 (20%) JAK2-positive patients never achieved a MR, yet had HRs with a median duration of 35 months (range, 6–82). Similar findings have been reported by others18,19. In addition, 95% of JAK2-negative patients achieved a HR of similar duration to that of JAK2-positive patients, and half of them are still on study, suggesting that PEG-IFN-α-2a has similar clinical efficacy in JAK2-positive and -negative MPNs.

5 months (range, 6–82). Similar findings have been reported by others18,19. In addition, 95% of JAK2-negative patients achieved a HR of similar duration to that of JAK2-positive patients, and half of them are still on study, suggesting that PEG-IFN-α-2a has similar clinical efficacy in JAK2-positive and -negative MPNs. Nine of ten patients who achieved a CMR had durable remissions (median, 69 months), while only 7 patients with a PMR or mMR maintained their best response. These findings are in contrast to those of Kiladjian et al. 20, showing no increase in JAK2 allele burden during follow-up (median follow-up, 31.4 months) regardless of depth of response. This difference may be explained in part by our longer follow-up, but it also reinforces the hypothesis that resistance may occur during the course of therapy through the acquisition of additional mutations (e.g., TET2, DNMT3a, ASXL1, IDH1, IDH2, EZH2, TP53) in clones that are not suppressed by PEG-IFN-α-2a 11,21,22. Our finding that no clinical, demographic or treatment-related factors were associated with the achievement of a MR or its duration further strengthens this hypothesis. Additional molecular testing of patients treated for more than 24 months and after long-term follow-up is currently being performed in our patients to further investigate this question. Twenty eight percent (n=10) of patients with a MR achieved a CMR, similar to previously reported results 23, suggesting that PEG-IFN-α-2a may eradicate the JAK2 clone in selected cases, providing a “functional cure.” Moreover, 3 patients have maintained their CMR after discontinuation of therapy for 1.5, 4.5 and 6.5 years, similar to what has been reported by others 24,25. However, studies have shown endogenous erythroid colony formation activity and/or the presence of the JAK2V617F allele at very low levels in more sensitive assays (level of detectability, 0.1%) 11,25,26 in patients who achieved a CMR with PEG-IFN-α. Therefore, we cannot assume the complete disappearance of the JAK2V617F clone in our patients. On the other hand, a very important clinical observation is that patients who achieved a CMR (our level of detectability) derived the longest clinical benefit. It remains to be determined whether deeper molecular responses really translate into better clinical outcome compared with patients that only maintain hematologic control. The fact that patients who achieved a CMR did not experience disease progression may support this hypothesis.

derived the longest clinical benefit. It remains to be determined whether deeper molecular responses really translate into better clinical outcome compared with patients that only maintain hematologic control. The fact that patients who achieved a CMR did not experience disease progression may support this hypothesis. Some reports have suggested that prolonged therapy with PEG-IFN-α could lead to a further reduction in JAK2V617F allele burden 20, a finding that was not confirmed in our study. In most of our patients, the JAK2V617F allele burden decreased within the first 24 months and subsequently increased over time; a continuous reduction was seen in only 3 patients (without a CMR) on long-term therapy. The long-term safety data from our study shows that the type and severity of late adverse events are similar to those observed earlier in the course of therapy; however, new late adverse events do occur even after 60 months on treatment, are often unrelated to dose or response status, and are therapy-limiting. The overall discontinuation rate due to toxicity was 22%, which is similar to that reported in other studies with shorter follow-up 20.

earlier in the course of therapy; however, new late adverse events do occur even after 60 months on treatment, are often unrelated to dose or response status, and are therapy-limiting. The overall discontinuation rate due to toxicity was 22%, which is similar to that reported in other studies with shorter follow-up 20. Unexpectedly, we found that 10% of patients in our study experienced thrombotic events on therapy, which is in contrast to previous reports showing low rates of thrombosis during treatment with PEG-IFN-α-2a 20. One can only speculate as to why we have observed such a phenomenon. Similarly, treatment did not reduce the rate of transformation to MF/AML compared with our historical control cohort. Limitations of our study include its retrospective nature and the fact that only 44 of 83 patients were followed after discontinuation and some follow-ups were in the form of telephone conversations; therefore, rates of transformation to MF and thrombotic events may be underestimated. In addition, evaluation of bone marrow response was not a prespecified endpoint. However, changes in BM histology and response assessment with reticulin and trichrome staining are being evaluated as a separate study.27 In addition, because the JAK2 testing was not performed with the most highly sensitive method, patients with CMR may have minimal residual disease that was not detected with our assay.

wever, changes in BM histology and response assessment with reticulin and trichrome staining are being evaluated as a separate study.27 In addition, because the JAK2 testing was not performed with the most highly sensitive method, patients with CMR may have minimal residual disease that was not detected with our assay. Taken together, our finding suggest that PEG-IFN-a remains a viable treatment option, especially for younger patients who want to avoid prolonged cytotoxic therapy. Lower doses minimize side effects while retaining efficacy. Patients with a history of autoimmune diseases and those with mood disorders (e.g., depression or anxiety) should be monitored more closely for side effects. Future studies of PEG-IFN-α-2a or its combination with novel immunomodulatory drugs are needed to identify patients who would derive the most benefit. Furthermore, additional objective response criteria, such as objective measurement of spleen size and changes in bone marrow histology, and symptom/quality of life assessments before and after therapy should be used to better assess clinical benefit. While no formal consensus on the optimum dose or dosing schedule exists, from our experience and that of others28,29, a starting dose of 45 mcg weekly is the optimal dosing strategy to limit AEs and maximize response. Whether novel forms of interferon 30 would be better tolerated, allowing patients to remain on therapy longer is of interest.

us on the optimum dose or dosing schedule exists, from our experience and that of others28,29, a starting dose of 45 mcg weekly is the optimal dosing strategy to limit AEs and maximize response. Whether novel forms of interferon 30 would be better tolerated, allowing patients to remain on therapy longer is of interest. Supplementary Material 1 2 Funding Source: This study was funded in part through the MD Anderson Cancer Center Support Grant P30 CA016672 from the National Cancer Institute Funding This research is supported in part by the MD Anderson Cancer Center Support Grant P30 CA016672 from the National Cancer Institute. Author Contributions: SV and HK designed and coordinated the original clinical trial. LM, KP and SV analyzed the data. LM, SV, and KJN wrote the manuscript. JC, GB, MK, ZE, and HK enrolled patients and conducted the research. All authors participated in the discussion and have reviewed and approved the current manuscript. Declaration of Interest: The authors declared no conflicts of interest. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

f the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Figure 1 Yearly discontinuation rates and reasons for treatment discontinuation. MF/AML= transformation to myelofibrosis/acute leukemia; Resp. = response, NR = no response; others [see explanation in the main text] Figure 2 Molecular responses over time stratified by A) response type and B) diagnosis. Figure 3 Cumulative probability of transformation to MF/AML (age, gender matched historical control) Figure 4 Correlation of toxicities with time and dose. Table 1 Demographic and clinical characteristics of the entire cohort. Characteristic PV (n=43) ET (n=40) total (n=83) p-value Median age, (IQR) 54 (44–63) 52 (39–62) 53.4 (43–62) p = 0.41 Males, n (%) 17 (40) 12 (30) 29 (35) p = 0.49 Females, n (%) 26 (60) 28 (70) 54 (65) p = 0.49 High risk disease, n (%) 14(33) 16(40) 30 (36) p < 0.0001 Time from diagnosis to study entry, months (IQR) 50 (13–89) 37 (14–115) 42 (14–98) p = 0.67 History of major thrombosis, n (%) 2 (4.6) 2 (5) 4 (5) p = 1.00 No. JAK2 V617F-positive patients (%) 41 (95%) 19 (48%) 60 (72)* p < 0.0001 JAK2 V617F allele burden, median (IQR) 65 (34–78) 23 (12–45) 46 (23–76)* p < 0.0001 Abnormal karyotype, n (%) 2 (5) 4 (10) 6 (7) p = 0.42 Median white blood cell count, 109/L (IQR) 11.2 (8–16) 7.2 (6–9) 9 (6–13) p < 0.0001

History of major thrombosis, n (%) 2 (4.6) 2 (5) 4 (5) p = 1.00 No. JAK2 V617F-positive patients (%) 41 (95%) 19 (48%) 60 (72)* p < 0.0001 JAK2 V617F allele burden, median (IQR) 65 (34–78) 23 (12–45) 46 (23–76)* p < 0.0001 Abnormal karyotype, n (%) 2 (5) 4 (10) 6 (7) p = 0.42 Median white blood cell count, 109/L (IQR) 11.2 (8–16) 7.2 (6–9) 9 (6–13) p < 0.0001 Median hemoglobin, g/dL (IQR) 14.2 (13–15) 13.1 (12–14) 13.8 (12–15) p = 0.06 Median platelet count, 109/L (IQR) 496 (338–786) 752 (473–1020) 592 (389–938) p < 0.0025 ¶ Significant splenomegaly, n (% of known) 7/42 (16) 0/39 (0) 7/81 (8.6) NA Median spleen size BCM in cm (IQR) 11 (4–14) NA 11 (4–14) NA Disease-related symptoms, n (%) 19 (44) 24 (60) 43 (52) p = 0.19 Phlebotomy, n (%) 32 (74%) NA 32 NA High risk disease = patients ≥ 60 or previous thrombosis; ¶ Significant splenomegaly defined as a palpable spleen > 5 cm below costal margin (BCM); NA = not applicable Table 2 Summary of clinical efficacy in all enrolled patients Characteristics Total, N= 83 PV, N= 43 ET, N= 40 P-value Median total follow-up time, months (IQR) 82.5 (69–94) 83 (65–92) 83 (74–95) 0.43 Median treatment duration, months (IQR) 68.8 (15–85) 59 (12–91) 76 (25–83) 0.49 INITIAL RESPONSE Hematologic Resp. (HR), n, (%) Overall HR 66 (80) 34 (79) 32 (80) 1.00 CHR 62 (75) 33 (77) 29 (73) 0.35 PHR 4 (5) 3 (7) 1 (3) 0.61 Median HR duration, months (IQR) Overall 65.7 (35–83) 65 (43–87) 58 (36–84) 0.38 CHR 67 (36–86) 65 (33–82) 58 (33–83) 0.21 PHR 30 (6–52) 35 (2–58) 25 NA Median time to response, months (IQR) 4 (1.3–7.3) 1.7 (1–4.7) 4.8 (2.1–12.5) 0.11

Hematologic Resp. (HR), n, (%) Overall HR 66 (80) 34 (79) 32 (80) 1.00 CHR 62 (75) 33 (77) 29 (73) 0.35 PHR 4 (5) 3 (7) 1 (3) 0.61 Median HR duration, months (IQR) Overall 65.7 (35–83) 65 (43–87) 58 (36–84) 0.38 CHR 67 (36–86) 65 (33–82) 58 (33–83) 0.21 PHR 30 (6–52) 35 (2–58) 25 NA Median time to response, months (IQR) 4 (1.3–7.3) 1.7 (1–4.7) 4.8 (2.1–12.5) 0.11 Molecular Resp. (MR), n (%) Overall MR 35/55 (63) 22 (63) 13 (37) 0.40 CMR 10 (18) 7 (20) 3 (9) 0.71 PMR 20 (36) 14 (40) 6 (17) 0.48 mMR 5 (9) 1 (3) 4 (11) 0.06 Median MR duration, months (IQR) Total 53.4 (24–70) 58 (38–77) 57 (14–60) 0.12 CMR 69 (54–77) 70 (59–77) 54 (11–76) 0.13 PMR 49 (24–65) 50 (22–68) 47 (32–60) 0.67 mMR 18 (4–46.4) 17.7 18 (6–51) NA Median time to MR, months (range) 24 (12–35) 24 (12–30) 23 (12–42) 0.99 RESPONSE AT LAST FOLLOW-UP Hematologic response, n (%) Overall HR 26/66 (39) 13/34 13/32 HR type, n (% of overall HR) CHR 25 (96) 12 (92) 13 (100) Molecular response, n (%) Overall MR 25/35 (71) 17 (68) 8 (32) MR subtype, n (% of overall MR) CMR 9 (36) 6 (35) 3 (38) PMR 6 (24) 5 (29) 1 (13) Comments: The response at last follow-up is reported as seen at the time of last follow-up regardless of initial response. Table 3 Adverse events occurring in ≥ 10% of patients (related and unrelated)

Molecular response, n (%) Overall MR 25/35 (71) 17 (68) 8 (32) MR subtype, n (% of overall MR) CMR 9 (36) 6 (35) 3 (38) PMR 6 (24) 5 (29) 1 (13) Comments: The response at last follow-up is reported as seen at the time of last follow-up regardless of initial response. Table 3 Adverse events occurring in ≥ 10% of patients (related and unrelated) SAFETY TYPE ET, N (%) PV, N (%) GRADE 1–2 N (%) GRADE 3 N (%) GRADE 4 N (%) GRADE 5 N (%) PATIENTS WITH AE, N (%) Any AE 40 (100) 43 (100) 26 (33) 53 (64) 4 (5) 3 (4) Recurrent AE 38 (95) 36 (84) 61 (84) 13 (16) AE SUBTYPES, N (%) Musculoskeletal 36 (90) 37 (86) 67 (92) 6 (8) Neurological 26 (65) 27 (63) 51 (96) 2 (4) Psychological 17 (43) 21 (49) 34 (89) 4 (11) GIT (except for LFT abn) 25 (63) 20 (47) 43 (80) 11 (20) Dermatologic 10 (25) 8 (19) 16 (89) 2 (11) Infection/fever 13 (33) 13 (30) 23 (88) 3 (12) Respiratory 10 (25) 13 (30) 21 (91) 2 (9) Cardiovascular 5 (13) 8 (19) 10 (77) 3 (23) Hypothyroidism 10 (25) 5 (12) 13 (87) 2 (13) SELECTED ABNORMAL LABS Leukopenia/neutropenia 17 (43) 20 (47) 16 (43) 17 (46) 4 (11) Thrombocytopenia 15 (35) 3 (7) a 17 (94) 1 (6) Anemia 16 (40) 20 (47) 35 (97) 1 (3) LFT elevation 17 (43) 10 (23) 22 (82) 5 (18) DEATHS ON STUDY* Central pontine myelinolysis 1 (33) Chronic cardiac disease 1 (33) Motor vehicle accident 1 (33) * None of the deaths were related to the study drug. Table 4 Summary of long-term safety

SAFETY TYPE ET, N (%) PV, N (%) GRADE 1–2 N (%) GRADE 3 N (%) GRADE 4 N (%) GRADE 5 N (%) PATIENTS WITH AE, N (%) Any AE 40 (100) 43 (100) 26 (33) 53 (64) 4 (5) 3 (4) Recurrent AE 38 (95) 36 (84) 61 (84) 13 (16) AE SUBTYPES, N (%) Musculoskeletal 36 (90) 37 (86) 67 (92) 6 (8) Neurological 26 (65) 27 (63) 51 (96) 2 (4) Psychological 17 (43) 21 (49) 34 (89) 4 (11) GIT (except for LFT abn) 25 (63) 20 (47) 43 (80) 11 (20) Dermatologic 10 (25) 8 (19) 16 (89) 2 (11) Infection/fever 13 (33) 13 (30) 23 (88) 3 (12) Respiratory 10 (25) 13 (30) 21 (91) 2 (9) Cardiovascular 5 (13) 8 (19) 10 (77) 3 (23) Hypothyroidism 10 (25) 5 (12) 13 (87) 2 (13) SELECTED ABNORMAL LABS Leukopenia/neutropenia 17 (43) 20 (47) 16 (43) 17 (46) 4 (11) Thrombocytopenia 15 (35) 3 (7) a 17 (94) 1 (6) Anemia 16 (40) 20 (47) 35 (97) 1 (3) LFT elevation 17 (43) 10 (23) 22 (82) 5 (18) DEATHS ON STUDY* Central pontine myelinolysis 1 (33) Chronic cardiac disease 1 (33) Motor vehicle accident 1 (33) * None of the deaths were related to the study drug. Table 4 Summary of long-term safety Long-term safety 3th year 4th year 5th year >6th year New AE with RX> 24 mos Total No of pts on RX 60 53 50 41 All grades Patients, n (%) 10 (17) 6 (11) 5 (10) 10 (24) Events, n 13 14 10 18 Grade 1 – 2 Patients, n (%) 6 (60) 2 (33) 4 (80) 9 (90) Events, n (%) 9 (69) 10 (71) 9 (90) 17 (94) Grade 3 Patients, n (%) 4 (40) 4 (67) 1 (20) 1 (10) Events, n (%) 4 (31) 4 (29) 1 (10) 1 (6) Grade 4 Patients, n (%) 1 (17) 2 (40) 1 (10) Events, n (%) 1 (7) 2 (20) 1 (6) Deaths on Study Patients, n (%) 1 (6) * Grade 4 toxicities occurred only in patients with ET.

(80) 9 (90) Events, n (%) 9 (69) 10 (71) 9 (90) 17 (94) Grade 3 Patients, n (%) 4 (40) 4 (67) 1 (20) 1 (10) Events, n (%) 4 (31) 4 (29) 1 (10) 1 (6) Grade 4 Patients, n (%) 1 (17) 2 (40) 1 (10) Events, n (%) 1 (7) 2 (20) 1 (6) Deaths on Study Patients, n (%) 1 (6) * Grade 4 toxicities occurred only in patients with ET. Research in context Evidence before this study Previous studies have shown that pegylated interferon is highly effective in patients with myeloproliferative neoplasms, particularly essential thrombocythemia and polycythemia vera. However, safety and tolerability has limited its use. All studies have shown a high rate of hematologic and molecular responses, with a continuous reduction in JAK2 allele burden, a unique finding not observed with other agents. All study reports were encouraging, showing disease control with no or minimal disease-related vascular complications and no progression to more aggressive myelofibrosis. However, reports of longer follow-up beyond 45 months, are lacking.

reduction in JAK2 allele burden, a unique finding not observed with other agents. All study reports were encouraging, showing disease control with no or minimal disease-related vascular complications and no progression to more aggressive myelofibrosis. However, reports of longer follow-up beyond 45 months, are lacking. Added value of this study In this long-term prospective analysis of patients with essential thrombocythemia and polycythemia vera treated with pegylated interferon alpha-2a, we report efficacy and safety data after a follow-up of 7 years, almost twice as long as in previously published studies. Our analysis confirmed the high efficacy (disease control with hematologic and molecular responses) of pegylated interferon alpha-2a, with a median response duration of up to 6 years. In contrast to earlier studies, a continuous reduction in JAK2 allele burden was not observed with longer follow-up. We could not identify any factors predictive of loss of response, which happens over time; however, the deepest responses lasted the longest. Late toxicities of a similar type and severity as those observed earlier on therapy do occur, regardless of dose, and may be treatment limiting. The discontinuation rate (~20%) was similar to that reported in studies with short term follow-up. Vascular events, and disease progression were observed while on therapy, and the rate of progression to MF or the blast phase (AML) was similar to that of a matched control group.

s of dose, and may be treatment limiting. The discontinuation rate (~20%) was similar to that reported in studies with short term follow-up. Vascular events, and disease progression were observed while on therapy, and the rate of progression to MF or the blast phase (AML) was similar to that of a matched control group. Implications of all the available evidence Pegylated interferon can provide excellent disease control in patients with essential thrombocythemia and polycythemia vera, with an average response duration of 6 years. A deep response was achieved in some patients and seems to provide the longest benefit, including protection against vascular complications and disease progression. The long term toxicity rate is acceptable; however, late toxicities may occur and therefore patients should be carefully monitored.

Introduction B-lymphoproliferative disorders include a wide range of both non-malignant and malignant conditions. The malignant diseases represent the most frequent adult haematological cancers, accounting for more than 50% of leukaemia and more than 90% of non-Hodgkin lymphoma diagnoses.1 However, major geographical variation exists in the reported incidence of B-lymphoproliferative disorders. According to GLOBOCAN 2012,2 age-standardised incidence per 100 000 person-years for leukaemia and non-Hodgkin lymphoma in the WHO African region is 6·4, compared with 18·2 in the WHO Americas region and 21·3 in the WHO European region. Chronic lymphocytic leukaemia (CLL) is the most common adult leukaemia in Europe and the USA but incidence seems to be low in Africa, with some data suggesting that other B-cell malignancies, such as lymphoplasmacytic lymphoma and marginal zone lymphomas, might be more frequent than CLL.3 Several factors can affect the reported differences in incidence, including variable quality of access to cancer care, differences in age distribution, and failure to diagnose B-lymphoproliferative disorders, especially in the elderly.2 Monoclonal B-cell lymphocytosis (MBL) is a precursor lesion of B-lymphoproliferative disorders and provides a useful model for the study of B-cell lymphomagenesis, since it can be screened for in otherwise healthy people,4, 5, 6 thereby allowing a comparison of prevalence across geographical regions in a way that is independent of health-care provision.

L) is a precursor lesion of B-lymphoproliferative disorders and provides a useful model for the study of B-cell lymphomagenesis, since it can be screened for in otherwise healthy people,4, 5, 6 thereby allowing a comparison of prevalence across geographical regions in a way that is independent of health-care provision. MBL with a CLL phenotype is common in Europe and the USA and, compared with other MBL subtypes, has been relatively well characterised in terms of phenotype, genotype, and B-cell receptor immunoglobulin gene repertoire.7, 8, 9, 10, 11, 12, 13 Furthermore, the incidence, rate of disease progression, and inherited susceptibility single-nucleotide polymorphism (SNP) patterns are well established in US and European populations.14, 15, 16, 17 MBL of non-CLL phenotype is also increasingly recognised, and preliminary data suggest that this subtype is less common than CLL-phenotype MBL6, 18, 19 but might be a precursor of entities within the spectrum of marginal zone lymphomas, typically of splenic type.20, 21 Research in context Evidence before this study

MBL with a CLL phenotype is common in Europe and the USA and, compared with other MBL subtypes, has been relatively well characterised in terms of phenotype, genotype, and B-cell receptor immunoglobulin gene repertoire.7, 8, 9, 10, 11, 12, 13 Furthermore, the incidence, rate of disease progression, and inherited susceptibility single-nucleotide polymorphism (SNP) patterns are well established in US and European populations.14, 15, 16, 17 MBL of non-CLL phenotype is also increasingly recognised, and preliminary data suggest that this subtype is less common than CLL-phenotype MBL6, 18, 19 but might be a precursor of entities within the spectrum of marginal zone lymphomas, typically of splenic type.20, 21 Research in context Evidence before this study Monoclonal B-cell lymphocytosis (MBL) is a precursor to B-cell haematological malignancy that can be used as a model for the study of B-cell lymphomagenesis. MBL with a chronic lymphocytic leukaemia phenotype is common in Europe and the USA and, compared with other MBL subtypes, has been relatively well characterised in terms of phenotype, genotype, and B-cell receptor immunoglobulin gene repertoire. Furthermore, the prevalence, rate of disease progression, and inherited susceptibility single-nucleotide polymorphisms patterns are well established in the USA and Europe. We found no reports of MBL prevalence in Africa on PubMed up to Dec 31, 2015 (search term “monoclonal B-cell lymphocytosis Africa”). Added value of this study

Monoclonal B-cell lymphocytosis (MBL) is a precursor to B-cell haematological malignancy that can be used as a model for the study of B-cell lymphomagenesis. MBL with a chronic lymphocytic leukaemia phenotype is common in Europe and the USA and, compared with other MBL subtypes, has been relatively well characterised in terms of phenotype, genotype, and B-cell receptor immunoglobulin gene repertoire. Furthermore, the prevalence, rate of disease progression, and inherited susceptibility single-nucleotide polymorphisms patterns are well established in the USA and Europe. We found no reports of MBL prevalence in Africa on PubMed up to Dec 31, 2015 (search term “monoclonal B-cell lymphocytosis Africa”). Added value of this study To our knowledge, this is the first study to provide data for the prevalence and phenotype of MBL in an African country. Overall MBL prevalence was slightly higher in rural Uganda than in the UK but chronic lymphocytic leukaemia-phenotype MBL was infrequent whereas CD5-negative MBL was much more common in rural Uganda and showed substantial phenotypic differences compared with the UK population. Implications of all the available evidence Although MBL is common in both rural Uganda and the UK, the variation in phenotype shows that differences in environmental exposure, inherited susceptibility, or both factors affect the type of B-cell neoplasms that can develop.

To our knowledge, this is the first study to provide data for the prevalence and phenotype of MBL in an African country. Overall MBL prevalence was slightly higher in rural Uganda than in the UK but chronic lymphocytic leukaemia-phenotype MBL was infrequent whereas CD5-negative MBL was much more common in rural Uganda and showed substantial phenotypic differences compared with the UK population. Implications of all the available evidence Although MBL is common in both rural Uganda and the UK, the variation in phenotype shows that differences in environmental exposure, inherited susceptibility, or both factors affect the type of B-cell neoplasms that can develop. Therefore, the study of MBL in Uganda might provide insights into the pathogenesis of B-lymphoproliferative disorders in sub-Saharan Africa, and comparison with data from the UK could deepen these insights. In this study, we aimed to assess the prevalence of MBL in a rural population in Uganda and compare these findings with those from a study of MBL in a UK population.

insights into the pathogenesis of B-lymphoproliferative disorders in sub-Saharan Africa, and comparison with data from the UK could deepen these insights. In this study, we aimed to assess the prevalence of MBL in a rural population in Uganda and compare these findings with those from a study of MBL in a UK population. Methods Study design and data collection In this cross-sectional study, we analysed samples from participants in rural Uganda and a UK hospital-based population (appendix pp 1–2). We recruited volunteers aged at least 45 years who were seronegative for HIV-1 from the Ugandan General Population Cohort, a community-based open dynamic cohort with more than 22 000 people at present. Whole-blood samples were obtained from the participants, stored at 4°C overnight, and analysed within 24 h. We used a previously reported method16 to obtain whole-blood samples from anonymised waste material of UK hospital outpatients and primary care patients with a normal blood count and no history of cancer. These patients were selected to match the age and sex distribution of the Ugandan participants. Ethics approval for sample collection in Uganda was provided by the Uganda Virus Research Institute Scientific Ethics Committee and the Uganda Council for Science and Technology, and for sample collection in the UK by the Leeds Teaching Hospitals National Health Service Trust Ethics Review Board. In Uganda, study personnel read the consent form to all participants, who signed or thumb-printed (for those who were illiterate) the forms.

Committee and the Uganda Council for Science and Technology, and for sample collection in the UK by the Leeds Teaching Hospitals National Health Service Trust Ethics Review Board. In Uganda, study personnel read the consent form to all participants, who signed or thumb-printed (for those who were illiterate) the forms. Procedures MBL is a condition in which the B-cell surface immunoglobulin light chain κ-to-λ ratio is heavily skewed in total B-cells or in a B-cell subset. In previous studies the detection of MBL representing a subset of B cells (ie, with predominantly polyclonal B cells) has been limited to CLL-phenotype cells whereas CD5-negative MBL has represented a skewed κ-to-λ ratio in total B cells.5, 6, 18, 19 However, in some individuals, immunoglobulin light-chain skewing could be detected within B-cell subsets defined by CXCR5 (CD185) and LAIR1 (CD305) expression. In European populations, these markers are strongly expressed in most normal B cells but show aberrant expression in most B-lymphoproliferative disorders.22 LAIR1 (CD305) expression varies according to the type of B-lymphoproliferative disorder: germinal-centre B-cell malignancies typically have no LAIR1 expression; Waldenström's macroglobulinaemia usually have either no LAIR1 expression or intermediate levels, often with both a LAIR1-negative and a LAIR1-positive component; and hairy cell leukaemia and splenic lymphomas usually have stronger than normal expression.22

-centre B-cell malignancies typically have no LAIR1 expression; Waldenström's macroglobulinaemia usually have either no LAIR1 expression or intermediate levels, often with both a LAIR1-negative and a LAIR1-positive component; and hairy cell leukaemia and splenic lymphomas usually have stronger than normal expression.22 We used flow cytometry to determine the presence of MBL, defined according to the standard diagnostic criteria23—κ-to-λ ratio greater than 3:1 or less than 0·3:1, or more than 25% of B cells without surface immunoglobulin or expressing low levels of surface immunoglobulin either in the total B-cell population (CD19 positive, CD20 positive), or in one or more pre-specified B-cell subsets: CD5 positive, CD20 weak, surface immunoglobulin weak (CLL-phenotye); CD10 positive; CD305 positive and CD185 negative; CD305 positive and CD185 positive; CD305 negative and CD185 negative; or CD305 negative and CD185 positive. The threshold for positivity was validated against internal cellular controls in each case.

: CD5 positive, CD20 weak, surface immunoglobulin weak (CLL-phenotye); CD10 positive; CD305 positive and CD185 negative; CD305 positive and CD185 positive; CD305 negative and CD185 negative; or CD305 negative and CD185 positive. The threshold for positivity was validated against internal cellular controls in each case. We analysed the immunoglobulin gene repertoire of 13 Ugandan individuals selected to represent each of the MBL categories (three with CLL-type MBL, three with CD5-negative MBL, three with a monoclonal B-cell subset) and individuals with polyclonal B-cells (four individuals). IGHV–IGHD–IGHJ gene rearrangements were amplified from genomic DNA (200 ng, extracted from at least 2 × 106 peripheral blood mononuclear cells) and sequenced according to the standard Roche 454 GS-Junior sequencing protocol. Productive, in-frame sequences were analysed using IMGT HighV-Quest software, version 1.3.0.24 Genotype data were generated with the Illumina HumanOmni2.5 BeadChip array at the Sanger Institute (Cambridge, UK) for 5000 participants of the General Population Cohort.25 The data were assessed for six SNPs reported to be associated with both CLL and MBL, and compared with published UK risk-allele frequencies.14 Comparisons with African American people were made using data from the 1000 Genomes project.

Institute (Cambridge, UK) for 5000 participants of the General Population Cohort.25 The data were assessed for six SNPs reported to be associated with both CLL and MBL, and compared with published UK risk-allele frequencies.14 Comparisons with African American people were made using data from the 1000 Genomes project. Statistical analysis No formal power calculations were done because of insufficient data from Africa, and sample size was determined pragmatically. We compared differences in the proportion of cases with CLL-phenotype MBL and CD5-negative MBL and in absolute monoclonal B-cell count between the Ugandan and UK samples. We assessed significance using Fisher's exact test (two-tailed) or the Wilcoxon-Mann-Whitney U-test (two-tailed) as appropriate, using IBM SPSS version 24. 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 Between Jan 15 and Dec 18, 2012, we obtained samples from 302 Ugandan volunteers and 302 age-and-sex-matched UK individuals. 121 (40%) participants from Uganda were women aged 40–60 years, 47 (16%) were men aged 40–60 years, 90 (30%) were women older than 60 years, and 44 (15%) were men older than 60 years. The UK population was matched exactly by sex and age.

ed samples from 302 Ugandan volunteers and 302 age-and-sex-matched UK individuals. 121 (40%) participants from Uganda were women aged 40–60 years, 47 (16%) were men aged 40–60 years, 90 (30%) were women older than 60 years, and 44 (15%) were men older than 60 years. The UK population was matched exactly by sex and age. MBL was detected in 42 (14%) of 302 Ugandan participants and 25 (8%) of 302 UK participants (p=0·038). Monoclonal CLL-phenotype B cells were detected in three (1%) Ugandan participants and in 21 (7%) UK participants (p=0·00021; figure 1). By contrast, CD5-negative MBL was more prevalent in the Ugandan participants (41 [14%] individuals, of whom two [5%] also had CLL-phenotype MBL detectable) than in the UK participants (six [2%], of whom two [33%] also had CLL-phenotype MBL detectable; p<0·0001; figure 1). A monoclonal B-cell subset was detectable in 93 (31%) Ugandan participants and 21 (7%) UK participants (p<0·0001; figure 1). CD10-positive monoclonal B-cell populations with a germinal centre phenotype were not detected in either population.Figure 1 Prevalence of MBL in the Ugandan and UK cohorts, by B-cell phenotype (A) Proportion of participants with detectable CLL-type MBL and CD5-negative MBL according to the 2005 diagnostic criteria.22 (B) Proportion of participants with a monoclonal B-cell subset. CLL=chronic lymphocytic leukaemia. MBL=monoclonal B-cell lymphocytosis.

MBL was detected in 42 (14%) of 302 Ugandan participants and 25 (8%) of 302 UK participants (p=0·038). Monoclonal CLL-phenotype B cells were detected in three (1%) Ugandan participants and in 21 (7%) UK participants (p=0·00021; figure 1). By contrast, CD5-negative MBL was more prevalent in the Ugandan participants (41 [14%] individuals, of whom two [5%] also had CLL-phenotype MBL detectable) than in the UK participants (six [2%], of whom two [33%] also had CLL-phenotype MBL detectable; p<0·0001; figure 1). A monoclonal B-cell subset was detectable in 93 (31%) Ugandan participants and 21 (7%) UK participants (p<0·0001; figure 1). CD10-positive monoclonal B-cell populations with a germinal centre phenotype were not detected in either population.Figure 1 Prevalence of MBL in the Ugandan and UK cohorts, by B-cell phenotype (A) Proportion of participants with detectable CLL-type MBL and CD5-negative MBL according to the 2005 diagnostic criteria.22 (B) Proportion of participants with a monoclonal B-cell subset. CLL=chronic lymphocytic leukaemia. MBL=monoclonal B-cell lymphocytosis. In addition to differences in the phenotype of monoclonal B cells between the two cohorts, the absolute counts of monoclonal B cells also differed significantly (figure 2). All three Ugandan participants with CLL-phenotype MBL had absolute monoclonal B-cell count below one cell per μL and close to the detection limit of the assay. By contrast, in the 21 UK participants with CLL-phenotype MBL, the median absolute number of circulating neoplastic cells was 5 cells per μL (IQR 2–12, range 1–1773). Seven (2%) UK participants had more than ten CLL-phenotype cells per μL, compared with none in the Ugandan participants (p=0·015). Although the prevalence of CD5-negative MBL was higher in the Ugandan participants than in the UK participants, the median absolute B-cell count was similar (227 [IQR 152–345, range 56–947] cells per μL in the Ugandan cohort vs 135 [105–177, 69–503] cells per μL in the UK cohort; p=0·13). However, the median absolute count of CD5-negative monoclonal B-cell subsets was significantly higher in the Ugandan cohort (seven [3–15, 1–53] cells per μL) than in the UK cohort (two [2–3, 1–17] cells per μL; p=0·012).Figure 2 Absolute monoclonal B-cell count, by phenotype

503] cells per μL in the UK cohort; p=0·13). However, the median absolute count of CD5-negative monoclonal B-cell subsets was significantly higher in the Ugandan cohort (seven [3–15, 1–53] cells per μL) than in the UK cohort (two [2–3, 1–17] cells per μL; p=0·012).Figure 2 Absolute monoclonal B-cell count, by phenotype CLL=chronic lymphocytic leukaemia. MBL=monoclonal B-cell lymphocytosis. Differences were also apparent in LAIR1 expression by CD5-negative MBL and monoclonal B-cell subsets in the two cohorts. Monoclonal B-cell expansions in the UK participants typically showed moderate to strong LAIR1 expression, whereas those in Uganda had weak or bimodal expression similar to the IgM-secreting lymphoplasmacytic disorder (figure 3). Of 41 CD5-negative MBL cases in the Ugandan cohort, 40 (98%) had LAIR1-negative cells or a combination of both LAIR1-negative and LAIR1-positive cells, compared with two (33%) of six UK cases (p=0·0040). Monoclonal B-cell subsets comprised only LAIR1-negative cells or a combination of both LAIR1-negative and LAIR1-positive cells in 85 (91%) of 93 Ugandan cases, compared with four (19%) of 21 UK cases (p<0·0001; figure 3).Figure 3 LAIR1 expression pattern in monoclonal CD5-negative B cells MBL=monoclonal B-cell lymphocytosis.

Differences were also apparent in LAIR1 expression by CD5-negative MBL and monoclonal B-cell subsets in the two cohorts. Monoclonal B-cell expansions in the UK participants typically showed moderate to strong LAIR1 expression, whereas those in Uganda had weak or bimodal expression similar to the IgM-secreting lymphoplasmacytic disorder (figure 3). Of 41 CD5-negative MBL cases in the Ugandan cohort, 40 (98%) had LAIR1-negative cells or a combination of both LAIR1-negative and LAIR1-positive cells, compared with two (33%) of six UK cases (p=0·0040). Monoclonal B-cell subsets comprised only LAIR1-negative cells or a combination of both LAIR1-negative and LAIR1-positive cells in 85 (91%) of 93 Ugandan cases, compared with four (19%) of 21 UK cases (p<0·0001; figure 3).Figure 3 LAIR1 expression pattern in monoclonal CD5-negative B cells MBL=monoclonal B-cell lymphocytosis. The high prevalence of CD5-negative MBL was unexpected; although not initially planned, we extended immunophenotyping on 18 samples from rural Uganda using a panel of 46 markers with a common backbone of CD19, CD20, CD27, CXCR5, and LAIR1 (appendix p 5). The samples were from five individuals with CD5-negative MBL, eight individuals with a monoclonal B-cell subset, and five individuals with polyclonal B-cell subsets. Although the analysis was not powered to show significant differences, some markers—including CD38, CD73, CD200, and CD307d (FcRL4)—had at least a five times difference in expression by CD5-negative MBL B cells compared with polyclonal B cells and at least a two times difference in expression by CD5-negative MBL cells compared with B cells from cases with a monoclonal B-cell subset (appendix p 9). This finding suggests that neoplastic expansions might be phenotypically discriminated in a similar manner to the detection of CLL-phenotype MBL populations.

cells and at least a two times difference in expression by CD5-negative MBL cells compared with B cells from cases with a monoclonal B-cell subset (appendix p 9). This finding suggests that neoplastic expansions might be phenotypically discriminated in a similar manner to the detection of CLL-phenotype MBL populations. Immunoglobulin gene repertoire analysis was done in 13 Ugandan individuals, with a median of 224 (range 39–2674, IQR 127-428) productive, in-frame IGHV–IGHD–IGHJ gene rearrangements sequenced per individual. Four of these individuals had no detectable MBL clones by flow cytometry and, concordantly, next-generation sequencing did not detect any significant clonal expansion. Three of these individuals were found to carry CD5-negative MBL by flow cytometry, and we detected a clonal population in each one using next-generation sequencing. Three individuals had light-chain restriction detectable by flow cytometry in a subset of B cells, and clonal expansion was identified in one of these individuals (appendix p 6).

als were found to carry CD5-negative MBL by flow cytometry, and we detected a clonal population in each one using next-generation sequencing. Three individuals had light-chain restriction detectable by flow cytometry in a subset of B cells, and clonal expansion was identified in one of these individuals (appendix p 6). In the three individuals with a suspected CLL-phenotype population present at the detection limit of the flow cytometry assay, next-generation sequencing showed no evidence of significant clonal expansion, with the most expanded clonotypes representing a 1·3%, 3·1%, and 3·9% of the B-cell receptor immunoglobulin repertoire, respectively in the three individuals. In a post-hoc analysis, sequences were cross-compared with a database of 19 464 CLL sequences including 668 stereotyped CLL subsets. In one of the three individuals with a suspected CLL-phenotype population, a single, non-expanded clonotype was found (IGHV1-2/IGHD3-3/IGHJ6, CysAlaLysGlyAlaGlnTyrTyrAspPheTrpSerGlyTyrLeuProTyrTyrTyrGlyMetAspValTrp) clustering to a minor CLL subset. In another individual with a monoclonal B-cell subset, an expanded clonotype consisting of four identical sequences was found (IGHV3-30/IGHD2-21/IGHJ4, CysValLysAspAspGlnTrpGlyProAspTyrTrp) clustering to another minor CLL subset. The third individual had no clonotype clustering to a stereotyped CLL subset.

CLL subset. In another individual with a monoclonal B-cell subset, an expanded clonotype consisting of four identical sequences was found (IGHV3-30/IGHD2-21/IGHJ4, CysValLysAspAspGlnTrpGlyProAspTyrTrp) clustering to another minor CLL subset. The third individual had no clonotype clustering to a stereotyped CLL subset. Genotype data generated for more than 5000 participants of the Ugandan General Population Cohort are shown in the appendix (p 7).25 The prevalence of six SNPs that have been shown to be significantly associated both with CLL and with CLL-phenotype MBL were compared with published data from the UK14 on prevalence in the general population and in African Americans. For all of the SNPs associated with CLL-phenotype MBL, prevalence is substantially lower in rural Uganda than in the UK and in African Americans (appendix p 7).

ated both with CLL and with CLL-phenotype MBL were compared with published data from the UK14 on prevalence in the general population and in African Americans. For all of the SNPs associated with CLL-phenotype MBL, prevalence is substantially lower in rural Uganda than in the UK and in African Americans (appendix p 7). Discussion In this cross-sectional study, we showed that the prevalence of MBL is marginally higher in rural Uganda than in the UK and that the phenotypes identified are different. By contrast, the incidence of B-lymphoproliferative disorders is reported to be significantly lower in Africa than in Europe and the USA. However, detailed epidemiological studies are compromised by the fact that most patients with indolent B-lymphoproliferative disorders such as CLL and marginal zone lymphomas are asymptomatic and might be less likely to be diagnosed in African countries than in European or North American countries.2 Therefore, more meaningful data could be obtained with population screening for the precursor disorders. MBL with a CLL phenotype is well characterised, and results from a follow-up study16 have established that annually roughly 1% of individuals with CLL-phenotype MBL will develop symptomatic CLL that requires treatment. CLL-phenotype B cells are detectable at very low cell counts (often termed low-count MBL, typically <0·01 × 109 cells per L) in 3–20% of adults in Europe and the USA, depending on assay sensitivity and age. Individuals with low-count MBL have a similar inherited susceptibility risk allele profile to those with CLL but no evidence of progression to clinical CLL.12, 13, 14, 16 CD5-negative MBL is less well characterised, mainly because of a lack of a lack of markers that differentiate neoplastic CD5-negative monoclonal B-cells from reactive monotypic expansions, but the available data suggest that it might be a precursor to marginal zone lymphoma, at least in some cases.20 However, the use of antigens such as LAIR1 and CXCR5 has improved the diagnosis of CD5-negative B-lymphoproliferative disorders,22 potentially facilitating the detection of precursor lesions.

pic expansions, but the available data suggest that it might be a precursor to marginal zone lymphoma, at least in some cases.20 However, the use of antigens such as LAIR1 and CXCR5 has improved the diagnosis of CD5-negative B-lymphoproliferative disorders,22 potentially facilitating the detection of precursor lesions. Therefore, we used flow cytometry to assess, in detail, the prevalence of MBL in a rural Ugandan population who were HIV negative. We also compared this prevalence with that in an age-and-sex-matched UK control population, since such a comparison might provide insights into the relative incidences of both CLL and marginal zone lymphomas, independent of differences in health-care provision. Monoclonal B-cell populations were found with a similar frequency in adults from rural Uganda and the UK, but the type of monoclonal B-cell expansions seemed distinct. Subpopulations of CLL-phenotype monoclonal B-cells were detected frequently in the UK cohort but very rarely in Ugandan participants. By contrast, a much higher proportion of Ugandan participants had detectable CD5-negative monoclonal B cells in the peripheral blood than adults from the UK.

expansions seemed distinct. Subpopulations of CLL-phenotype monoclonal B-cells were detected frequently in the UK cohort but very rarely in Ugandan participants. By contrast, a much higher proportion of Ugandan participants had detectable CD5-negative monoclonal B cells in the peripheral blood than adults from the UK. In a preliminary immunophenotypic analysis, we identified some differences in expression of markers, including CD38, CD73, CD200, and CD307d. In individuals with a predominantly monoclonal B-cell population, the expression pattern suggested neoplastic expansion; by contrast, in individuals with polyclonal B cells or light-chain restriction in a subset of B-cells, the pattern was suggestive of a reactive expansion. The data confirmed that phenotypically identified CD5-negative MBL is consistently associated with clonal B-cell receptor immunoglobulin expansion. We could not confirm the presence of CLL-phenotype monoclonal B cells at the molecular level. Moreover, the presence of non-CLL B-cell subpopulations with light-chain restriction does not necessarily correspond to monoclonal expansions and might reflect a reactive expansion. Further work to identify markers that discriminate between reactive and neoplastic expansions would greatly benefit the diagnostic process for post-germinal centre B-cell disorder.

ubpopulations with light-chain restriction does not necessarily correspond to monoclonal expansions and might reflect a reactive expansion. Further work to identify markers that discriminate between reactive and neoplastic expansions would greatly benefit the diagnostic process for post-germinal centre B-cell disorder. Further insights have been provided by immunoglobulin gene repertoire analysis and SNP data. Although the numbers of samples included in this study were small, our results suggest that CLL-specific IGHV–IGHD–IGHJ gene rearrangements might be detected in the rural Ugandan population without being associated with the detection of a CLL-phenotype MBL. This finding is notable because CLL-derived B-cell receptor immunoglobulin molecules have a central role in CLL pathogenesis by transducing antigen-driven26 or antigen-independent cell-autonomous signals, or both.27 Therefore, the presence of CLL-specific rearrangements without clonal expansion indicates that additional or alternative mechanisms are also necessary for the development and expansion of CLL-phenotype MBL. We also confirmed clonality in most CD5-negative MBL cases, although all cases had multiple expanded clonotypes, indicating that reactive expansions might also be common in rural Ugandans. Inherited susceptibility is an important factor in CLL: 30 common genetic variants have been identified so far and six have been confirmed in MBL.14, 28 SNP analysis in the Ugandan cohort confirmed HapMap data showing that CLL susceptibility loci are very rare in this population. Although the absence of CLL susceptibility loci does not preclude development of CLL,29 some of the SNP are putatively causal for CLL.30 This finding, in combination with the MBL data, provides strong support for the hypothesis that inherited susceptibility affects the development of specific types of B-lymphoproliferative disorder.

bsence of CLL susceptibility loci does not preclude development of CLL,29 some of the SNP are putatively causal for CLL.30 This finding, in combination with the MBL data, provides strong support for the hypothesis that inherited susceptibility affects the development of specific types of B-lymphoproliferative disorder. In the UK, the prevalence of MBL subtypes closely reflects the incidence of clinical disease. CLL accounts for a much higher proportion of B-lymphoproliferative disorders with detectable circulating disease than do other CD5-negative B-cell disorders.1 In this study, the prevalence of CD5-negative MBL in the UK cohort was similar to that reported in previous studies;5, 6, 18, 19 however, the inclusion of LAIR1 in the screening process permitted detection of a larger proportion of cases with a monoclonal B-cell subset than has previously been reported. LAIR1 is negative or weakly expressed in most B-cell disorders, moderately expressed on most polyclonal B cells, and strongly expressed in hairy cell leukaemia.22 Therefore, the use of LAIR1 expression greatly improves the sensitivity for detection of non-CLL-type monoclonal B-cell expansions, although whether these expansions represent a neoplastic precursor lesion or a reactive expansion remains to be determined. The differences seen between the Ugandan cohort and UK cohort in regards to LAIR1 expression might reflect a further predisposition of the Ugandan population to the development of particular types of B-cell neoplasms, such as Waldenström's macroglobulinaemia, which is predominantly LAIR1 negative or comprises both LAIR1-positive and LAIR1-negative B-cell subpopulations, whereas hairy cell leukaemia typically shows strong LAIR1 expression.22 The use of a hospital-based study population has been shown to have a similar MBL prevalence to population-based studies using a similar screening sensitivity.5, 6 However, the age and sex distribution of the UK study population in our study was selected to match the Ugandan cohort and might therefore under-represent population MBL prevalence. Therefore, further studies of MBL in the general UK population would be informative.

ed studies using a similar screening sensitivity.5, 6 However, the age and sex distribution of the UK study population in our study was selected to match the Ugandan cohort and might therefore under-represent population MBL prevalence. Therefore, further studies of MBL in the general UK population would be informative. Some data support the case that CD5-negative B-lymphoproliferative disorders are more common than CLL in Africa.3 The results of our study, with a higher prevalence of CD5-negative MBL than CLL-phenotype MBL, provide further evidence of geographical variation. This finding mirrors the results of Landgren and colleagues' study,31 which showed that the prevalence and incidence patterns of myeloma are reflected in a much higher incidence of the precursor lesion monoclonal gammopathy of undetermined significance in Africans and African Americans than in Caucasians. Although the use of LAIR1 and CXCR5 expression helps to identify monoclonal or monotypic B-cell expansions, neoplastic populations are difficult to define accurately with only these markers because both markers show heterogeneous—and, for LAIR1, frequently bimodal—expression in non-CLL B-lymphoproliferative disorders. Therefore, we propose that future studies should investigate the distribution of abnormal B-cell expansions by combining the sensitive approach of clonality with LAIR1 expression and a more specific phenotypic analysis—eg, the strong CD25 and weak CD22 expression seen in Waldenström's macroglobulinaemia.32 This combined approach might help to improve understanding of the potential distribution of B-cell malignancies in the general African population.

approach of clonality with LAIR1 expression and a more specific phenotypic analysis—eg, the strong CD25 and weak CD22 expression seen in Waldenström's macroglobulinaemia.32 This combined approach might help to improve understanding of the potential distribution of B-cell malignancies in the general African population. The use of an environmentally distinct population in rural Uganda in our study has some advantages for understanding the inherited susceptibility, but a limitation is that the rural Ugandan population might not be representative of the whole population in Uganda. In future studies, we aim to compare the rural and urban populations in Uganda, which would offer insights into the relative contribution of inherited susceptibility and environmental factors to the development of B-cell neoplasms. Although efforts were made to exclude people with immune suppression from this study, we cannot completely rule out the possibility that mild immune modulation was present in some participants. The small numbers of participants in our study relative to some studies in high-income countries might also affect results.

s. Although efforts were made to exclude people with immune suppression from this study, we cannot completely rule out the possibility that mild immune modulation was present in some participants. The small numbers of participants in our study relative to some studies in high-income countries might also affect results. To conclude, the prevalence of MBL is broadly similar in rural Uganda and the UK, but substantial qualitative differences exist, with a lower prevalence of CLL-phenotype MBL and higher prevalence of CD5-negative MBL in the Ugandan cohort than in the UK cohort. These differences are likely to reflect variation in susceptibility alleles and antigenic exposure, which might affect the type of B-lymphoproliferative disorders that become clinically relevant. We believe that the study of MBL is an ideal platform to identify and study apparent differences in B-lymphoproliferative disorders in a way that is independent of health-care provision and referral biases, and to potentially allow validation of putative driver mutations using peripheral blood cells.33 Additional data are required for confirmation of our findings, and more detailed biological data from patients with symptomatic disease, particularly splenic marginal zone lymphomas, are also needed. Supplementary Material Supplementary appendix

To conclude, the prevalence of MBL is broadly similar in rural Uganda and the UK, but substantial qualitative differences exist, with a lower prevalence of CLL-phenotype MBL and higher prevalence of CD5-negative MBL in the Ugandan cohort than in the UK cohort. These differences are likely to reflect variation in susceptibility alleles and antigenic exposure, which might affect the type of B-lymphoproliferative disorders that become clinically relevant. We believe that the study of MBL is an ideal platform to identify and study apparent differences in B-lymphoproliferative disorders in a way that is independent of health-care provision and referral biases, and to potentially allow validation of putative driver mutations using peripheral blood cells.33 Additional data are required for confirmation of our findings, and more detailed biological data from patients with symptomatic disease, particularly splenic marginal zone lymphomas, are also needed. Supplementary Material Supplementary appendix Acknowledgments The Ugandan General Population Cohort study is jointly funded by the UK Medical Research Council (MRC) and the UK Department for International Development (DFID) under the MRC/DFID Concordat agreement. Further funding was from a Pump Priming Award from the Hull York Medical School (York, UK), the Wellcome Trust (090132), and the Ellis family. The genotype data from Uganda was obtained with support from MRC (grants G0901213-92157 and G0801566) and the Wellcome Trust Sanger Institute (WT098051). The immunogenetic analysis was partly supported by Horizon 2020 AEGLE, An analytics framework for integrated and personalised health-care services in Europe, from the European Commission. RN and TL are senior visiting scientists at the WHO International Agency for Research on Cancer, Lyon, France.

Institute (WT098051). The immunogenetic analysis was partly supported by Horizon 2020 AEGLE, An analytics framework for integrated and personalised health-care services in Europe, from the European Commission. RN and TL are senior visiting scientists at the WHO International Agency for Research on Cancer, Lyon, France. Contributors ACR and RN designed the study. ACR, AS, RdT, CD, DN, AV, PASE, KS, KW, AK, GA, and RN collected, analysed, and interpreted the data. All authors prepared the report and approved the final version. Declaration of interests We declare no competing interests.