Browse the corpus

Introduction Early recognition of major trauma enables emergency medical services (EMS) to accurately triage and transport injured patients to an appropriate hospital. Field triage, however, remains a challenge due to occult injuries, the unpredictable evolution of symptoms and complexities of evaluating patients in difficult circumstances. A combined literature review and US national expert panel consensus resulted in ‘Guidelines for Field Triage of Injured Patients’1, 2. This presented a stepwise evaluation of trauma victims for physiological instability, obvious anatomical injury, mechanism of injury and co-morbidity. The report recommended that tiered trauma care should be provided according to the probability of having sustained major trauma. Norway is sparsely populated with weather-dependent and time-consuming patient transport. Some 50 Norwegian hospitals receive patients with major injuries, most with low admission rates3. In an attempt to optimize patient outcome4, immediate resuscitation is increasingly being delivered via multidisciplinary one-tiered trauma teams. However, several studies indicate a trend for imprecise activation of such teams5–8.

orwegian hospitals receive patients with major injuries, most with low admission rates3. In an attempt to optimize patient outcome4, immediate resuscitation is increasingly being delivered via multidisciplinary one-tiered trauma teams. However, several studies indicate a trend for imprecise activation of such teams5–8. If patients with major injuries are deprived access to the possible benefits of immediate resuscitation and expert evaluation provided by a trauma team (undertriage), avoidable deaths may occur9. Conversely, if the trauma team attends patients with minor injuries (overtriage), scarce financial and human resources are consumed. To improve triage efficacy, a two-tiered trauma team activation (TTA) response has been recommended1. A full trauma team should attend patients suffering from obvious major injury, but a reduced trauma team may systematically evaluate patients where the extent of injury is unclear. A growing body of evidence suggests that a tiered response is safe and cost-effective10–21. The American College of Surgeons considers 5 per cent undertriage associated with 25–50 per cent overtriage as acceptable22. An unpublished registry-based analysis of the informal one-tiered TTA practice at Stavanger University Hospital (SUH) revealed unacceptably high undertriage and overtriage rates. For this reason, a two-tiered TTA protocol was developed and implemented at this trauma centre according to international recommendations1. The impact of this system revision on medical resource utilization and triage precision was evaluated using trauma registry data.

nacceptably high undertriage and overtriage rates. For this reason, a two-tiered TTA protocol was developed and implemented at this trauma centre according to international recommendations1. The impact of this system revision on medical resource utilization and triage precision was evaluated using trauma registry data. Methods SUH is a 630-bed primary trauma centre for a mixed rural/urban population of approximately 330 000 inhabitants and the trauma referral centre for an additional 120 000 people living in Rogaland county in southwestern Norway. The hospital admits each year approximately 140 adult and paediatric patients with a New Injury Severity Score23 (NISS) greater than 1524, 25. A hospital-based trauma registry has been fully operational since 2004. An Association for the Advancement of Automotive Medicine-certified Abbreviated Injury Scale (AIS) coder (a registered nurse) manually searches the hospital administrative data system for relevant patients (Table 1) and annually codes data on approximately 360 patients. Table 1 Inclusion and exclusion criteria for the Stavanger University Hospital trauma registry

Methods SUH is a 630-bed primary trauma centre for a mixed rural/urban population of approximately 330 000 inhabitants and the trauma referral centre for an additional 120 000 people living in Rogaland county in southwestern Norway. The hospital admits each year approximately 140 adult and paediatric patients with a New Injury Severity Score23 (NISS) greater than 1524, 25. A hospital-based trauma registry has been fully operational since 2004. An Association for the Advancement of Automotive Medicine-certified Abbreviated Injury Scale (AIS) coder (a registered nurse) manually searches the hospital administrative data system for relevant patients (Table 1) and annually codes data on approximately 360 patients. Table 1 Inclusion and exclusion criteria for the Stavanger University Hospital trauma registry Inclusion criteria Exclusion criteria Absolute criteria Patients not fulfilling the absolute Activated trauma team criteria Penetrating injury to or Head Isolated fracture with skin injury Neck (AIS 1) in Trunk  Upper extremity Extremities proximal to  Lower extremity knee or elbow  Floor of orbita Relative criteria Chronic subdural haematoma ISS ≥ 10 Drowning, inhalation injury, NISS > 15*  asphyxia-related injury (hanging, strangulation) Secondary admission to SUH > 24 h after injury * After implementing the Utstein template for uniform reporting of data following major trauma. AIS, Abbreviated Injury Scale; ISS, Injury Severity Score; NISS, New Injury Severity Score; SUH, Stavanger University Hospital.

asphyxia-related injury (hanging, strangulation) Secondary admission to SUH > 24 h after injury * After implementing the Utstein template for uniform reporting of data following major trauma. AIS, Abbreviated Injury Scale; ISS, Injury Severity Score; NISS, New Injury Severity Score; SUH, Stavanger University Hospital. Prehospital emergency care in the SUH catchment area is provided by on-call general practitioners, vehicle ambulance units staffed by paramedics and emergency medical technicians, and anaesthetist-manned rapid response cars and helicopters26. Until February 2009, the hospital practised informal activation of a one-tiered 13-personnel multidisciplinary trauma team.

t area is provided by on-call general practitioners, vehicle ambulance units staffed by paramedics and emergency medical technicians, and anaesthetist-manned rapid response cars and helicopters26. Until February 2009, the hospital practised informal activation of a one-tiered 13-personnel multidisciplinary trauma team. The Rogaland Trauma System Study Group was established by SUH in 2008 in cooperation with the Norwegian Air Ambulance Foundation research department. The group comprised clinical representatives from the emergency department, dispatch, surgery, anaesthesiology, and ground and air ambulance units in addition to researchers. They developed guidelines on field triage and TTA based on available evidence1, 5 and multidisciplinary consensus on optimal local practice. EMS providers were empowered to assign patients into two tiers of TTA according to field triage criteria (Table 2). Activation of the full multidisciplinary trauma team was based on physiological or anatomical criteria. The purpose of the full team was to provide immediate resuscitation and rapid evaluation, and initiation of definitive care. A reduced team was initiated in patients not meeting the criteria for the full team but when there was either one mechanism of injury or one co-morbidity criterion present (Table 3). The purpose of the reduced team was rapidly to assess physiologically stable patients for occult injuries. When two or more mechanisms of injury or co-morbidity criteria were fulfilled the full team was activated. The reduced team was capable of rapid upgrading to a full team if potentially severe injures were detected. Both full and reduced teams were led by the same surgeon with a minimum of 2 years of experience in surgery and certified as an Advanced Trauma Life Support provider. The remaining team members had no formal competence requirements. Additional surgical subspecialty resources were available at the team leader's discretion.

h full and reduced teams were led by the same surgeon with a minimum of 2 years of experience in surgery and certified as an Advanced Trauma Life Support provider. The remaining team members had no formal competence requirements. Additional surgical subspecialty resources were available at the team leader's discretion. Table 2 Triage criteria for tiered trauma team activation (full and reduced)

m revision to be a quality improvement initiative not in need of formal approval (2009/228-CAG). The Norwegian Social Science Data Services approved access to aggregate anonymous data on relevant patients in the hospital-based trauma registry (20 840 KS/LR). The study was registered in clinicaltrials.gov (NCT00876564). Statistical analysis Patients were classified as major trauma victims if they had an NISS above 1528. The evaluation of triage precision was based on the assumption that all patients with major injury benefit from assessment by a trauma team upon arrival at hospital. Sensitivity was defined as the probability for major trauma victims to be assessed by a full and/or reduced trauma team. Undertriage was defined as the contrary event (1–sensitivity), the probability of not being examined by a trauma team (full and/or reduced) despite having a major injury. To calculate specificity and thereby the conventional definition of overtriage (1—specificity)32, the number of patients with minor injuries admitted without an activated trauma team (true negatives; group d in Table 4) must be identified. As SUH annually treats a large number of patients (approximately 3400 subjects) with only minor injuries, the classical definition is of limited usefulness. This substantial and not easily definable group of patients is rarely considered in need of assessment by a trauma team, and would strongly bias a computation of overtriage based on specificity. Overtriage was therefore defined as the complement of the positive predictive value, 1 − PPV, where PPV represents the probability of a patient suffering from major trauma when the trauma team is activated (Table 4)33. This is equivalent to the proportion of patients without major trauma among those who were triaged to a trauma team.

h full and reduced teams were led by the same surgeon with a minimum of 2 years of experience in surgery and certified as an Advanced Trauma Life Support provider. The remaining team members had no formal competence requirements. Additional surgical subspecialty resources were available at the team leader's discretion. Table 2 Triage criteria for tiered trauma team activation (full and reduced) Full trauma team Reduced trauma team 1. Physiology 5. Co-morbidity 1·1 RTS ≤ 11 5·1 Age > 60 years 1·2 GCS < 14 5·2 Age < 6 years 1·3 Respiratory rate < 9/min 5·3 Severe co-morbidity (e.g. 1·4 Respiratory rate > 25/min   COPD, congestive heart 1·5 Spo2 < 90%   failure) 1·6 Intubated/attempted 5·4 Pregnancy intubation 5·5 Increased risk of haemorrhage 1·7 Obvious massive   (anticoagulant drugs, haemorrhage   coagulopathy) 1·8 Systolic blood pressure < 90 mmHg 6. Mechanism of injury 6·1 Co-passenger killed 2. Anatomy 6·2 Entrapped person 2·1 Facial injury with risk for 6·3 Person ejected from airway obstruction   vehicle/motorcycle 2·2 Flail chest 6·4 Pedestrian, cyclist run down 2·3 Suspected pneumothorax   at > 30 km/h or thrown up 2·4 Stab or gunshot wound   in the air proximal to knee or elbow 6·5 Collision speed > 50 km/h 2·5 Suspected pelvic fracture 6·6 Deformed vehicle 2·6 Crushed, mangled or   compartment amputated extremity 6·7 Airbag set off 2·7 Two or more long bone 6·8 Vehicle roll-over fractures 6·9 Fall > 5 m (adults) 2·8 Open fracture with 6·10 Fall > 3 m (children) ongoing haemorrhage 2·9 Open skull fracture or 7. Interhospital transfer impression fracture 7·1 Interhospital transfer and 2·10 Suspected spinal cord   < 24 h since time of injury injury 2·11 Burn injury (≥ grade II) Note: If two or more criteria under > 15% total body surface   list 5 or 6 are fulfilled, activate area   full trauma team 3. Several patients 3·1 Accident with several severely injured (suspected or confirmed) 4. Upgrade to full trauma team 4·1 When two or more criteria for reduced trauma team (list 5 or 6) are fulfilled 4·2 When reduced trauma team finds a perceived stable patient to be unstable RTS, Revised Trauma Score; GCS, Glasgow Coma Scale; COPD, chronic obstructive pulmonary disease; Spo2, oxygen saturation measured by pulse oximetry.

trauma team 4·1 When two or more criteria for reduced trauma team (list 5 or 6) are fulfilled 4·2 When reduced trauma team finds a perceived stable patient to be unstable RTS, Revised Trauma Score; GCS, Glasgow Coma Scale; COPD, chronic obstructive pulmonary disease; Spo2, oxygen saturation measured by pulse oximetry. Table 3 Trauma team composition (full and reduced) Full trauma team (13 members) Reduced trauma team (4 members) Team leader surgeon* Team leader surgeon* Orthopaedic surgeon† Orthopaedic surgeon† Theatre nurse 2 ED nurses 3 ED nurses Anaesthetist† Nurse anaesthetist Radiologist† 2 radiographers Laboratory technician Orderly * Minimum of 2 years' experience with surgery and certified Advanced Trauma Life Support provider. †No formal competence requirements. ED, emergency department. The trauma registry was upgraded to prospectively collect data necessary to compare practice after introduction of the two-tiered guidelines. The guidelines were launched on 3 February 2009 under the direction of the Rogaland Trauma System Study Group. Throughout the implementation period, instructors addressed specific aspects of the system revision during educational outreach visits. Information posters and periodical newsletters were used to increase understanding and awareness of the system revision.

under the direction of the Rogaland Trauma System Study Group. Throughout the implementation period, instructors addressed specific aspects of the system revision during educational outreach visits. Information posters and periodical newsletters were used to increase understanding and awareness of the system revision. The trial was designed as a prospective interventional study utilizing SUH trauma registry data and was divided into an analysis of the ‘before’ period, which consisted of patients subject to the informal one-tiered practice (1 January 2004 to 31 December 2008), and an analysis of the ‘after’ period, which consisted of patients subject to the two-tiered TTA protocol (1 July 2009 to 31 December 2010). The implementation period (1 January 2009 to 30 June 2009) was excluded from the analysis. Consecutive patients admitted to SUH during the study period who were registered in the SUH trauma registry and assigned one or more AIS codes were included if they had major trauma (NISS over 15) and/or had been triaged to meet the trauma team (Table 4, groups a, b and c). The AIS 1998 catalogue was used for all patients27. Interhospital transfers to SUH and patients admitted by non-healthcare personnel were excluded. Survival status 30 days after injury28 was obtained from patient records and the Norwegian Population Registry. The Standards for Quality Improvement Reporting (SQUIRE)29, Standards for Reporting of Diagnostic Accuracy (STARD) statement30 and Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were used31.

after injury28 was obtained from patient records and the Norwegian Population Registry. The Standards for Quality Improvement Reporting (SQUIRE)29, Standards for Reporting of Diagnostic Accuracy (STARD) statement30 and Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were used31. Table 4 Injury severity and trauma team activation Major trauma Not major trauma Total TTA a b a + b No TTA c d c + d Total a + c b + d n Sensitivity = a/(a + c); specificity = d/(b + d); positive predictive value (PPV) = a/(a + b); undertriage = 1 − sensitivity = c/(a + c); overtriage = 1 − PPV = b/(a + b). TTA, trauma team activation. The Regional Committee for Medical and Health Research Ethics deemed the system revision to be a quality improvement initiative not in need of formal approval (2009/228-CAG). The Norwegian Social Science Data Services approved access to aggregate anonymous data on relevant patients in the hospital-based trauma registry (20 840 KS/LR). The study was registered in clinicaltrials.gov (NCT00876564).

who require conversion to open operation may have had poorer postoperative outcomes in some series15,21,22. The present study found that 30-day postoperative mortality, length of hospital stay and 1-year survival was better in laparoscopically treated patients (irrespective of whether a procedure was converted or not). Patients undergoing laparoscopic surgery appeared to have a better prognosis than those receiving open surgery, although it is impossible to separate the effect of earlier disease in the laparoscopic group from any advantages arising directly from the approach. Nevertheless, appropriate selection for any surgical technique remains of paramount importance. It is apparent that some who undergo open surgery would simply not be suitable for a laparoscopic approach and it is to be expected that they have a worse prognosis. This study has highlighted some of the advantages arising from the implementation of a national programme of laparoscopic surgery for colorectal cancer. It does not conclude that laparoscopic surgery is superior to open surgery for all individuals, for which more detailed clinical information would be required.

therefore defined as the complement of the positive predictive value, 1 − PPV, where PPV represents the probability of a patient suffering from major trauma when the trauma team is activated (Table 4)33. This is equivalent to the proportion of patients without major trauma among those who were triaged to a trauma team. In addition to direct comparison of overtriage rates ‘before’ and ‘after’ system revision, skilled hours' expenditure on overtriage per major trauma victim was measured. For each member of the trauma team, 30 min per unnecessary activation was allocated (full trauma team, 13 members = 6·5 skilled hours; reduced trauma team, 4 members = 2 skilled hours; Table 3). Probability of survival was calculated using the Trauma Score—Injury Severity Score (TRISS) methodology34 with 1995 coefficients35. The W statistic36 (expressing excess survivors per 100 patients compared with TRISS model predictions) with 95 per cent confidence interval (c.i.) was used to compare outcomes from the two study periods33. Non-overlapping 95 per cent c.i. were considered to indicate significant differences in survival. Categorical variables were compared with Fisher's exact test, whereas continuous variables were analysed using the Mann–Whitney U test. Assumed predictors of overtriage and undertriage were tested in a multiple logistic regression analysis. All data were analysed using STATA/SE™ version 10.1 (StataCorp LP, College Station, Texas, USA) and StatView version 5.0.1 (SAS Institute, Cary, North Carolina, USA). Statistical significance was assumed for P < 0·050.

umed predictors of overtriage and undertriage were tested in a multiple logistic regression analysis. All data were analysed using STATA/SE™ version 10.1 (StataCorp LP, College Station, Texas, USA) and StatView version 5.0.1 (SAS Institute, Cary, North Carolina, USA). Statistical significance was assumed for P < 0·050. Results During the study period (1 January 2004 to 31 December 2010), 2327 patients were entered in the SUH trauma registry. Some 364 injured patients who were transferred to SUH from other hospitals, admitted by non-healthcare personnel or admitted during the new TTA criteria implementation period (1 January 2009 to 30 June 2009) were excluded. A further 151 patients who had neither sustained major trauma nor been triaged to a trauma team (true-negatives) were also excluded. In total, 1812 patients met the inclusion criteria and were enrolled in the study. There was a missing probability of survival for seven patients and lack of documentation of TTA criteria in 123, but otherwise data were complete. Table 5 shows population characteristics of included patients in the ‘before’ and ‘after’ study periods. Distribution of age and sex, proportion of accidents involving motor vehicles and the proportion of penetrating versus blunt injuries did not change significantly between the two study periods. Table 5 Patients included in the ‘before’ and ‘after’ study periods

Table 5 shows population characteristics of included patients in the ‘before’ and ‘after’ study periods. Distribution of age and sex, proportion of accidents involving motor vehicles and the proportion of penetrating versus blunt injuries did not change significantly between the two study periods. Table 5 Patients included in the ‘before’ and ‘after’ study periods Before After P† Included patients (TTA and/or major trauma) 1255 557 Age (years)* 31 (19–51) 34 (20–53) 0·280 Sex ratio (F:M) 354:901 155:402 0·910 Falls 273 (21·8) 164 (29·4) 0·001 Motor vehicle-related accidents 498 (39·7) 204 (36·6) 0·230 Dominant injury (penetrating:blunt) 58:1197 (4·8:95·2) 22:535 (3·9:96·1) 0·620 NISS* 12 (5–26) 8 (3–18) < 0·001 Major trauma 585 (46·6) 183 (32·9) < 0·001 Prehospital anaesthetist (yes:no) 737:518 (58·7:41·3) 271:286 (48·7:51·3) < 0·001 TTA 1089 (86·8) 522 (93·7) < 0·001 Deaths (unadjusted) 78 (6·2) 16 (2·9) 0·003 Values in parentheses are percentages unless otherwise stated; * values are median (interquartile range). TTA, trauma team activation; NISS, New Injury Severity Score; major trauma, NISS > 15. †Fisher's exact test for categorical variables; Mann–Whitney U test for continuous variables.

Before After P† Included patients (TTA and/or major trauma) 1255 557 Age (years)* 31 (19–51) 34 (20–53) 0·280 Sex ratio (F:M) 354:901 155:402 0·910 Falls 273 (21·8) 164 (29·4) 0·001 Motor vehicle-related accidents 498 (39·7) 204 (36·6) 0·230 Dominant injury (penetrating:blunt) 58:1197 (4·8:95·2) 22:535 (3·9:96·1) 0·620 NISS* 12 (5–26) 8 (3–18) < 0·001 Major trauma 585 (46·6) 183 (32·9) < 0·001 Prehospital anaesthetist (yes:no) 737:518 (58·7:41·3) 271:286 (48·7:51·3) < 0·001 TTA 1089 (86·8) 522 (93·7) < 0·001 Deaths (unadjusted) 78 (6·2) 16 (2·9) 0·003 Values in parentheses are percentages unless otherwise stated; * values are median (interquartile range). TTA, trauma team activation; NISS, New Injury Severity Score; major trauma, NISS > 15. †Fisher's exact test for categorical variables; Mann–Whitney U test for continuous variables. In the ‘after’ period, there was a significant increase in the proportion of traumas due to falls. The proportion of patients who met an anaesthetist before hospital decreased significantly and a higher proportion of the included patients had been triaged to receive a full or reduced trauma team. Median NISS score, proportion of patients with major trauma and number of deaths in ‘after’ patients were significantly lower. Triage categories of included patients are shown in Table 6. Among the 1255 patients included in the ‘before’ study period, 1089 (86·8 per cent) were triaged to a trauma team. In the ‘after’ study period, 522 of 557 patients (93·7 per cent) were triaged to a team, 232 to the full team and 290 to the reduced team.

In the ‘after’ period, there was a significant increase in the proportion of traumas due to falls. The proportion of patients who met an anaesthetist before hospital decreased significantly and a higher proportion of the included patients had been triaged to receive a full or reduced trauma team. Median NISS score, proportion of patients with major trauma and number of deaths in ‘after’ patients were significantly lower. Triage categories of included patients are shown in Table 6. Among the 1255 patients included in the ‘before’ study period, 1089 (86·8 per cent) were triaged to a trauma team. In the ‘after’ study period, 522 of 557 patients (93·7 per cent) were triaged to a team, 232 to the full team and 290 to the reduced team. Table 6 Triage categories and prehospital response types Before After TTA Not TTA TTA Not TTA Total (MT:not MT) Total (MT) Total (MT:not MT) Full team (MT:not MT) Reduced team (MT:not MT) Total (MT) All 419:670 166 148:374 108:124 40:250 35 Prehospital anaesthetist 338:364 35 99:165 80:73 19:92 7 No prehospital anaesthetist 81:306 131 49:209 28:51 21:158 28 TTA, trauma team activation; MT, major trauma (New Injury Severity Score > 15). Undertriage and overtriage In the ‘before’ period, 166 of the 585 patients with major trauma (28·4 per cent) were not triaged to a trauma team, and this fell to 35 of 183 (19·1 per cent) in the ‘after’ period (P < 0·001). There was a 41·2 per cent relative reduction in undertriage rate in responses without anaesthetists, whereas the decrease in the low rate of undertriage performed by prehospital anaesthetists was not significant.

not triaged to a trauma team, and this fell to 35 of 183 (19·1 per cent) in the ‘after’ period (P < 0·001). There was a 41·2 per cent relative reduction in undertriage rate in responses without anaesthetists, whereas the decrease in the low rate of undertriage performed by prehospital anaesthetists was not significant. The proportion of patients triaged to a trauma team who had not suffered major trauma increased from 670 of 1089 (61·5 per cent) in the ‘before’ study period to 374 of 522 (71·6 per cent) in the ‘after’ period (P < 0·001). The increase was most pronounced in prehospital responses with an anaesthetist, although responses without anaesthetists still had the highest rate (Table 7). Table 7 Changes in triage categories by prehospital response types Before (%) After (%) Absolute change (%) Relative change (%) P* Undertriage All 28·4 19·1 − 9·3 − 32·6 < 0·001 Prehospital anaesthetist 9·4 6·6 − 2·8 − 29·6 0·155 No prehospital anaesthetist 61·8 36·4 − 25·4 − 41·2 < 0·001 Overtriage, total All 61·5 71·6 10·1 16·5 < 0·001 Prehospital anaesthetist 51·9 62·5 10·6 20·5 0·001 No prehospital anaesthetist 79·1 81·0 1·9 2·5 < 0·001 Overtriage, full team All 53·4 Prehospital anaesthetist 47·7 No prehospital anaesthetist 64·6 Overtriage, reduced team All 86·2 Prehospital anaesthetist 82·9 No prehospital anaesthetist 88·3 * Fisher's exact test.

1 Prehospital anaesthetist 51·9 62·5 10·6 20·5 0·001 No prehospital anaesthetist 79·1 81·0 1·9 2·5 < 0·001 Overtriage, full team All 53·4 Prehospital anaesthetist 47·7 No prehospital anaesthetist 64·6 Overtriage, reduced team All 86·2 Prehospital anaesthetist 82·9 No prehospital anaesthetist 88·3 * Fisher's exact test. The proportion of patients who had not suffered major trauma was particularly high in patients assigned to receive reduced teams (250 of 290, 86·2 per cent) compared with 124 of 232 (53·4 per cent) in patients triaged to receive full teams (P < 0·001) (Table 7). The mean number of skilled hours spent per overtriaged patient was reduced from 6·5 to 3·5 (P < 0·001), whereas the number of skilled hours spent per major trauma victim was reduced from 7·4 to 7·1 (P < 0·001). After initially finding an association between age and mistriage (Fig. 1), age was included as an independent variable in the logistic regression models, along with sex, fall, motor vehicle-related accident, prehospital response type (with versus without anaesthetist) and study period (‘after’ versus ‘before’). Results are shown in Table 8. Fig. 1 Relationship between patient age and triage category. Box plots depict medians and interquartile ranges; whiskers represent 10th and 90th percentiles. Note non-overlapping 95 per cent confidence intervals for medians (notches) Table 8 Odds ratios for undertriage and overtriage in the logistic regression model

After initially finding an association between age and mistriage (Fig. 1), age was included as an independent variable in the logistic regression models, along with sex, fall, motor vehicle-related accident, prehospital response type (with versus without anaesthetist) and study period (‘after’ versus ‘before’). Results are shown in Table 8. Fig. 1 Relationship between patient age and triage category. Box plots depict medians and interquartile ranges; whiskers represent 10th and 90th percentiles. Note non-overlapping 95 per cent confidence intervals for medians (notches) Table 8 Odds ratios for undertriage and overtriage in the logistic regression model Odds ratio P Undertriage* Age (per decade) 1·28 (1·18, 1·39) < 0·001 Sex (F versus M) 1·26 (0·86, 1·87) 0·241 Fall (yes versus no) 2·46 (1·71, 3·55) < 0·001 Motor vehicle-related 0·09 (0·04, 0·18) < 0·001 accident (yes versus no) Prehospital anaesthetist 0·16 (0·11, 0·24) < 0·001 (yes versus no) Period (after versus before) 0·26 (0·17, 0·40) < 0·001 Overtriage* Age (per decade) 0·79 (0·75, 0·83) < 0·001 Sex (F versus M) 1·38 (1·10, 1·74) 0·006 Fall (yes versus no) 0·67 (0·52, 0·87) 0·003 Motor vehicle-related 2·07 (1·64, 2·62) < 0·001 accident (yes versus no) Prehospital anaesthetist 0·55 (0·45, 0·68) < 0·001 (yes versus no) Period (after versus before) 1·97 (1·57, 2·46) < 0·001 Values in parentheses are 95 per cent confidence intervals. * Overall model R2 for undertriage 0·101; for overtriage 0·291.

Odds ratio P Undertriage* Age (per decade) 1·28 (1·18, 1·39) < 0·001 Sex (F versus M) 1·26 (0·86, 1·87) 0·241 Fall (yes versus no) 2·46 (1·71, 3·55) < 0·001 Motor vehicle-related 0·09 (0·04, 0·18) < 0·001 accident (yes versus no) Prehospital anaesthetist 0·16 (0·11, 0·24) < 0·001 (yes versus no) Period (after versus before) 0·26 (0·17, 0·40) < 0·001 Overtriage* Age (per decade) 0·79 (0·75, 0·83) < 0·001 Sex (F versus M) 1·38 (1·10, 1·74) 0·006 Fall (yes versus no) 0·67 (0·52, 0·87) 0·003 Motor vehicle-related 2·07 (1·64, 2·62) < 0·001 accident (yes versus no) Prehospital anaesthetist 0·55 (0·45, 0·68) < 0·001 (yes versus no) Period (after versus before) 1·97 (1·57, 2·46) < 0·001 Values in parentheses are 95 per cent confidence intervals. * Overall model R2 for undertriage 0·101; for overtriage 0·291. All but one variable showed consistent and significant effects on triage. Increasing age clearly increased risk for undertriage and decreased risk for overtriage. For mechanisms of injury, falls showed increased risk for undertriage and decreased risk for overtriage, whereas motor vehicle-related accidents showed the opposite effects. Patients triaged by the emergency medical communication centre to a prehospital response involving an anaesthetist had reduced risk for both undertriage and overtriage. In the ‘after’ study period, risk for undertriage was reduced whereas risk for overtriage was increased. In this multiple logistic regression model, sex showed inconsistent effects on triage, possibly owing to a correlation between female sex, advanced age and trauma due to falls.

k for both undertriage and overtriage. In the ‘after’ study period, risk for undertriage was reduced whereas risk for overtriage was increased. In this multiple logistic regression model, sex showed inconsistent effects on triage, possibly owing to a correlation between female sex, advanced age and trauma due to falls. Analysis of individual TTA criteria in the ‘after’ study period for usage and overtriage showed that for reduced teams mechanism of injury criteria were associated with 89·4 per cent overtriage and co-morbidity criteria with 68 per cent overtriage (Table 9). Criteria were undocumented for 70 (24·5 per cent) of 286 reduced teams (79 per cent overtriage). For full teams, criteria pertaining to physiology were associated with 41 per cent overtriage, and criteria depicting anatomical injury with 59 per cent overtriage. Criteria were undocumented for 53 (23·1 per cent) of 229 full teams (62 per cent overtriage). Upgraded TTA due to the patient being unstable was applied to five patients of whom one had suffered minor injuries only (20 per cent overtriage). Four patients had falls and one was involved in a motor vehicle accident. Table 9 Trauma team activation criteria in the ‘after’ period: frequency and overtriage

Analysis of individual TTA criteria in the ‘after’ study period for usage and overtriage showed that for reduced teams mechanism of injury criteria were associated with 89·4 per cent overtriage and co-morbidity criteria with 68 per cent overtriage (Table 9). Criteria were undocumented for 70 (24·5 per cent) of 286 reduced teams (79 per cent overtriage). For full teams, criteria pertaining to physiology were associated with 41 per cent overtriage, and criteria depicting anatomical injury with 59 per cent overtriage. Criteria were undocumented for 53 (23·1 per cent) of 229 full teams (62 per cent overtriage). Upgraded TTA due to the patient being unstable was applied to five patients of whom one had suffered minor injuries only (20 per cent overtriage). Four patients had falls and one was involved in a motor vehicle accident. Table 9 Trauma team activation criteria in the ‘after’ period: frequency and overtriage n Overtriage Full team Physiology RTS ≤ 11 18 4 (22) GCS < 14 37 18 (49) Respiratory rate < 9/min 0 0 (0) Respiratory rate > 25/min 5 4 (80) Spo2 < 90% 0 0 (0) Intubated/attempted intubation 14 4 (29) Obvious massive haemorrhage 1 1 (100) Systolic blood pressure < 90 mmHg 0 0 (0) Physiology total 75 31 (41) Anatomy Facial injury with risk for airway obstruction 7 4 (57) Flail chest 2 1 (50) Suspected pneumothorax 21 9 (43) Stab or gunshot wound proximal to knee or elbow 10 7 (70) Suspected pelvic fracture 10 7 (70) Crushed, mangled or amputated extremity 2 1 (50) Two or more long bone fractures 4 1 (25) Open fracture with ongoing haemorrhage 0 0 (0) Open skull fracture or impression fracture 2 1 (50) Suspected spinal cord injury 14 11 (79) Burn injury > 15% total body surface area 2 2 (100) Anatomy total 74 44 (59) Other Several severely injured (suspected or 14 8 (57) confirmed) Two or more criteria for reduced trauma 8 6 (75) team are fulfilled Reduced team finds perceived stable 5 1 (20) patient unstable Other total 27 15 (56) Undocumented criteria 53 33 (62) Full team total 229 123 (53·7) Reduced team Co-morbidity Age > 60 years 9 7 (78) Age < 6 years 7 6 (86) Severe co-morbidity 8 4 (50) Pregnancy 0 0 (0) Increased risk for haemorrhage 4 2 (50) Co-morbidity total 28 19 (68) Mechanism of injury Co-passenger dead 1 1 (100) Entrapped person 4 3 (75) Ejected from vehicle/motorcycle 27 23 (85) Pedestrian, cyclist run down at > 30 km/h 33 28 (85) or thrown in the air Collision speed > 50 km/h 61 61 (100) Deformed vehicle compartment 8 8 (100) Airbag set off 14 14 (100) Vehicle roll-over 8 8 (100) Fall > 5 m (adults) 27 17 (63) Fall > 3 m (children) 5 5 (100) Mechanism of injury total 188 168 (89·4) Undocumented criteria 70 55 (79) Reduced team total 286 242 (84·6) Values in parentheses are percentages. RTS, Revised Trauma Score; GCS, Glasgow Coma Scale. Spo2, oxygen saturation measured by pulse oximetry.

(100) Fall > 5 m (adults) 27 17 (63) Fall > 3 m (children) 5 5 (100) Mechanism of injury total 188 168 (89·4) Undocumented criteria 70 55 (79) Reduced team total 286 242 (84·6) Values in parentheses are percentages. RTS, Revised Trauma Score; GCS, Glasgow Coma Scale. Spo2, oxygen saturation measured by pulse oximetry. Mortality No deaths were registered in patients triaged to reduced teams. Median time from activation of reduced team to full team upgrade for the five affected patients was 11 (range 0–21) min. Median NISS was 17 (range 6–50), and one upgraded patient died. There were 12 deaths among undertriaged patients, eight (4·8 per cent) in the ‘before’ and four (11 per cent) in the ‘after’ study period (P = 0·229). The median age of patients who died was 80 (range 66–90) years and median NISS 46 (range 27–59). All had falls. For the total population of included patients, the W statistic (excess survivors per 100 patients compared with TRISS model predictions) did not change significantly: 2·123 (95 per cent c.i. 1·070 to 3·177) ‘before’ versus 2·510 (1·127 to 3·892) ‘after’. Discussion The present study found that the introduction of a formalized TTA protocol with a two-tiered response was associated with reduced undertriage and increased overtriage. Trauma team resource consumption was significantly reduced. For the study period as a whole, increasing age and falls increased risk for undertriage and decreased risk for overtriage, whereas motor vehicle-related accidents showed the opposite effects.

e was associated with reduced undertriage and increased overtriage. Trauma team resource consumption was significantly reduced. For the study period as a whole, increasing age and falls increased risk for undertriage and decreased risk for overtriage, whereas motor vehicle-related accidents showed the opposite effects. Triage precision before implementation of the TTA protocol was poor. Informal activation of trauma teams did not correctly identify victims of major trauma. A relative reduction in overall undertriage of 32·6 per cent followed system revision. The current undertriage rate of 19·1 per cent is still considered unacceptable and continued efforts to further improve triage precision are essential. The death of one upgraded patient with an NISS of 50 emphasizes that the practice of upgrading a reduced team to a full team requires constant monitoring. There was a highly significant 41·2 per cent relative reduction in undertriage in prehospital responses without an anaesthestist but only a non-significant trend towards less undertriage when an anaesthetist was present. When studied in the logistic regression model, prehospital responses involving an anaesthetist had a higher overall triage precision with reduced risk for undertriage as well as overtriage. In the Norwegian prehospital system, anaesthetist-manned units normally attend patients considered severely injured by either dispatch or paramedic-manned units already at the scene, whereas paramedics respond to a considerably less preselected patient population. Direct comparison between the two EMS provider categories was therefore considered both unreasonable and counterproductive.

lly attend patients considered severely injured by either dispatch or paramedic-manned units already at the scene, whereas paramedics respond to a considerably less preselected patient population. Direct comparison between the two EMS provider categories was therefore considered both unreasonable and counterproductive. This undertriage rate in responses without an anaesthestist remains high, but is also seen in other organized trauma systems5, 10, 12. Initiatives such as increasing the number of employees with a certificate of competence in prehospital care have been launched to improve quality of care, but further studies on the reasons for undertriage are called for37. Triage precision should also be addressed in responses with an anaesthetist, although an undertriage rate of 5–10 per cent is considered acceptable22. All 12 patients who died in the undertriaged group were over 66 years old and had falls. The logistic regression model showed that increasing age and falls were both found to increase risk for undertriage and decrease risk for overtriage. Velmahos et al.38 have previously found that unintoxicated patients over 55 years of age with low-level falls had a high likelihood of significant injuries. Others have recommended that age over 69 years should be a criterion for TTA39 or a need for enhanced focus on apparently low-impact injuries in this population5.

Velmahos et al.38 have previously found that unintoxicated patients over 55 years of age with low-level falls had a high likelihood of significant injuries. Others have recommended that age over 69 years should be a criterion for TTA39 or a need for enhanced focus on apparently low-impact injuries in this population5. It was expected that a reduction in undertriage would be accompanied by increased overtriage. Although TTA is beneficial for trauma victims, it may lead to suboptimal care for other patients40. The two-tier TTA system was designed to reduce excess resource consumption due to overtriage. Skilled hours spent on overtriage per major trauma victim, reflecting the exploitation of manpower on minor trauma cases, were reduced from 7·4 to 7·1 after implementation of this system. This is of particular interest given the current focus on improvement of quality and cost reduction in healthcare. Much emphasis has been put on mechanism of injury as a criterion for TTA1, as it can contribute to the effectiveness of the triage tool in the absence of changes in vital signs or obvious anatomical injury41. Consequently, the findings that motor vehicle-related accidents were associated with both reduced risk for undertriage and increased risk for overtriage were expected. It was alarming, however, to find that falls carried an odds ratio for undertriage of 2·46. Educational efforts are obviously needed to reduce undertriage in this patient group.

s that motor vehicle-related accidents were associated with both reduced risk for undertriage and increased risk for overtriage were expected. It was alarming, however, to find that falls carried an odds ratio for undertriage of 2·46. Educational efforts are obviously needed to reduce undertriage in this patient group. The present study has a number of limitations. The ‘before’ study period involved a review of trauma registry data restricted to variables already defined in the trauma registry. Missing documentation of TTA criteria remained a challenge throughout the study period. A short 18-month ‘after’ period compared with a 60-month long ‘before’ period increases the risk for type II errors. The study is also susceptible to the Hawthorne effect42. The simultaneous introduction of revised TTA criteria and the two-tiered response also complicated the evaluation of the study outcome. Even though major trauma defines the threshold against which triage protocols are tested, several conflicting definitions exist43. An NISS of over 15 was used to define major trauma and adhere to the inclusion criteria recommended by the Utstein template for uniform reporting of trauma data28. This implies that undertriaged patients were those included in this group who were not met by a full or reduced trauma team. In contrast, Curtis et al.44 considered all patients with an ISS of more than 15 assessed by a trauma standby (similar to the SUH reduced team) to be undertriaged. The different definitions highlight the difficulties of comparing data. The way in which definitions of major trauma influence calculations of triage precision merit investigation.

l.44 considered all patients with an ISS of more than 15 assessed by a trauma standby (similar to the SUH reduced team) to be undertriaged. The different definitions highlight the difficulties of comparing data. The way in which definitions of major trauma influence calculations of triage precision merit investigation. Implementation of system revisions can be a challenging enterprise with over 250 barriers identified in the literature45. To improve implementation of the new TTA criteria a teaching programme was developed addressing specific aspects of system revision. The programme was included in hospital and prehospital educational outreach visits arranged by trained instructors, a periodical newsletter was published and information posters were designed to remind staff of the new system for tiered TTA. To reduce the impact of failures related to lack of experience with the protocol, all patients from the 6-month implementation phase were excluded. However, examples of misapplication of the triage protocol were found throughout the entire ‘after’ period and act as reminders that implementation is a continuous process. Converting from an informal one-tiered TTA to a formalized two-tiered TTA lowered the threshold for immediate access to high-quality trauma care by reducing undertriage rates. Although the introduction of a reduced trauma team increased the overtriage rate, the number of work hours spent per major trauma victim was reduced.

rting from an informal one-tiered TTA to a formalized two-tiered TTA lowered the threshold for immediate access to high-quality trauma care by reducing undertriage rates. Although the introduction of a reduced trauma team increased the overtriage rate, the number of work hours spent per major trauma victim was reduced. The Norwegian Air Ambulance Foundation funded the study. The funder had no involvement in study design, data collection, data analysis, manuscript preparation and publication decision. The authors had complete access to the study data that support the publication. We acknowledge and thank all the participating EMS providers and Stavanger University Hospital staff for their willingness to participate and support this project, and for their continued dedication to improve trauma care. The authors thank Signe Søvik, MD PhD, for her contributions to the statistical analyses and for invaluable comments on the manuscript. We also thank trauma coder Morten Hestnes, RN, for his valuable comments. Disclosure: The authors declare no conflict of interest. Collaborators The members of the Rogaland Trauma System Study Collaborating Group were Espen Fevang, Kjetil Søreide, Eldar Søreide, Johannes Lokøy, Pieter Oord, Carina Lavransdatter Fossåen, Pål Stokkeland and Kristian Strand.

Introduction The anatomical and physiological characteristics of haemorrhoids have not been elucidated fully. Microscopically, haemorrhoidal piles are sinusoids (vascular structures without a muscular wall)1. Direct arteriovenous communications have been demonstrated histologically and radiologically, and some authors have noted a resemblance to erectile tissue2. Traditionally, haemorrhoidal piles frequently appear to be localized to the left lateral, right posterolateral and right anterolateral sites in the anal canal circumference with the patient in the lithotomy position; however, this configuration is demonstrated in less than 20 per cent of patients3. In reality, a wider network of arterial and venous vessels has been described4, although the distribution and relationship to rectal and anal layers is unclear. Recently, haemorrhoidal disease (HD) has often been treated using non-excisional procedures. Some surgical techniques address the reduction of arterial inflow to haemorrhoids. Transanal haemorrhoidal dearterialization (THD) and Doppler-guided haemorrhoidal artery ligation (DG-HAL) are the main surgical procedures with this aim, and use specifically designed devices for arterial ligation in the lower rectum guided by a Doppler signal5. Stapled haemorrhoidopexy (SH) divides the haemorrhoidal arteries in the suture line6. Assessment of the optimal site for these surgical approaches should improve the clinical efficacy. The purpose of this study was to localize precisely the arteries running into the rectum and directed to haemorrhoids.

Transanal haemorrhoidal dearterialization (THD) and Doppler-guided haemorrhoidal artery ligation (DG-HAL) are the main surgical procedures with this aim, and use specifically designed devices for arterial ligation in the lower rectum guided by a Doppler signal5. Stapled haemorrhoidopexy (SH) divides the haemorrhoidal arteries in the suture line6. Assessment of the optimal site for these surgical approaches should improve the clinical efficacy. The purpose of this study was to localize precisely the arteries running into the rectum and directed to haemorrhoids. Methods The local institutional review board approved this study. Patients with HD were enrolled prospectively. Each patient signed an informed consent form regarding the procedures and purpose of the study. All patients had anal bleeding with or without haemorrhoidal prolapse. Before inclusion in the study, an accurate diagnostic assessment, including patient history, physical examination, anoproctoscopic and colonoscopic findings, if indicated, confirmed HD. Patients with chronic bowel inflammatory disease, anal fissures, anal fistulas or abscesses, and a history of pelvic surgery and/or radiotherapy were excluded.

tudy, an accurate diagnostic assessment, including patient history, physical examination, anoproctoscopic and colonoscopic findings, if indicated, confirmed HD. Patients with chronic bowel inflammatory disease, anal fissures, anal fistulas or abscesses, and a history of pelvic surgery and/or radiotherapy were excluded. The enrolled patients underwent endoanal–endorectal ultrasonography (ERUS) and colour duplex imaging performed by a single operator. An ultrasound system (Pro-Focus Green™; BK Medical, Herlev, Denmark) fitted with endoanal–endorectal probes (models 2052 and 8848; BK Medical) was used. Before ultrasound examinations, the patients were prepared with two enemas to flush the rectum. During ERUS, the proximal edge of the puborectalis sling was identified to localize the anorectal junction (ARJ). The ARJ was regarded as the best reference point during anorectal ultrasonography. The anal dentate line cannot be identified using ultrasound techniques, and in patients with HD the anal pecten can frequently be displaced. The lower rectal circumference was subdivided into six sectors (left anterolateral, left lateral, left posterolateral, right posterolateral, right lateral and right anterolateral) (Fig. 1). From the upper limit (6 cm above the ARJ), the same procedure was repeated every 1 cm until the lower limit was reached (1 cm above the ARJ) (Fig. 1). A total of 300 sectors were studied for each of the six rectal levels. Fig. 1 Topographic schematic diagram showing different levels of the colour duplex imaging examination

The enrolled patients underwent endoanal–endorectal ultrasonography (ERUS) and colour duplex imaging performed by a single operator. An ultrasound system (Pro-Focus Green™; BK Medical, Herlev, Denmark) fitted with endoanal–endorectal probes (models 2052 and 8848; BK Medical) was used. Before ultrasound examinations, the patients were prepared with two enemas to flush the rectum. During ERUS, the proximal edge of the puborectalis sling was identified to localize the anorectal junction (ARJ). The ARJ was regarded as the best reference point during anorectal ultrasonography. The anal dentate line cannot be identified using ultrasound techniques, and in patients with HD the anal pecten can frequently be displaced. The lower rectal circumference was subdivided into six sectors (left anterolateral, left lateral, left posterolateral, right posterolateral, right lateral and right anterolateral) (Fig. 1). From the upper limit (6 cm above the ARJ), the same procedure was repeated every 1 cm until the lower limit was reached (1 cm above the ARJ) (Fig. 1). A total of 300 sectors were studied for each of the six rectal levels. Fig. 1 Topographic schematic diagram showing different levels of the colour duplex imaging examination Using combined colour duplex imaging, the courses of arteries that reached haemorrhoidal piles were followed carefully. All perirectal arteries that were not directed to haemorrhoids (vaginal, prostatic, and seminal vesicle arteries) were excluded from the study.

Fig. 1 Topographic schematic diagram showing different levels of the colour duplex imaging examination Using combined colour duplex imaging, the courses of arteries that reached haemorrhoidal piles were followed carefully. All perirectal arteries that were not directed to haemorrhoids (vaginal, prostatic, and seminal vesicle arteries) were excluded from the study. Arteries were classified according to their location in the rectal wall: running within the submucosa, between the submucosa and the rectal muscle, within the rectal muscle, between the rectal muscle and the perirectal fat, or outside the rectal wall. The distance between the centre of the arterial lumen and the ultrasound probe surface (defined as ‘arterial depth’) was calculated. Close contact was maintained with the rectal mucosa, but care was taken to avoid applying excessive pressure to the rectal wall with the ultrasound probe to minimize any distortion of the ultrasonographic and Doppler signals owing to arterial occlusion or compression. For each sector investigated, at least one picture was obtained for review after the examination. Statistical analysis The mean(s.d.) value was calculated for each recorded parameter. One-way ANOVA was used to compare means. The Bonferroni method was used for multiple comparisons, when appropriate. P < 0·050 was considered statistically significant. Results Fifty patients (36 men, 14 women) with a mean(s.d.) age of 47·1(13·1) years were studied. Five patients (10 per cent) had grade II, 41 (82 per cent) had grade III and four (8 per cent) had grade IV haemorrhoids.

Statistical analysis The mean(s.d.) value was calculated for each recorded parameter. One-way ANOVA was used to compare means. The Bonferroni method was used for multiple comparisons, when appropriate. P < 0·050 was considered statistically significant. Results Fifty patients (36 men, 14 women) with a mean(s.d.) age of 47·1(13·1) years were studied. Five patients (10 per cent) had grade II, 41 (82 per cent) had grade III and four (8 per cent) had grade IV haemorrhoids. Significantly fewer sectors in the upper part of the low rectum had an arterial supply directed to the haemorrhoids than in the lower part (64·3, 66·0 and 66·0 per cent at 6, 5 and 4 cm above the ARJ respectively versus 98·3, 99·3 and 99·7 per cent at 3, 2 and 1 cm respectively; P < 0·001). Fig. 2 shows colour duplex imaging samples of different artery locations in relation to the rectal wall layers. The distribution of haemorrhoidal arteries in relation to rectal layers and distance from the ARJ is shown in Table 1. In the majority of the upper sectors (97·9 per cent at 6 cm and 90·9 per cent at 5 cm from the ARJ), haemorrhoidal arteries were located in the perirectal fat, and only occasionally within the bowel wall. At 4 cm above the ARJ, a greater number of sectors had arteries located in the rectal muscle. At 3 cm, the arteries were shown to run into the submucosa in the majority of sectors, whereas at 2 and 1 cm above the ARJ almost all of the arteries had a submucosal location (in 96·6 and 100 per cent of sectors respectively); the differences were statistically significant (P < 0·001).

es located in the rectal muscle. At 3 cm, the arteries were shown to run into the submucosa in the majority of sectors, whereas at 2 and 1 cm above the ARJ almost all of the arteries had a submucosal location (in 96·6 and 100 per cent of sectors respectively); the differences were statistically significant (P < 0·001). Fig. 2 Colour duplex imaging examples of different locations of arteries in relation to the rectal wall: a,b perirectal fat; c perirectal fat–rectal muscle; d,e rectal muscle–submucosa; e submucosa Table 1 Distribution of detectable haemorrhoidal arteries in relation to rectal sectors and wall layers No. of rectal sectors Distance from anorectal junction Haemorrhoidal artery location 6 cm 5 cm 4 cm 3 cm 2 cm 1 cm P* Perirectal fat 189 (97·9) 180 (90·9) 84 (42·4) 8 (2·7) 0 (0) 0 (0) < 0·001 Perirectal fat–rectal muscle 0 (0) 5 (2·5) 48 (24·2) 9 (3·1) 0 (0) 0 (0) < 0·001 Rectal muscle 4 (2·1) 9 (4·5) 34 (17·2) 23 (7·8) 5 (1·7) 0 (0) < 0·001 Rectal muscle–submucosa 0 (0) 3 (1·5) 27 (13·6) 57 (19·3) 5 (1·7) 0 (0) < 0·001 Submucosa 0 (0) 1 (0·5) 5 (2·5) 198 (67·1) 288 (96·6) 299 (100) < 0·001 Values in parentheses are percentages. * One-way ANOVA and Bonferroni tests.

Haemorrhoidal artery location 6 cm 5 cm 4 cm 3 cm 2 cm 1 cm P* Perirectal fat 189 (97·9) 180 (90·9) 84 (42·4) 8 (2·7) 0 (0) 0 (0) < 0·001 Perirectal fat–rectal muscle 0 (0) 5 (2·5) 48 (24·2) 9 (3·1) 0 (0) 0 (0) < 0·001 Rectal muscle 4 (2·1) 9 (4·5) 34 (17·2) 23 (7·8) 5 (1·7) 0 (0) < 0·001 Rectal muscle–submucosa 0 (0) 3 (1·5) 27 (13·6) 57 (19·3) 5 (1·7) 0 (0) < 0·001 Submucosa 0 (0) 1 (0·5) 5 (2·5) 198 (67·1) 288 (96·6) 299 (100) < 0·001 Values in parentheses are percentages. * One-way ANOVA and Bonferroni tests. No haemorrhoidal arteries were detected in the left and right anterolateral sectors at 6, 5 and 4 cm above the ARJ, whereas such arteries were identified in the other sectors. At the lower three levels (3, 2 and 1 cm above the ARJ), haemorrhoidal arteries were identified in nearly all circumferential sectors (Table 2). The mean haemorrhoidal arterial depth was significantly lower in more distal sectors than in more proximal sectors; a statistical comparison between each level showed all differences to be significant (P < 0·001) (Table 3). This was a consistent finding in each rectal sector (Table 4). When mean arterial depths at each rectal level were compared, the differences between sectors were not statistically different at 6 cm above (P = 0·674) or 1 cm below (P = 0·865) the ARJ, whereas differences between sectors were statistically significant at 5, 4, 3 and 2 cm above the ARJ (P = 0·022, P = 0·020, P < 0·001 and P = 0·005 respectively). Table 2 Rectal sectors with detectable haemorrhoidal arteries in relation to distance from anorectal junction

No haemorrhoidal arteries were detected in the left and right anterolateral sectors at 6, 5 and 4 cm above the ARJ, whereas such arteries were identified in the other sectors. At the lower three levels (3, 2 and 1 cm above the ARJ), haemorrhoidal arteries were identified in nearly all circumferential sectors (Table 2). The mean haemorrhoidal arterial depth was significantly lower in more distal sectors than in more proximal sectors; a statistical comparison between each level showed all differences to be significant (P < 0·001) (Table 3). This was a consistent finding in each rectal sector (Table 4). When mean arterial depths at each rectal level were compared, the differences between sectors were not statistically different at 6 cm above (P = 0·674) or 1 cm below (P = 0·865) the ARJ, whereas differences between sectors were statistically significant at 5, 4, 3 and 2 cm above the ARJ (P = 0·022, P = 0·020, P < 0·001 and P = 0·005 respectively). Table 2 Rectal sectors with detectable haemorrhoidal arteries in relation to distance from anorectal junction Distance from anorectal junction Rectal sector 6 cm 5 cm 4 cm 3 cm 2 cm 1 cm Left anterolateral 0 (0) 0 (0) 0 (0) 47 (94) 50 (100) 50 (100) Left lateral 47 (94) 49 (98) 49 (98) 50 (100) 50 (100) 50 (100) Left posterolateral 49 (98) 50 (100) 50 (100) 50 (100) 50 (100) 50 (100) Right posterolateral 48 (96) 49 (98) 49 (98) 49 (98) 49 (98) 49 (98) Right lateral 49 (98) 50 (100) 50 (100) 50 (100) 50 (100) 50 (100) Right anterolateral 0 (0) 0 (0) 0 (0) 49 (98) 49 (98) 50 (100) Values in parentheses are percentages.

50 (100) Left posterolateral 49 (98) 50 (100) 50 (100) 50 (100) 50 (100) 50 (100) Right posterolateral 48 (96) 49 (98) 49 (98) 49 (98) 49 (98) 49 (98) Right lateral 49 (98) 50 (100) 50 (100) 50 (100) 50 (100) 50 (100) Right anterolateral 0 (0) 0 (0) 0 (0) 49 (98) 49 (98) 50 (100) Values in parentheses are percentages. Table 3 Arterial depth in relation to level of rectal circumference Distance from anorectal junction 6 cm 5 cm 4 cm 3 cm 2 cm 1 cm P* Arterial depth (mm) 8·3(1·5) 6·6(1·4) 5·1(1·2) 3·3(0·9) 2·4(0·6) 1·9(0·4) < 0·001 Values are mean(s.d.). * One-way ANOVA and Bonferroni tests. Table 4 Arterial depth in relation to level of rectal circumference and sector Arterial depth (mm) Distance from anorectal junction Rectal sector 6 cm 5 cm 4 cm 3 cm 2 cm 1 cm P* Left anterolateral — — — 2·6(0·7) 2·2(0·6) 1·9(0·5) < 0·001 Left lateral 8·0(1·7) 6·1(1·4) 4·6(0·9) 3·4(0·7) 2·4(0·6) 1·9(0·4) < 0·001 Left posterolateral 8·5(1·4) 7·0(1·4) 5·3(1·3) 3·5(0·9) 2·4(0·6) 1·9(0·4) < 0·001 Right posterolateral 8·3(1·5) 6·7(1·5) 5·2(1·2) 3·6(0·8) 2·5(0·5) 2·0(0·4) < 0·001 Right lateral 8·3(1·5) 6·7(1·3) 5·2(1·2) 3·6(1·0) 2·6(0·8) 1·9(0·4) < 0·001 Right anterolateral — — — 2·9(0·8) 2·2(0·5) 1·8(0·4) < 0·001 P* 0·674 0·022 0·020 < 0·001 0·005 0·865 — Values are mean(s.d.). * One-way ANOVA and Bonferroni tests.

Rectal sector 6 cm 5 cm 4 cm 3 cm 2 cm 1 cm P* Left anterolateral — — — 2·6(0·7) 2·2(0·6) 1·9(0·5) < 0·001 Left lateral 8·0(1·7) 6·1(1·4) 4·6(0·9) 3·4(0·7) 2·4(0·6) 1·9(0·4) < 0·001 Left posterolateral 8·5(1·4) 7·0(1·4) 5·3(1·3) 3·5(0·9) 2·4(0·6) 1·9(0·4) < 0·001 Right posterolateral 8·3(1·5) 6·7(1·5) 5·2(1·2) 3·6(0·8) 2·5(0·5) 2·0(0·4) < 0·001 Right lateral 8·3(1·5) 6·7(1·3) 5·2(1·2) 3·6(1·0) 2·6(0·8) 1·9(0·4) < 0·001 Right anterolateral — — — 2·9(0·8) 2·2(0·5) 1·8(0·4) < 0·001 P* 0·674 0·022 0·020 < 0·001 0·005 0·865 — Values are mean(s.d.). * One-way ANOVA and Bonferroni tests. Discussion The pathogenesis of HD is unclear, but is probably multi-factorial. A number of elements have been claimed to be causative or predisposing factors. Disruption of supportive tissue surrounding haemorrhoids is considered to be an important factor in haemorrhoidal prolapse7 and a number of inflammatory mediators have also been cited8, 9. A hypertonic internal anal sphincter has frequently been associated with HD and is regarded as a possible cause of haemorrhoidal symptoms10. Haemorrhoidal vascularization appears to play a central role in the pathophysiology of HD.

haemorrhoidal prolapse7 and a number of inflammatory mediators have also been cited8, 9. A hypertonic internal anal sphincter has frequently been associated with HD and is regarded as a possible cause of haemorrhoidal symptoms10. Haemorrhoidal vascularization appears to play a central role in the pathophysiology of HD. Hyperplasia of the arteriovenous network within the anorectal submucosa (corpus cavernosum recti, CCR) results in increased vascular pressure. Blood overflow to the CCR should also cause increased intravascular pressure, and is thus a significant predisposing factor for HD11. Aigner and colleagues12 confirmed the relationship between arterial overflow and HD. Using a transperineal Doppler probe to investigate haemorrhoidal arteries, they found a significantly higher arterial calibre and flow velocity in patients with HD compared with controls. They then hypothesized that the coordinated filling and drainage of the anorectal vascular plexus is regulated by the intrinsic vascular sphincter mechanism, and that the morphological and functional failure of this vascular system may contribute to the development of HD13.

y in patients with HD compared with controls. They then hypothesized that the coordinated filling and drainage of the anorectal vascular plexus is regulated by the intrinsic vascular sphincter mechanism, and that the morphological and functional failure of this vascular system may contribute to the development of HD13. A comprehensive understanding of anorectal vascularization should contribute to outlining the pathophysiology of HD. A recent study by Schuurman and co-workers14 highlighted how vascularization of the CCR is provided almost exclusively by branches of the superior rectal artery (SRA), a terminal branch of the inferior mesenteric artery. A previous study by Shafik and Mostafa15 indicated that the lower half of the rectum is vascularized by the terminal branches of the SRA (two or three main branches), with plexiform patterns at the ends. The middle rectal artery has been reported in only 50 per cent of cadaver specimens15, and the functional role of this artery seems negligible in light of these anatomical inconsistencies. DiDio and colleagues16 also studied the middle rectal artery in 30 cadavers; it was present in 56·7 per cent of specimens, bilaterally (36·7 per cent) or unilaterally (20·0 per cent). The middle rectal artery arose from the internal pudendal artery in 40 per cent of specimens, the inferior gluteal artery in 26·7 per cent, and the internal iliac artery in 16·8 per cent. The consistent findings of the above studies appear to demonstrate that the SRA branches play a predominant role in CCR vascularization. Therefore, it is particularly important to define the topography of these vessels within the rectal–perirectal area. Aigner et al.17 analysed five macroscopic preparations of human pelvis; they described the division of the SRA into left and right branches, then into three to five terminal branches penetrating the rectal wall in the middle and lower rectum. On examining microscopic preparations from 27 fetuses, they identified two to four terminal vessels penetrating the rectal wall and reaching the submucosa, especially in the posterolateral position (71 per cent of specimens)17.

three to five terminal branches penetrating the rectal wall in the middle and lower rectum. On examining microscopic preparations from 27 fetuses, they identified two to four terminal vessels penetrating the rectal wall and reaching the submucosa, especially in the posterolateral position (71 per cent of specimens)17. In the present study, the majority of arterial branches at the three highest levels (6, 5 and 4 cm from the ARJ) were located outside the rectal wall in the right lateral, right posterolateral, left posterolateral and left lateral sectors, where these vessels arise. In contrast, no haemorrhoidal arteries were detected in several sectors in the higher three levels; in particular, none was found in the right and left anterolateral sectors. These findings suggest that the arterial pulses detected by Doppler ultrasonography in the anterior highest three levels of the low rectum during surgical procedures using this technology can be regarded as being generated by vessels that are not directed to haemorrhoids. In contrast, in the lower 2 cm, haemorrhoidal arteries were detected in 98 per cent of sectors; specifically, at 2 and 1 cm from the ARJ, arteries were identified in the submucosa in 96·6 per cent and 100 per cent of sectors respectively. These features can be confirmed easily during Doppler-guided surgical procedures. Investigation of the position of the arteries in relation to rectal layers and levels showed that the mean arterial depth decreased significantly from the highest to the lowest level, reaching the shallowest depth at the most distal 2 cm of the rectum where nearly all of the arteries were in the submucosa; this feature was invariably found regardless of the circumferential sector investigated.

nd levels showed that the mean arterial depth decreased significantly from the highest to the lowest level, reaching the shallowest depth at the most distal 2 cm of the rectum where nearly all of the arteries were in the submucosa; this feature was invariably found regardless of the circumferential sector investigated. Both anatomical and physiological evidence obtained from the literature and the present study has implications for the various therapeutic approaches that are currently available. In this regard, the most innovative surgical techniques are SH (also known as Longo's technique) and Doppler-guided ligation of haemorrhoidal arteries, including THD and DG-HAL techniques. The goal of the first method is to treat haemorrhoidal prolapse by resecting the rectal mucosa approximately 3–4 cm above the dentate line18; however, the level of anastomosis is frequently unpredictable as it is affected by the traction applied to the previously performed rectal purse-string. In fact, it has been established that, even though SH is performed according to well established technical guidelines, the intended location of the staple line is too difficult to standardize, as demonstrated by the wide range of anatomicopathological results reported by Ohana and co-workers19. Based on these data and the present findings, only a suture located within the distal 2 cm of the low rectum plays a role in the control of arterial overflow in patients treated with SH. However, even if SH and Doppler-guided ligation of haemorrhoidal arteries were applied, some doubts about Longo's procedure remain with respect to the circumferential suture performed with the specifically designed stapling device. This type of suture cannot ensure selective ligation of haemorrhoidal arteries as it can also involve both major and minor arterial vessels. Moreover, the circumferential suture could generate an unpredictable risk of venous outflow blockage, thus damaging the drainage system, as described previously13.

gnosis. This study has highlighted some of the advantages arising from the implementation of a national programme of laparoscopic surgery for colorectal cancer. It does not conclude that laparoscopic surgery is superior to open surgery for all individuals, for which more detailed clinical information would be required. This paper is a contribution from the National Cancer Intelligence Network (http://www.ncin.org.uk). It is based on information collected and quality assured by the regional cancer registries in England, specifically the Eastern Cancer Registration and Information Centre (Jem Rashbass), the Northern and Yorkshire Cancer Registry and Information Service (John Wilkinson and Brian Ferguson), the North West Cancer Intelligence Service (Tony Moran), the Oxford Cancer Intelligence Unit (Monica Roche), the South West Cancer Intelligence Service (Julia Verne), the Thames Cancer Registry (Elizabeth Davies), the Trent Cancer Registry (David Meechan) and the West Midlands Cancer Intelligence Unit (Gill Lawrence). While undertaking this work E.J.A.M was supported by the Cancer Research UK Bobby Moore Fund and P.Q. by Yorkshire Cancer Research. Disclosure: The authors declare no conflict of interest.

tapling device. This type of suture cannot ensure selective ligation of haemorrhoidal arteries as it can also involve both major and minor arterial vessels. Moreover, the circumferential suture could generate an unpredictable risk of venous outflow blockage, thus damaging the drainage system, as described previously13. The correlation between SH and rectal vascularization was highlighted in another study in which a perineal Doppler probe was used in patients who underwent SH for HD and a group of healthy subjects20. Baseline measurements differed significantly between patient and control groups. Postoperative follow-up showed no significant alterations in physiological parameters. Patients with a higher rate of recurrence of HD had higher baseline arterial flow velocity values. The study showed that SH did not reduce arterial inflow in the vessels feeding the anorectal vascular plexus. The present data may explain the reasons for the failure to reduce vascular overflow. In that study20, the anastomosis was performed 3·5–4 cm above the dentate line, a level at which most terminal arterial branches are not in the submucosa. Indeed, a meta-analysis of large-scale studies of patients undergoing SH demonstrated that these patients are more likely to develop recurrent HD with prolapse and bleeding at any time than those having conventional haemorrhoidectomy21, 22.

ine, a level at which most terminal arterial branches are not in the submucosa. Indeed, a meta-analysis of large-scale studies of patients undergoing SH demonstrated that these patients are more likely to develop recurrent HD with prolapse and bleeding at any time than those having conventional haemorrhoidectomy21, 22. The goal of THD and DG-HAL is significantly to reduce arterial overflow to haemorrhoidal piles by dearterialization, that is Doppler-guided ligation of the haemorrhoidal arteries in the upper part of the low rectum. The results of these operations seem promising23–31. In particular, most studies have shown that recurrent bleeding is limited to a minority of patients (5–20 per cent after THD; 1–21 per cent after DG-HAL)23–31. However, traditional dearterialization might fail to include the haemorrhoidal arteries in some sites owing to their deep location (within the muscularis propria or in perirectal fat), particularly on the anterior side of the rectum. The reported frequency of recurrent bleeding in patients undergoing dearterialization alone using the ‘high arterial ligation’ technique (31 per cent)26 supports this view. When mucopexy is included in THD or DG-HAL procedures, the possibility of excluding arteries may be lower as the running suture (even one that begins in the upper part of the low rectum to perform high ligation of haemorrhoidal arteries) is usually continued by transfixing the mucosa and submucosa to the ARJ, thus involving arterial branches directed to the haemorrhoidal piles.

ossibility of excluding arteries may be lower as the running suture (even one that begins in the upper part of the low rectum to perform high ligation of haemorrhoidal arteries) is usually continued by transfixing the mucosa and submucosa to the ARJ, thus involving arterial branches directed to the haemorrhoidal piles. By selectively ligating the haemorrhoidal arteries using a very precise Doppler system26, the THD technique can accurately identify the location of arterial vessels in the submucosa of the low rectum, thus achieving a significant reduction in arterial overflow to haemorrhoids. Based on the present findings, dearterialization should be more effective if performed 1 and 2 cm from the ARJ, where almost all of the arteries are localized in the submucosa, with a mean depth of 1·9–2·4 mm. In contrast, arterial ligation 3 cm from the ARJ may not be effective in certain sectors. At this level, 67·1 per cent of the vessels identified were located in the submucosa. Above this level, a smaller percentage of submucosal arteries was found, possibly making dearterialization less accurate at higher levels.

4 mm. In contrast, arterial ligation 3 cm from the ARJ may not be effective in certain sectors. At this level, 67·1 per cent of the vessels identified were located in the submucosa. Above this level, a smaller percentage of submucosal arteries was found, possibly making dearterialization less accurate at higher levels. This study provides new insight into the functional anatomy of haemorrhoids, with a direct impact on pathophysiology and treatment. The location of almost all the branches of the SRA in the submucosa provides a clear target for surgical treatments that recognize vascular overflow as a fundamental factor in the aetiology of HD. Optimizing dearterialization is essential for improving clinical outcome. In this regard, Doppler imaging plays a pivotal role during the surgical procedure by providing precise identification, then guided ligation of arteries. Clinical trials are required to confirm the therapeutic implications of these findings. The authors declare no conflict of interest.

Introduction Suspected appendicitis is the most common indication for surgery for non-obstetric conditions during pregnancy, and occurs in approximately one in 635 to one in 500 pregnancies per year1, 2. Appendicitis occurs more frequently in the second trimester than in the first or third trimester of pregnancy2–6. Abdominal surgery during pregnancy, particularly appendicectomy7, may increase the risk of poor pregnancy outcomes8. Fetal loss usually occurs in 3–15 per cent of women with complicated appendicectomy during the first trimester. However, the rate may be as high as 20–30 per cent9–11, with a premature delivery rate of 15–45 per cent12, and a significantly increased risk of spontaneous abortion, premature labour, and perinatal morbidity and mortality13. Miscarriage and infant mortality occur more frequently in women with perforated appendicitis14. However, the maternal mortality rate is very low15, 16 as a result of the use of advanced antibiotics, close perioperative monitoring, cooperation between specialties and improvements in perioperative management17, 18.

ty13. Miscarriage and infant mortality occur more frequently in women with perforated appendicitis14. However, the maternal mortality rate is very low15, 16 as a result of the use of advanced antibiotics, close perioperative monitoring, cooperation between specialties and improvements in perioperative management17, 18. Although guidelines for laparoscopic procedures during pregnancy have been established19, concern remains over the safety of the procedure, with reports of an increased risk of intra-abdominal abscess, particularly in perforated appendicitis. Assessment for open appendicectomy is related to gestational age as the appendix progressively relocates during pregnancy, typically from McBurney's point upwards from the iliac crest to near the gallbladder. Open appendicectomy is an established and safe operation with acceptable morbidity and low mortality rates.

is. Assessment for open appendicectomy is related to gestational age as the appendix progressively relocates during pregnancy, typically from McBurney's point upwards from the iliac crest to near the gallbladder. Open appendicectomy is an established and safe operation with acceptable morbidity and low mortality rates. Previous studies11, 13–15, 17, 18 were underpowered to detect any benefit of laparoscopic appendicectomy over the traditional open approach, resulting in conflicting results regarding the efficacy of laparoscopic appendicectomy for appendicitis during pregnancy. Only one randomized clinical trial comparing laparoscopic with open appendicectomy in pregnant women has been performed, with quality of life as the primary outcome20. One previous systematic review, including 28 observational studies that documented 637 laparoscopic appendicectomies, suggested that the laparoscopic procedure was associated with a higher rate of fetal loss but a similar or lower rate of preterm delivery compared with open appendicectomy21. However, the magnitude of treatment effects was not quantified. A systematic review and meta-analysis was therefore carried out, with the primary aim of estimating and comparing pregnancy outcomes including rates of fetal loss, preterm delivery, Apgar score and low birth weight.

ery compared with open appendicectomy21. However, the magnitude of treatment effects was not quantified. A systematic review and meta-analysis was therefore carried out, with the primary aim of estimating and comparing pregnancy outcomes including rates of fetal loss, preterm delivery, Apgar score and low birth weight. Methods Study selection Studies published between January 1990 and 11 July 2011 were identified from MEDLINE and Scopus databases using PubMed and Scopus search engines respectively. Search terms used were: pregnancy, pregnant women, laparoscopy, laparoscopic appendectomy, laparoscopic management, open appendectomy, conventional appendectomy, maternal outcome, premature labor pain, preterm labor, abortion, fetal loss, gestational age, fetal outcome, birth weight, Apgar score, surgical outcome, hospital stay, length of stay, hospitalization length, operative time, operation time, duration of operation, infection, wound infection, surgical infection and negative appendectomy. Search strategies are described in Table S1 (supporting information).

l age, fetal outcome, birth weight, Apgar score, surgical outcome, hospital stay, length of stay, hospitalization length, operative time, operation time, duration of operation, infection, wound infection, surgical infection and negative appendectomy. Search strategies are described in Table S1 (supporting information). Inclusion criteria Studies were included in the review if they met the following criteria: studied patients were pregnant women with suspected appendicitis; the intervention and comparator were laparoscopic and open appendicectomy respectively; at least one pregnancy (for example preterm delivery, fetal loss, birth weight, Apgar score) or surgical (duration of operation, length of hospital stay, wound infection, negative appendicectomy) outcome was reported; and the study was published in English. Studies were excluded if a hybrid procedure or single-trocar technique was used rather than the standard laparoscopic appendicectomy. The reference lists of all relevant studies were also reviewed. If studies were duplicates, the one with the most complete data was included. For studies that reported insufficient data, the corresponding authors were contacted and invited to provide more information. Two attempts were made to contact authors and if no response was obtained the study was excluded from the review.

. If studies were duplicates, the one with the most complete data was included. For studies that reported insufficient data, the corresponding authors were contacted and invited to provide more information. Two attempts were made to contact authors and if no response was obtained the study was excluded from the review. Outcomes The primary outcomes of interest were the pregnancy outcomes fetal loss and preterm delivery. Secondary outcomes were: birth weight, Apgar score and surgical outcomes, including duration of operation, length of hospital stay, wound infection and negative appendicectomy (appendicitis not proven pathologically). Data extraction Two investigators independently extracted the data from each study using a standard data extraction form. Information extracted included general data (author, year of publication, journal), study characteristics (study design, setting), patient characteristics (age, gestational age at surgery, gravida, body temperature, white blood cell count, number of subjects per group), and outcome as described above. Any disagreement was discussed and resolved by consensus with a third reviewer.

ublication, journal), study characteristics (study design, setting), patient characteristics (age, gestational age at surgery, gravida, body temperature, white blood cell count, number of subjects per group), and outcome as described above. Any disagreement was discussed and resolved by consensus with a third reviewer. Assessment of risk of bias The quality of studies was assessed independently by three reviewers on the basis of representativeness of studied subjects, information bias (ascertainment of outcome and interventions) and confounding bias (Table S2, supporting information). Each item was graded as having a low risk of bias, a high risk of bias, or an unclear risk if there was insufficient information to judge22. Any disagreement between the reviewers was discussed and resolved by consensus.

scertainment of outcome and interventions) and confounding bias (Table S2, supporting information). Each item was graded as having a low risk of bias, a high risk of bias, or an unclear risk if there was insufficient information to judge22. Any disagreement between the reviewers was discussed and resolved by consensus. Statistical analysis All analyses were performed using Stata version 11.1 (StataCorp LP, College Station, Texas, USA)23. For dichotomous outcomes (preterm delivery, fetal loss and wound infection), the relative risk (RR) of the outcome between laparoscopic versus open appendicectomy and its 95 per cent confidence interval (c.i.) were estimated for each study. If one cell in the 2 × 2 table contained zero, a continuity correction was performed by adding 0·5 to each cell24. Heterogeneity of RRs across studies was assessed using Cochran's Q test and the degree of heterogeneity was estimated using the I2 statistic. If the heterogeneity was significant or I2 exceeded 25 per cent, a random-effects model using the DerSimonian and Laird method was applied for pooling ORs; otherwise the inverse variance method was used25, 26. For continuous variables (duration of operation, length of hospital stay, Apgar score, birth weight) the unstandardized mean difference in outcomes between groups along with its 95 per cent c.i. was estimated and the values were pooled. Heterogeneity of the mean difference across studies was assessed as described above.

For continuous variables (duration of operation, length of hospital stay, Apgar score, birth weight) the unstandardized mean difference in outcomes between groups along with its 95 per cent c.i. was estimated and the values were pooled. Heterogeneity of the mean difference across studies was assessed as described above. Meta-regression analysis was used to assess the source of heterogeneity by fitting age and gestational age at surgery in the meta-regression model. A funnel plot with or without contour enhancement was used to detect publication bias owing to small study effects. The asymmetry of the funnel plot was assessed by means of Egger's test. The trim-and-fill method was used to impute missing studies if there was evidence of asymmetry of the funnel. P < 0·050 was considered statistically significant, except for the heterogeneity test, for which P < 0·100 was used. Results The initial literature search identified 88 and 196 studies from MEDLINE and Scopus databases respectively. Sixty-one studies were duplicates, leaving 223 for title or abstract review. After exclusion of 212 ineligible articles, 11 studies remained for analysis (Fig. 1). Agreement in data extraction between the two reviewers was 93·9 per cent (κ = 0·93, P < 0·001) and 92·2 per cent (κ = 0·0·92, P = 0 < 0·001) for dichotomous and continuous outcomes respectively. Fig 1 Identification of studies for inclusion in review

Results The initial literature search identified 88 and 196 studies from MEDLINE and Scopus databases respectively. Sixty-one studies were duplicates, leaving 223 for title or abstract review. After exclusion of 212 ineligible articles, 11 studies remained for analysis (Fig. 1). Agreement in data extraction between the two reviewers was 93·9 per cent (κ = 0·93, P < 0·001) and 92·2 per cent (κ = 0·0·92, P = 0 < 0·001) for dichotomous and continuous outcomes respectively. Fig 1 Identification of studies for inclusion in review The 11 included studies contained a total of 3415 patients (599 in laparoscopic and 2816 in open group) (Table 1)12, 27–36. Eight studies were comparative prospective cohort studies and three were comparative retrospective medical record reviews. Nine of the 11 studies were from the USA. The mean patient age ranged from 23·4 to 29·5 years. Gestational age at surgery was mostly in the second trimester, except in the study by Upadhyay and colleagues30. Four studies reported failure of laparoscopic appendicectomy, and the need to convert to open surgery in between one and three patients in each study27, 29, 33, 36. Two of these studies carried out intention-to-treat analysis, including the patient in the laparoscopic group27, 29. One study applied a per-protocol analysis33 and the other excluded patients whose operation was converted36. Table 1 Baseline characteristics of included studies

The 11 included studies contained a total of 3415 patients (599 in laparoscopic and 2816 in open group) (Table 1)12, 27–36. Eight studies were comparative prospective cohort studies and three were comparative retrospective medical record reviews. Nine of the 11 studies were from the USA. The mean patient age ranged from 23·4 to 29·5 years. Gestational age at surgery was mostly in the second trimester, except in the study by Upadhyay and colleagues30. Four studies reported failure of laparoscopic appendicectomy, and the need to convert to open surgery in between one and three patients in each study27, 29, 33, 36. Two of these studies carried out intention-to-treat analysis, including the patient in the laparoscopic group27, 29. One study applied a per-protocol analysis33 and the other excluded patients whose operation was converted36. Table 1 Baseline characteristics of included studies No. of women Reference Year No. of women Age (years)* Gestational age (weeks)* Negative appendicectomy (%) Laparoscopic Open Outcomes Corneille et al.27 2010 49 25·6(6·4) 15·9(8·4) NA 9 40 Fetal loss, hospital stay, preterm delivery Sadot et al.28 2010 57 29·5(5·9) 19·7(7·2) 24 41 16 Apgar score, birth weight, fetal loss, hospital stay, duration of operation, preterm delivery, wound infection Kirshtein et al.29 2009 42 28·4 13·9(6) 12 23 19 Apgar score, birth weight, fetal loss, hospital stay, duration of operation, wound infection McGory et al.12 2007 3133 NA NA 23·1 454 2679 Fetal loss Upadhyay et al.30 2007 6 27·2(3·3) 32(2·6) 17 4 2 Fetal loss, preterm delivery Carver et al.31 2005 28 23·4(5·8) 14(5·4) NA 17 11 Apgar score, birth weight, fetal loss, hospital stay, preterm delivery, wound infection Lyass et al.32 2001 22 28·5(15·2) 20(6·3) NA 11 11 Fetal loss, hospital stay, duration of operation, preterm delivery Affleck et al.33 1999 37 NA NA NA 19 18 Fetal loss, preterm delivery Conron et al.34 1999 21 NA NA NA 12† 9‡ Apgar score, birth weight, fetal loss, hospital stay, duration of operation Gurbuz and Peetz35 1997 9 24·5(1·5) 20·1(9) 22 5 4 Fetal loss, hospital stay, duration of operation, preterm delivery Curet et al.36 1996 11 NA NA NA 4 7 Fetal loss * Values are mean(s.d.).

eterm delivery Conron et al.34 1999 21 NA NA NA 12† 9‡ Apgar score, birth weight, fetal loss, hospital stay, duration of operation Gurbuz and Peetz35 1997 9 24·5(1·5) 20·1(9) 22 5 4 Fetal loss, hospital stay, duration of operation, preterm delivery Curet et al.36 1996 11 NA NA NA 4 7 Fetal loss * Values are mean(s.d.). † Includes laparoscopic cholecystectomy; ‡ includes open cholecystectomy. NA, not available. The negative appendicectomy rate ranged from 12 to 24 per cent. Fetal loss was reported in all 11 studies12, 27–36 and preterm labour in seven27, 28, 30–33, 35. Four studies reported on Apgar scores28, 29, 31, 34, but only three had sufficient data to pool28, 31, 34, and four reported birth weight28, 29, 31, 34. Risk of bias Assessment of risk of bias is reported in Table 2. The agreement between two reviewers was 95·5 per cent with a κ statistic of 0·94 (P < 0·001). Among 11 studies, the risk of selection bias from the use of non-representative cases was low in seven and unclear in four studies. The ascertainment of all outcomes was clearly described (except for wound infection) in six studies. Ascertainment of surgical technique was clear in seven studies. Unclear ascertainment in four studies was due to conversion from laparoscopic appendicectomy to an open technique. Confounding bias was likely to be present in ten studies. Table 2 Quality assessment of included studies

Risk of bias Assessment of risk of bias is reported in Table 2. The agreement between two reviewers was 95·5 per cent with a κ statistic of 0·94 (P < 0·001). Among 11 studies, the risk of selection bias from the use of non-representative cases was low in seven and unclear in four studies. The ascertainment of all outcomes was clearly described (except for wound infection) in six studies. Ascertainment of surgical technique was clear in seven studies. Unclear ascertainment in four studies was due to conversion from laparoscopic appendicectomy to an open technique. Confounding bias was likely to be present in ten studies. Table 2 Quality assessment of included studies Reference Representativeness of cohorts Ascertainment of outcome Ascertainment of intervention Confounding bias Note Corneille et al.27 Low risk Low risk* High risk High risk 2 operations in LA group converted to OA Sadot et al.28 Low risk Low risk* Low risk High risk Kirshtein et al.29 Low risk Low risk High risk High risk 1 operation in LA group converted to OA McGory et al.12 Low risk Low risk* Low risk Low risk Applied logistic regression, adjusted for age and race Upadhyay et al.30 Unclear Low risk* Low risk High risk Carver et al.31 Low risk Unclear Low risk High risk Lyass et al.32 Low risk Unclear Low risk High risk Affleck et al.33 Low risk Low risk* High risk High risk 2 operations in LA group converted to OA Conron et al.34 Unclear Unclear Low risk High risk Gurbuz and Peetz35 Unclear Unclear Low risk High risk Curet et al.36 Unclear Unclear High risk High risk 3 operations in LA group converted to OA * Except wound infection. LA, laparosopic appendicectomy; OA, open appendicectomy.

2 operations in LA group converted to OA Conron et al.34 Unclear Unclear Low risk High risk Gurbuz and Peetz35 Unclear Unclear Low risk High risk Curet et al.36 Unclear Unclear High risk High risk 3 operations in LA group converted to OA * Except wound infection. LA, laparosopic appendicectomy; OA, open appendicectomy. Fetal loss All 11 studies (3415 women) reported fetal loss after appendicectomy12, 27–36, which allowed quantitative pooled analysis. The RRs were homogeneous (χ2 = 2·44, 10 d.f., P = 0·992; I2 = 0 per cent) with a pooled value (laparoscopic versus open appendicectomy) of 1·91 (95 per cent c.i. 1·31 to 2·77) (Table 3, Fig. 2a). This suggested that the odds of fetal loss was almost twice as high in the laparoscopy group as in the open appendicectomy group. Fig 2 Meta-analysis of pregnancy outcomes a fetal loss and b preterm labour after laparoscopic (LA) versus open (OA) appendicectomy. Relative risks are shown with 95 per cent confidence intervals Table 3 Comparisons of fetal loss and preterm labour between laparoscopic and open appendicectomy in pregnancy

Fetal loss All 11 studies (3415 women) reported fetal loss after appendicectomy12, 27–36, which allowed quantitative pooled analysis. The RRs were homogeneous (χ2 = 2·44, 10 d.f., P = 0·992; I2 = 0 per cent) with a pooled value (laparoscopic versus open appendicectomy) of 1·91 (95 per cent c.i. 1·31 to 2·77) (Table 3, Fig. 2a). This suggested that the odds of fetal loss was almost twice as high in the laparoscopy group as in the open appendicectomy group. Fig 2 Meta-analysis of pregnancy outcomes a fetal loss and b preterm labour after laparoscopic (LA) versus open (OA) appendicectomy. Relative risks are shown with 95 per cent confidence intervals Table 3 Comparisons of fetal loss and preterm labour between laparoscopic and open appendicectomy in pregnancy Laparoscopic Open Reference Yes No Yes No Relative risk Fetal loss Corneille et al.27 0 9 3 37 0·59 (0·03, 10·45) Sadot et al.28 1 40 0 16 1·21 (0·05, 28·35) Kirshtein et al.29 1 22 1 18 0·83 (0·06, 12·35) McGory et al.12 31 423 88 2591 2·08 (1·40, 3·09) Upadhyay et al.30 0 4 0 2 0·60 (0·02, 23·07) Carver et al.31 2 15 0 11 3·33 (0·17, 63·51) Lyass et al.32 0 11 0 11 1·00 (0·02, 46·40) Affleck et al.33 0 19 0 18 0·95 (0·02, 45·51) Conron et al.34 0 12 0 9 0·77 (0·02, 35·51) Gurbuz and Peetz35 0 5 0 4 0·83 (0·02, 34·94) Curet et al.36 0 4 0 7 1·60 (0·04, 68·53) Pooled relative risk 1·91 (1·31, 2·77) Preterm labour Corneille et al.27 1 8 5 35 0·89 (0·12, 6·71) Sadot et al.28 12 29 3 13 1·56 (0·51, 4·81) Upadhyay et al.30 1 3 0 2 1·80 (0·10, 31·52) Carver et al.31 2 15 0 11 3·33 (0·17, 63·51) Lyass et al.32 0 11 0 11 1·00 (0·02, 46·40) Affleck et al.33 3 16 2 16 1·42 (0·27, 7·54) Gurbuz and Peetz35 0 5 0 4 0·83 (0·02, 34·94) Pooled relative risk 1·44 (0·68, 3·06) Values in parentheses are 95 per cent confidence intervals.

dhyay et al.30 1 3 0 2 1·80 (0·10, 31·52) Carver et al.31 2 15 0 11 3·33 (0·17, 63·51) Lyass et al.32 0 11 0 11 1·00 (0·02, 46·40) Affleck et al.33 3 16 2 16 1·42 (0·27, 7·54) Gurbuz and Peetz35 0 5 0 4 0·83 (0·02, 34·94) Pooled relative risk 1·44 (0·68, 3·06) Values in parentheses are 95 per cent confidence intervals. Egger's test suggested asymmetry of the funnel (coefficient − 0·47, s.e. 0·13, P = 0·005). A contour-enhanced funnel plot was therefore created (Fig. 3a); this showed that all studies were in the non-significant area except that by McGory and colleagues12 in which laparoscopic appendicectomy had a significantly higher risk. Despite this asymmetry, application of a non-parametric trim-and-fill method could not identify any missing study. Fig 3 Contour-enhanced funnel plots for studies comparing a fetal loss and b preterm labour after laparoscopic versus open appendicectomy Preterm delivery Among seven studies (208 women) that reported preterm labour27, 28, 30–33, 35, the RRs were homogeneous across studies (χ2 = 0·69, 6 d.f., P = 1·000; I2 = 0 per cent) (Table 3). The pooled RR was 1·44 (0·68 to 3·06) (Fig. 2b), indicating that the odds of preterm labour was 44 per cent higher in the laparoscopy than the open appendicectomy group; however, this was not statistically significant. Egger's test did not suggest publication bias (coefficient − 0·89, s.e. 0·34, P = 0·802) and this was supported by a symmetrical contour-enhanced funnel plot (Fig. 3b).

e odds of preterm labour was 44 per cent higher in the laparoscopy than the open appendicectomy group; however, this was not statistically significant. Egger's test did not suggest publication bias (coefficient − 0·89, s.e. 0·34, P = 0·802) and this was supported by a symmetrical contour-enhanced funnel plot (Fig. 3b). Other pregnancy outcomes Among four studies that reported birth weight (n = 148)28, 29, 31, 34, there was no heterogeneity (χ2 = 0·66, 3 d.f., P = 0·882; I2 = 0 per cent). The unstandardized pooled mean difference was 0·06 (95 per cent c.i. − 0·05 to 0·16) kg, suggesting that birth weights were similar in the two groups (Table 4). Apgar scores in the laparoscopic and open appendicectomy groups were compared in three studies (n = 106)28, 31, 34. The data were heterogeneous (χ2 = 9·70, 2 d.f., P = 0·008; I2 = 78·6 per cent), with an unstandardized mean difference of 0·05 (−0·18 to 0·27), indicating no significant different in Apgar scores between groups (Table 4). Table 4 Comparison of secondary outcomes between laparoscopic and open appendicectomy No. of women No. of included studies Laparoscopic Open I2 (%) Pooled effect* Birth weight (kg) 4 93 55 0 0·06 (−0·05, 0·16) Apgar score 3 70 36 78·6 0·05 (−0·18, 0·27) Wound infection 3 81 46 0 0·91 (0·12, 7·18) Duration of operation (min) 5 92 59 59·5 5·88 (−1·58, 13·33) Hospital stay (days) 7 118 110 90·9 − 0·49 (−1·76, − 0·78) Values in parentheses are 95 per cent confidence intervals. * Pooled relative risk for wound infection and pooled mean difference for other outcomes.

No. of women No. of included studies Laparoscopic Open I2 (%) Pooled effect* Birth weight (kg) 4 93 55 0 0·06 (−0·05, 0·16) Apgar score 3 70 36 78·6 0·05 (−0·18, 0·27) Wound infection 3 81 46 0 0·91 (0·12, 7·18) Duration of operation (min) 5 92 59 59·5 5·88 (−1·58, 13·33) Hospital stay (days) 7 118 110 90·9 − 0·49 (−1·76, − 0·78) Values in parentheses are 95 per cent confidence intervals. * Pooled relative risk for wound infection and pooled mean difference for other outcomes. Surgical outcomes Wound infection, duration of operation and hospital stay were also pooled across studies (Table 4). Pooling wound infection in three studies (n = 127)28, 29, 31 yielded a pooled RR of 0·91 (0·12 to 7·18), suggesting little difference in the risk of wound infection between interventions. The duration of operation was longer in the laparoscopy group, by a mean of 5·88 (−1·58 to 13·33) min, but the difference was not significant. The length of hospital stay was significantly shorter in the laparoscopy group by almost half a day (95 per cent c.i. − 1·76 to − 0·78 days). Discussion The results of this systematic review and meta-analysis suggest that laparoscopic appendicectomy in pregnancy results in an almost twofold significantly higher risk of fetal loss compared with open appendicectomy. No significant differences were observed between groups in preterm delivery, birth weight, Apgar score, wound infection after surgery or duration of operation.

is suggest that laparoscopic appendicectomy in pregnancy results in an almost twofold significantly higher risk of fetal loss compared with open appendicectomy. No significant differences were observed between groups in preterm delivery, birth weight, Apgar score, wound infection after surgery or duration of operation. The higher risk of fetal loss after laparoscopic compared with open appendicectomy needs to be addressed in the era of laparoscopic surgery, and has been discussed in many reports of the relative safety of laparoscopy in pregnancy10, 13, 37. However, this finding was largely dominated by the study of McGory and colleagues12, which had largest sample size and greatest power in detection of an association. After exclusion of this study from the pooled analysis, there was no effect of laparoscopic appendicectomy on fetal loss. The major consideration in laparoscopic appendicectomy in pregnancy is the effect of increased intra-abdominal pressure and fetal acidosis during carbon dioxide pneumoperitoneum. Increasing abdominal pressure from the pneumoperitoneum can lead to decreased venous return, especially in women with impaired cardiac output38, and result in maternal hypotension and hypoxia39. In addition, it has been reported that carbon dioxide is also absorbed across the peritoneum, which leads to fetal acidosis40. However, this is in contrast the findings of another study that reported no substantial adverse effect on the fetus when the maximum pneumoperitoneal pressure was as high as 10–12 mmHg and the duration less than 30 min41.

orted that carbon dioxide is also absorbed across the peritoneum, which leads to fetal acidosis40. However, this is in contrast the findings of another study that reported no substantial adverse effect on the fetus when the maximum pneumoperitoneal pressure was as high as 10–12 mmHg and the duration less than 30 min41. Although not statistically significant, the present results suggest that there may be an increased risk of preterm delivery in those undergoing laparoscopic appendicectomy compared with open appendicectomy. It is likely that this analysis did not have sufficient statistical power to detect a significant difference, given that a sample size of 749 would be required in each group to detect a RR of 1·44. Although the mean operating time was 5·88 min longer in the laparoscopic group, this was not statistically significant. The length of hospital stay was approximately half a day shorter after laparoscopic compared with the open appendicectomy, but this result depends heavily on one outlier study and cannot be considered robust. This requires further investigation for health service use planning, but a shorter hospital stay after a laparoscopic appendicectomy might not be advantageous clinically because of the need to monitor the patient for the adverse events noted above. This meta-analysis quantified the effects of laparoscopic and open appendicectomy on pregnancy and surgical outcomes. A previous review did not pool data and most included studies were non-comparative, with only one group21. The present review included the most relevant pregnancy and surgical outcomes.

Although the mean operating time was 5·88 min longer in the laparoscopic group, this was not statistically significant. The length of hospital stay was approximately half a day shorter after laparoscopic compared with the open appendicectomy, but this result depends heavily on one outlier study and cannot be considered robust. This requires further investigation for health service use planning, but a shorter hospital stay after a laparoscopic appendicectomy might not be advantageous clinically because of the need to monitor the patient for the adverse events noted above. This meta-analysis quantified the effects of laparoscopic and open appendicectomy on pregnancy and surgical outcomes. A previous review did not pool data and most included studies were non-comparative, with only one group21. The present review included the most relevant pregnancy and surgical outcomes. One major limitation is that all studies included in the pooled analysis were observational, and summary data published within each article were included in the review. Many other factors (such as patient age, duration of pregnancy, weight gain, complicated appendicitis, surgeon's skill, clinical setting) may affect the outcomes following appendicectomy, and confounding bias cannot be ruled out as the studies were not randomized. To adjust for confounding bias, individual patient data would be required from each study. There were no available data on complicated appendicitis (perforated and gangrenous) and it was not possible to assess whether the effects of laparoscopic appendicectomy on fetal loss were confounded by complicated appendicitis. Agreement between the present results and a meta-analysis of randomized trials or a subsequent large-scale trial is needed to confirm the present findings42. Given that pregnant women are subject to human-subject protection in clinical studies, it will be difficult to conduct a randomized trial. However, the authors believe that the direction of bias is probably conservative: those with more co-morbidity and who are considered high risk are likely to undergo open appendicectomy, making the laparoscopic approach look spuriously superior. The increased risk of fetal loss seen here is therefore likely to be an underestimate. It was not possible to identify a statistically significant difference for preterm delivery and infection owing to the limited number of studies available for pooling. Finally, as the severity of appendicitis was not reported consistently in the pooled studies, a subgroup analysis to identify specific subgroups of women who might benefit from, or be harmed by, laparoscopic appendicectomy was not possible.

ery and infection owing to the limited number of studies available for pooling. Finally, as the severity of appendicitis was not reported consistently in the pooled studies, a subgroup analysis to identify specific subgroups of women who might benefit from, or be harmed by, laparoscopic appendicectomy was not possible. Disclosure The authors declare no conflict of interest. Supporting information Additional supporting information may be found in the online version of this article: Table S1 Search strategy for PubMed and Scopus (Word document) Table S2 Assessment of risk of bias (Word document) Please note: John Wiley & Sons Ltd is not responsible for the functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

Introduction There have been extensive studies of laparoscopic resection for colonic cancer, including randomized clinical trials showing short-term advantages for a minimally invasive approach. Some have also studied health-related quality of life (HRQL), reporting the superiority of laparoscopic surgery1,2. Laparoscopic surgery for rectal cancer has been studied less extensively. The impact of a permanent stoma on HRQL has been described, as has the change of HRQL over time in patients treated for rectal cancer3–5. In a prospective comparison of the effects of laparoscopic versus open surgery, Li and colleagues6 found improved HRQL 1 week after laparoscopic surgery, but not after 1 year. The present study compared HRQL 1 year after laparoscopic or open surgery for rectal cancer in a subset of patients from the international multicentre randomized clinical trial COlorectal cancer Laparoscopic or Open Resection (COLOR) II7.

ues6 found improved HRQL 1 week after laparoscopic surgery, but not after 1 year. The present study compared HRQL 1 year after laparoscopic or open surgery for rectal cancer in a subset of patients from the international multicentre randomized clinical trial COlorectal cancer Laparoscopic or Open Resection (COLOR) II7. Methods The COLOR II trial The patients in this HRQL study constituted a subset of the COLOR II trial cohort7. Thirty hospitals in eight countries (Belgium, Canada, Denmark, Germany, the Netherlands, South Korea, Spain and Sweden) participated in COLOR II, but inclusion in the HRQL study was optional. The primary endpoint of the trial is local recurrence rate, and the trial was designed as a non-inferiority study. Patients were randomized between laparoscopic and open surgery in the ratio 2 : 1, and the trial was stratified according to centre, preoperative radiation and type of operation. The inclusion criteria focused on selection of patients undergoing elective surgery for potentially curable rectal cancer, T1–T3, best treated by partial mesorectal excision, total mesorectal excision or abdominoperineal resection. Exclusion criteria included transanal resection. The protocol of the COLOR II trial was approved by the appropriate ethics committees7, and registered at http://ClinicalTrials.gov (NCT0029779).

le rectal cancer, T1–T3, best treated by partial mesorectal excision, total mesorectal excision or abdominoperineal resection. Exclusion criteria included transanal resection. The protocol of the COLOR II trial was approved by the appropriate ethics committees7, and registered at http://ClinicalTrials.gov (NCT0029779). Patients Twelve hospitals in five countries (Canada, Denmark, Germany, the Netherlands and Sweden) participated in the HRQL component of the COLOR II trial. Inability to understand the questionnaires was an exclusion criterion. Patients who agreed to participate were asked to complete the preoperative questionnaire within 5 days before the operation, then 4 weeks, 6, 12 and 24 months after surgery. In Dutch hospitals, patients were also asked to complete EuroQol – 5D (EQ-5D™; EuroQol Group, Rotterdam, The Netherlands) questionnaires 3, 7 and 14 days after operation. The results at 24 months will be published separately. Demographic details, data on complications, tumour stage as classified in the pathology report on the resected specimen, reoperations, postoperative adjuvant chemotherapy, sexual function, and urinary and faecal continence, as recorded in clinical record forms at follow-up outpatient visits, were retrieved from the COLOR II database in Halifax, Canada. An analysis of sexual and urinary function will be presented separately, including European Organization for Research and Treatment of Cancer (EORTC) QLQ-PR25 together with data from clinical follow-up.

n clinical record forms at follow-up outpatient visits, were retrieved from the COLOR II database in Halifax, Canada. An analysis of sexual and urinary function will be presented separately, including European Organization for Research and Treatment of Cancer (EORTC) QLQ-PR25 together with data from clinical follow-up. Health-related quality-of-life instruments The instruments used and reported here were EQ-5D™, EORTC QLQ-C30 and EORTC QLQ-CR38. Validated Swedish, Dutch, Danish, English and German translations of the instrument were used8. EuroQol – 5D The EQ-5D™ is a generic measure of health status. It is a standardized non-disease-specific (generic) instrument for assessing self-reported health status, allowing for comparisons across disease groups9. It comprises a description of the patient's health in five dimensions (mobility, self-care, daily activity, pain/discomfort and anxiety/depression). One of three levels is chosen for each dimension; the first level denotes no problems or a low level of symptoms, whereas the third level denotes severe problems or a high level of symptoms. Also included in the instrument is a vertical ‘thermometer’ (EQ-VAS) in which the patient is asked to assess their global health on a visual analogue scale from 0 (worst imaginable health state) to 100 (best imaginable). Respondents were requested to assess their health status on the day they filled out the questionnaire.

o included in the instrument is a vertical ‘thermometer’ (EQ-VAS) in which the patient is asked to assess their global health on a visual analogue scale from 0 (worst imaginable health state) to 100 (best imaginable). Respondents were requested to assess their health status on the day they filled out the questionnaire. European Organization for Research and Treatment of Cancer QLQ-C30 and QLQ-CR38 The EORTC QLQ-C30 is a questionnaire developed to assess the quality of life of patients with cancer. The instrument available at the start of the study (2004) was version 3.0, a 30-item instrument designed for self-administration. The validated Swedish, English, Dutch, Danish and German translations were used10,11. This instrument has cross-cultural validity and the psychometric properties are considered satisfactory12. Normative data are available for German13 and Swedish14 patients as well as reference values15.

signed for self-administration. The validated Swedish, English, Dutch, Danish and German translations were used10,11. This instrument has cross-cultural validity and the psychometric properties are considered satisfactory12. Normative data are available for German13 and Swedish14 patients as well as reference values15. The QLQ-C30 questionnaire consists of 30 questions16. Both multi-item and single-item scales are constructed from the questions. There are five functional scales (physical, role, emotional, cognitive and social functioning), three symptom scales (fatigue, nausea/vomiting and pain), six single-item questions (about dyspnoea, insomnia, loss of appetite, constipation, diarrhoea and financial difficulties) and a global health/quality-of-life index. The latter assesses overall health and overall quality of life on a seven-point scale, where 1 indicating very poor and 7 indicating excellent. All other questions have four possible answers: ‘not at all’, ‘a little’, ‘quite a bit’ and ‘very much’. The time frame was ‘during the past week’.

ealth/quality-of-life index. The latter assesses overall health and overall quality of life on a seven-point scale, where 1 indicating very poor and 7 indicating excellent. All other questions have four possible answers: ‘not at all’, ‘a little’, ‘quite a bit’ and ‘very much’. The time frame was ‘during the past week’. The EORTC QLQ-CR38 questionnaire is used to measure more specific information about quality of life in patients with colorectal cancer. It is constructed in a similar manner to QLQ-C30. Thirty-eight questions cover four functional scales/single items (body image, sexual functioning, sexual enjoyment, future perspective) and eight symptom scales/items (micturition problems, chemotherapy side-effects, gastrointestinal symptoms, male sexual problems, female sexual problems, defaecation problems, stoma-related problems and weight loss). At the start of the study in 2004, QLQ-CR38 was available in the appropriate languages. For both instruments individual scores were converted to a score ranging from 0 to 100, according to the EORTC scoring manuals. A high score for the symptom/item scales represents a high level of symptoms/problems, whereas a high score for the functional scales and the global health/general quality-of-life index represents a high level of functioning, overall health and quality of life.

ing from 0 to 100, according to the EORTC scoring manuals. A high score for the symptom/item scales represents a high level of symptoms/problems, whereas a high score for the functional scales and the global health/general quality-of-life index represents a high level of functioning, overall health and quality of life. Statistical analysis Because the study was piggy-backed on to a randomized trial with power calculated for the primary endpoint, no power calculation was performed for the HRQL component. Missing data were handled as instructed in the EORTC scoring manual. All statistical analysis of demographic data, relevant clinical outcome measures and differences between study groups was carried out using SPSS® 20 software (IBM, Armonk, New York, USA). Comparisons of groups at baseline were made using Student's t test, χ2 test and, where appropriate, Fisher's exact test. EQ-5D™ global health was analysed at each assessment by means of the independent t test and repeated-measurement ANOVA was used for analysis over time. Proportions of patients reporting each level of the five dimensions were analysed by χ2 test or Fisher's exact test. As few patients reported problems at level 3 (severe problems), levels 2 and 3 were pooled in most analyses. QLQ-C30 and QLQ-CR38 global quality-of-life, functional and symptom scales were analysed using ANCOVA with baseline (preoperative score) as a co-variable and surgical procedure as a factor. The results are presented as mean changes, adjusted for baseline, with 95 per cent confidence intervals.

pooled in most analyses. QLQ-C30 and QLQ-CR38 global quality-of-life, functional and symptom scales were analysed using ANCOVA with baseline (preoperative score) as a co-variable and surgical procedure as a factor. The results are presented as mean changes, adjusted for baseline, with 95 per cent confidence intervals. All statistical analyses were carried out on the basis of intention to treat. P < 0·050 was considered statistically significant. Owing to the explorative nature of this study, significant P values should be interpreted with care, and considered as interesting findings rather than conclusive evidence. Results The COLOR II trial included 1103 patients between 2004 and 2010. In all, 617 patients were eligible for the HRQL study (Fig. 1). Thirty-three patients were excluded from the COLOR II trial after randomization as they did not conform to the inclusion criteria, and another 199 were primarily eligible but were not included owing to logistical difficulties in retrieving preoperative HRQL data, organizing preoperative radiation, language difficulties, patients' cognitive disabilities or lack of consent. Thus, 385 patients were included in the study (260 laparoscopic and 125 open). The included patients had a lower American Society of Anesthesiologists grade and fewer had undergone preoperative radiation compared with eligible patients who were not included. Basic demographic characteristics and clinical data did not differ between the laparoscopic and open groups (Table 1). Fig. 1 Study flow chart. HRQL, health-related quality of life; EQ-5D™, EuroQol – 5D

Results The COLOR II trial included 1103 patients between 2004 and 2010. In all, 617 patients were eligible for the HRQL study (Fig. 1). Thirty-three patients were excluded from the COLOR II trial after randomization as they did not conform to the inclusion criteria, and another 199 were primarily eligible but were not included owing to logistical difficulties in retrieving preoperative HRQL data, organizing preoperative radiation, language difficulties, patients' cognitive disabilities or lack of consent. Thus, 385 patients were included in the study (260 laparoscopic and 125 open). The included patients had a lower American Society of Anesthesiologists grade and fewer had undergone preoperative radiation compared with eligible patients who were not included. Basic demographic characteristics and clinical data did not differ between the laparoscopic and open groups (Table 1). Fig. 1 Study flow chart. HRQL, health-related quality of life; EQ-5D™, EuroQol – 5D Table 1 Demographics for health-related quality-of-life study and for those not included in this study

Results The COLOR II trial included 1103 patients between 2004 and 2010. In all, 617 patients were eligible for the HRQL study (Fig. 1). Thirty-three patients were excluded from the COLOR II trial after randomization as they did not conform to the inclusion criteria, and another 199 were primarily eligible but were not included owing to logistical difficulties in retrieving preoperative HRQL data, organizing preoperative radiation, language difficulties, patients' cognitive disabilities or lack of consent. Thus, 385 patients were included in the study (260 laparoscopic and 125 open). The included patients had a lower American Society of Anesthesiologists grade and fewer had undergone preoperative radiation compared with eligible patients who were not included. Basic demographic characteristics and clinical data did not differ between the laparoscopic and open groups (Table 1). Fig. 1 Study flow chart. HRQL, health-related quality of life; EQ-5D™, EuroQol – 5D Table 1 Demographics for health-related quality-of-life study and for those not included in this study Not included in HRQL study (n = 199) Included in HRQL study (n = 385) P§ Laparoscopic (n = 260) Open (n = 125) P§ Age (years)* 67 (66·0, 69·4) 67·1 (66·1, 68·1) 0·696¶ 67·4 (66·1, 68·6) 66·6 (64·8, 68·4) 0·487¶ Sex ratio (M : F) 123 : 76 239 : 146 0·949 162 : 98 77 : 48 0·893 Body mass index (kg/m2)* 25·9 (25·3, 26·5) 26·0 (25·6, 26·5) 0·750¶ 26·0 (25·4, 26·6) 26·1 (25·3, 26·8) 0·898¶ ASA fitness grade 0·624 I 37 (18·9) 103 (26·8) 0·008 69 (26·5) 34 (27·2) II 101 (50·8) 224 (58·2) 149 (57·3) 75 (60·0) III 48 (24·1) 55 (14·3) 40 (15·4) 15 (12·0) IV 1 (0·5) 2 (0·5) 2 (0·8) 0 (0) Unknown 12 (6·0) 1 (0·3) 0 (0) 1 (0·8) Tumour stage† 0·552 I 8 (4·0) 22 (5·7) 0·262 18 (6·9) 4 (3·2) II 71 (35·7) 135 (35·1) 93 (35·8) 42 (33·6) III 101 (50·8) 207 (53·8) 135 (51·9) 72 (57·6) IV 7 (3·5) 12 (3·1) 9 (3·5) 3 (2·4) Unknown 12 (6·0) 9 (2·3) 5 (1·9) 4 (3·2) Type of resection 0·956 Partial mesorectal excision 15 (7·5) 42 (10·9) 0·632 27 (10·4) 15 (12·0) Total mesorectal excision 112 (56·3) 219 (56·9) 147 (56·5) 72 (57·6) Abdominoperineal excision 64 (32·2) 116 (30·1) 80 (30·8) 36 (28·8) Other 3 (1·5) 6 (1·6) 4 (1·5) 2 (1·6) Unknown 5 (2·5) 2 (0·5) 2 (0·8) 0 (0) Preop. radiotherapy‡ 133 (66·8) 216 (56·1) 0·001 150 (57·7) 66 (52·8) 0·409 Short 110 (55·3) 157 (40·8) 110 (42·3) 47 (37·6) Long 23 (11·6) 40 (10·4) 27 (10·4) 13 (10·4) None 60 (30·2) 168 (43·6) 110 (42·3) 58 (46·4) Unknown 6 (3·0) 1 (0·3) 0 (0) 1 (0·8) Preop. chemotherapy 0·895 0·816 Yes 31 (15·6) 64 (16·6) 44 (16·9) 20 (16·0) No 145 (72·9) 290 (75·3) 195 (75·0) 95 (76·0) Unknown 23 (11·6) 31 (8·1) 21 (8·1) 10 (8·0) Conversion – – 65 (25·0) – Values in parentheses are percentages unless indicated otherwise;

) 110 (42·3) 58 (46·4) Unknown 6 (3·0) 1 (0·3) 0 (0) 1 (0·8) Preop. chemotherapy 0·895 0·816 Yes 31 (15·6) 64 (16·6) 44 (16·9) 20 (16·0) No 145 (72·9) 290 (75·3) 195 (75·0) 95 (76·0) Unknown 23 (11·6) 31 (8·1) 21 (8·1) 10 (8·0) Conversion – – 65 (25·0) – Values in parentheses are percentages unless indicated otherwise; * values are mean (95 per cent confidence interval). † Stage in the pathology report of the resected specimen. ‡ Short regimen comprised 5 × 5 Gy or less, and long programmes more than 5 days. Among those included in the health-related quality-of-life (HRQL) study, the dose specification was missing for 13 patients (5·0 per cent) in the laparoscopic group and six (4·8 per cent) in the open group. ASA, American Society of Anesthesiologists. § χ2 test, except ¶ Student's t test. The intention was to analyse the change in HRQL from baseline (preoperative data) over time and compare the groups. Analysis of stoma-related problems was therefore excluded from this part of the study. The actual results at 4 weeks, 6 months and 12 months regarding these problems, with comparisons between groups, are presented, but for obvious reasons without comparison with preoperative data (see Table 5).

and compare the groups. Analysis of stoma-related problems was therefore excluded from this part of the study. The actual results at 4 weeks, 6 months and 12 months regarding these problems, with comparisons between groups, are presented, but for obvious reasons without comparison with preoperative data (see Table 5). Compliance in answering the questionnaires was generally around 90 per cent at baseline and diminished over time to around 80 per cent at 12 months (Fig. 1). The compliance for EQ-5D™ was lower than this in the open group, being around 80 per cent at baseline and 70 per cent at 12 months. Compliance with the EQ-5D™ global health part was lower than for EQ-5D™ dimensions or EORTC questionnaires. For EORTC QLQ-C30 and QLQ-CR38 the answer rates were between 88 and 85 per cent at 4 weeks and 6 months, and 76–78 per cent at 12 months. There were no significant differences between the two groups at any time in overall health measured by EQ-5D™ (Table 2), nor was the repeated-measurement analysis significant (P = 0·171–0·966). Regarding the five dimensions, the only significant difference was in ‘daily activity’; a higher proportion of patients in the open group reported problems before treatment (level 2–3) (Table 3). Table 2 EuroQol – 5D global health scores

There were no significant differences between the two groups at any time in overall health measured by EQ-5D™ (Table 2), nor was the repeated-measurement analysis significant (P = 0·171–0·966). Regarding the five dimensions, the only significant difference was in ‘daily activity’; a higher proportion of patients in the open group reported problems before treatment (level 2–3) (Table 3). Table 2 EuroQol – 5D global health scores Preop. 4 weeks 6 months 12 months Mean(s.d.) EQ-5D™ score Laparoscopic 77·3(16·6) 64·2(20·8) 77·5(16·2) 79·4(15·9) Open 74·9(16·6) 62·6(20·4) 75·7(18·0) 78·7(15·1) Mean change* 2·4 (−1·5, 6·3) 1·6 (−3·3, 6·5) 1·7 (−2·4, 5·9) 0·6 (−3·4, 4·7) P† 0·228 0·981 0·815 0·646 * Values in parentheses are 95 per cent confidence intervals. EQ-5D™, EuroQol – 5D. † Independent t test. Table 3 Results for the five health dimensions of the EuroQol – 5D % of patients Preop. 4 weeks 6 months 12 months

Preop. 4 weeks 6 months 12 months Mean(s.d.) EQ-5D™ score Laparoscopic 77·3(16·6) 64·2(20·8) 77·5(16·2) 79·4(15·9) Open 74·9(16·6) 62·6(20·4) 75·7(18·0) 78·7(15·1) Mean change* 2·4 (−1·5, 6·3) 1·6 (−3·3, 6·5) 1·7 (−2·4, 5·9) 0·6 (−3·4, 4·7) P† 0·228 0·981 0·815 0·646 * Values in parentheses are 95 per cent confidence intervals. EQ-5D™, EuroQol – 5D. † Independent t test. Table 3 Results for the five health dimensions of the EuroQol – 5D % of patients Preop. 4 weeks 6 months 12 months Laparoscopic Open Laparoscopic Open Laparoscopic Open Laparoscopic Open Mobility Level 1 91 88 68 65 85 82 87 88 Level 2 9 12 30 34 14 18 13 11 Level 3 0 0 2 1 1 0 0 1 Self-care Level 1 99 98 87 85 96 92 96 97 Level 2 1 2 12 13 4 7 4 3 Level 3 0 0 1 2 0 1 0 0 Daily activity Level 1 89 80 40 37 73 72 80 80 Level 2 10 15 43 48 24 23 18 17 Level 3 1 5 17 15 3 5 2 2 Pain/discomfort Level 1 58 49 28 30 49 53 59 55 Level 2 40 49 66 68 49 44 40 44 Level 3 2 2 6 2 2 3 1 1 Anxiety/depression Level 1 60 51 55 52 68 68 72 68 Level 2 37 42 42 47 29 32 26 32 Level 3 3 7 3 1 3 0 2 0 Level 1, no problems; level 2, low level of symptoms; level 3, severe problems or high level of symptoms. The only significant difference between groups was in daily activity before treatment (P = 0·024, χ2 test).

Anxiety/depression Level 1 60 51 55 52 68 68 72 68 Level 2 37 42 42 47 29 32 26 32 Level 3 3 7 3 1 3 0 2 0 Level 1, no problems; level 2, low level of symptoms; level 3, severe problems or high level of symptoms. The only significant difference between groups was in daily activity before treatment (P = 0·024, χ2 test). HRQL measured by the cancer-specific EORTC QLQ-CR30 showed no statistically significant differences between groups in any dimension (global quality of life, five functional scales and three symptom scales) either before, or 4 weeks, 6 months and 12 months after surgery (Table 4). There were changes in most functional scales and symptoms between baseline and 4 weeks after surgery within both treatment groups. Global quality of life was restored by 12 months after both types of surgery, as were scores on most functional scales and symptoms, whereas emotional function had improved by 12 months. Table 4 Changes in function and symptom scores on European Organization for Research and Treatment of Cancer QLQ-C30 4 weeks 6 months 12 months Mean preop. score Mean change from preop. Adjusted mean difference (laparoscopic – open) Mean change from preop. Adjusted mean difference (laparoscopic – open) Mean change from preop.

Table 4 Changes in function and symptom scores on European Organization for Research and Treatment of Cancer QLQ-C30 4 weeks 6 months 12 months Mean preop. score Mean change from preop. Adjusted mean difference (laparoscopic – open) Mean change from preop. Adjusted mean difference (laparoscopic – open) Mean change from preop. Adjusted mean difference (laparoscopic – open) Global quality of life* Laparoscopic 72·8 (70·2, 75·3) −14.8 0·3 (−4·7, 5·3) −1.9 −2·2 (−6·8, 2·4) 2.1 −1·8 (−6·1, 2·4) Open 68·6 (64·7, 72·6) −11.9 3.0 6.3 Physical function* Laparoscopic 88·7 (86·8, 90·6) −21.6 0·2 (−4·8, 5·1) −6.7 −0·5 (−3·8, 2·8) −3.4 0 (−3·1, 3·0) Open 88·8 (86·0, 91·5) −21.6 −6.0 −3.3 Role function* Laparoscopic 80·9 (77·5, 84·4) −34.9 −1·7 (−9·0, 5·6) −4.9 1·7 (−4·4, 7·6) −0.8 −0·9 (−6·4, 4·6) Open 81·9 (76·8, 87·0) −33.7 −7.7 0.6 Emotional function* Laparoscopic 77·2 (74·4, 80·0) −2.5 −1·7 (−6·5, 3·0) 6.1 −2·4 (−6·4, 1·6) 7.1 −2·7 (−7·1, 1·6) Open 74·2 (70·1, 78·3) 1.2 10.2 12.0 Cognitive function* Laparoscopic 88·9 (86·8, 91·0) −8.4 −1·2 (−6·2, 3·7) −0.8 3·5 (−0·1, 7·1) −1.2 0 (−3·9, 3·8) Open 89·3 (86·5, 92·0) −7.1 −4.2 −0.8 Social function* Laparoscopic 87·0 (84·4, 89·5) −22.4 −0·4 (−7·0, 6·2) −8.1 0 (−5·5, 5·5) −3.1 0·7 (−4·7, 6·1) Open 84·4 (80·4, 88·5) −20.7 −7.1 −2.4 Fatigue† Laparoscopic 22·8 (19·8, 25·7) 25.0 2·1 (−3·8, 8·0) 5.2 −1·0 (−6·1, 4·0) 0.5 −0·1 (−4·5, 4·3) Open 25·8 (21·9, 29·6) 21.0 4.7 −1.6 Nausea and vomiting† Laparoscopic 4·9 (3·3, 6·5) 2.7 −2·8 (−6·5, 1·0) −1.4 −1·0 (−3·8, 1·9) −2.6 1·2 (−0·6, 2·9) Open 4·1 (2·4, 5·9) 6.0 0.4 −3.3 Pain† Laparoscopic 14·3 (11·6, 17·1) 18.5 −0·1 (−6·5, 6·3) 2.3 1·8 (−3·1, 6·8) 0.3 0·1 (−4·3, 4·4) Open 13·9 (9·9, 17·9) 18.5 0.4 −1.1 Dyspnoea† Laparoscopic 10·8 (8·4, 13·2) 10.0 −0·1 (−6·3, 6·1) 5.6 2·5 (−2·3, 7·3) 3.5 3·0 (−1·6, 7·7) Open 12·8 (8·2, 17·3) 9.4 1.8 −0.8 Insomnia† Laparoscopic 26·1 (22·5, 29·8) 4.5 −3·9 (−10·9, 3·0) −5.0 0·7 (−4·8, 6·3) −6.4 2·3 (−2·9, 7·5) Open 26·8 (21·6, 32·0) 8.1 −5.8 −9.8 Appetite loss† Laparoscopic 9·7 (7·0, 12·5) 17.1 −4·3 (−11·7, 3·2) −2.3 −2·0 (−6·7, 2·8) −3.9 1·8 (−1·8, 5·4) Open 10·3 (6·3, 14·2) 21.1 −0.4 −6.6 Constipation† Laparoscopic 12·8 (9·7, 16·0) −2.4 −0·7 (−5·9, 4·5) −4.5 −0·8 (−5·3, 3·6) −5.1 0·3 (−4·1, 4·7) Open 9·3 (5·2, 13·4) 1.0 1.1 −1.6 Diarrhoea† Laparoscopic 27·1 (23·0, 31·2) −11.0 2·5 (−3·5, 8·6) −8.1 1·6 (−4·9, 8·1) −7.9 6·0 (−0·4, 12·4) Open 30·5 (24·2, 36·9) −17.9 −15.1 −18.4 Financial difficulties† Laparoscopic 6·7 (4·0, 9·4) 4.4 −1·2 (−6·0, 3·7) 2.0 −0·4 (−5·1, 4·3) −1.3 −2·1

8 (−5·3, 3·6) −5.1 0·3 (−4·1, 4·7) Open 9·3 (5·2, 13·4) 1.0 1.1 −1.6 Diarrhoea† Laparoscopic 27·1 (23·0, 31·2) −11.0 2·5 (−3·5, 8·6) −8.1 1·6 (−4·9, 8·1) −7.9 6·0 (−0·4, 12·4) Open 30·5 (24·2, 36·9) −17.9 −15.1 −18.4 Financial difficulties† Laparoscopic 6·7 (4·0, 9·4) 4.4 −1·2 (−6·0, 3·7) 2.0 −0·4 (−5·1, 4·3) −1.3 −2·1 (−5·7, 1·6) Open 4·7 (1·5, 7·8) 6.2 3.1 2.1 Values in parentheses are 95 per cent confidence intervals. * A high value is positive to the patient; † a high value is negative to the patient. Table 5 Changes in scores on European Organization for Research and Treatment of Cancer QLQ-CR38 4 weeks 6 months 12 months

8 (−5·3, 3·6) −5.1 0·3 (−4·1, 4·7) Open 9·3 (5·2, 13·4) 1.0 1.1 −1.6 Diarrhoea† Laparoscopic 27·1 (23·0, 31·2) −11.0 2·5 (−3·5, 8·6) −8.1 1·6 (−4·9, 8·1) −7.9 6·0 (−0·4, 12·4) Open 30·5 (24·2, 36·9) −17.9 −15.1 −18.4 Financial difficulties† Laparoscopic 6·7 (4·0, 9·4) 4.4 −1·2 (−6·0, 3·7) 2.0 −0·4 (−5·1, 4·3) −1.3 −2·1 (−5·7, 1·6) Open 4·7 (1·5, 7·8) 6.2 3.1 2.1 Values in parentheses are 95 per cent confidence intervals. * A high value is positive to the patient; † a high value is negative to the patient. Table 5 Changes in scores on European Organization for Research and Treatment of Cancer QLQ-CR38 4 weeks 6 months 12 months Mean preop. score Mean change from preop. Adjusted mean difference (laparoscopic – open) Mean change from preop. Adjusted mean difference (laparoscopic – open) Mean change from preop. Adjusted mean difference (laparoscopic – open) Body image† Laparoscopic 90·3 (88·2, 92·4) −17.8 0·4 (−5·9, 6·6) −13.8 −2·0 (−7·9, 3·9) −11.5 −2·8 (−8·7, 3·0) Open 87·4 (83·8, 90·9) −17.1 −10.1 −6.6 Future perspective† Laparoscopic 57·1 (53·5, 60·7) 5.6 2·0 (−4·5, 8·4) 10.2 −2·4 (−8·6, 3·8) 11.8 −2·7 (−8·9, 3·6) Open 54·0 (48·2, 59·8) 5.7 14.3 16.7 GI symptoms‡ Laparoscopic 17·6 (15·7, 19·4) 6.9 2·6 (−1·1, 6·3) −0.6 0·5 (−2·7, 3·7) −0.8 0·1 (−3·0, 3·2) Open 17·1 (14·6, 19·6) 4.5 −0.5 −1.1 Defaecation problems‡ Laparoscopic 26·5 (24·1, 29·0) 7.2 2·7 (−5·7, 11·0) 2.4 5·9 (0·2, 11·6) −1.2 4·2 (−0·4, 8·7) Open 26·0 (22·2, 29·8) 6.7 −3.4 −5.8 Weight loss‡ Laparoscopic 14·7 (11·7, 17·7) 22.8 −3·7 (−11·2, 3·9) −0.7 −1·6 (−7·3, 4·2) −5.6 1·6 (−2·8, 6·0) Open 14·5 (10·2, 18·9) 26.5 −0.4 −8.0 Chemotherapy side-effects‡ Laparoscopic 8·8 (7·2, 10·5) 13.8 −0·9 (−5·8, 4·0) 6.8 −0·8 (−5·0, 3·5) 2.7 0 (−3·7, 3·6) Open 10·5 (7·3, 13·7) 13.5 6.6 1.3 Stoma-related problems*‡ Laparoscopic – 30.6 −1·0 (−6·7, 4·7) 25.2 −4·8 (−11·0, 1·4) 27.5 −1·3 (9·4, 6·7) Open – 31.6 30.0 28.8 Values in parentheses are 95 per cent confidence intervals.

ffects‡ Laparoscopic 8·8 (7·2, 10·5) 13.8 −0·9 (−5·8, 4·0) 6.8 −0·8 (−5·0, 3·5) 2.7 0 (−3·7, 3·6) Open 10·5 (7·3, 13·7) 13.5 6.6 1.3 Stoma-related problems*‡ Laparoscopic – 30.6 −1·0 (−6·7, 4·7) 25.2 −4·8 (−11·0, 1·4) 27.5 −1·3 (9·4, 6·7) Open – 31.6 30.0 28.8 Values in parentheses are 95 per cent confidence intervals. * Only six patients had a stoma before surgery. Values at 4 weeks, 6 months and 12 months are mean scores instead of mean change in score. † A high value is positive to the patient; ‡ a high value is negative to the patient. GI, gastrointestinal. There were no differences between groups in EORTC QLQ-CR38 data at any time point measured (Table 5). Future perspective scores improved over time in both groups, with no difference between the two surgical techniques.

† A high value is positive to the patient; ‡ a high value is negative to the patient. GI, gastrointestinal. There were no differences between groups in EORTC QLQ-CR38 data at any time point measured (Table 5). Future perspective scores improved over time in both groups, with no difference between the two surgical techniques. Discussion This study has shown no difference in the changes to HRQL within 12 months after laparoscopic and open surgery for rectal cancer. It is important to evaluate what constitutes a clinically significant difference. In regard to EORTC QLQ-C30, several studies have examined the minimal important change (MID) implicating a change that is clinically meaningful to the patient. Osoba17 has suggested that the MID is in the range of 5–10 points on the 100-point scale, whereas over 20 points indicates a substantial change. In the present study, the changes reported for most functional scales and symptoms, in both the EORTC QLQ-C30 and QLQ-C38, were substantial or moderate after 4 weeks, and gradually diminished over time. All results were within narrow confidence intervals, which supports the validity of the results, and also excludes any ‘clinically relevant’ differences between the groups.

ctional scales and symptoms, in both the EORTC QLQ-C30 and QLQ-C38, were substantial or moderate after 4 weeks, and gradually diminished over time. All results were within narrow confidence intervals, which supports the validity of the results, and also excludes any ‘clinically relevant’ differences between the groups. Physical functioning, role functioning, social function and fatigue measured by QLQ-C30 showed substantial deterioration 4 weeks after surgery. All of these functional/symptom scales improved after 6 months and were fully recovered at 12 months. The time frame differed from that in laparoscopic surgery for colonic cancer, where physical function and role function were reduced after 2 weeks, but partially recovered within 4 weeks1,2. It appears that patients with rectal cancer require a longer time to recover after curative surgery. There was a selection bias in the present study cohort as participants were somewhat healthier in general than the entire COLOR II trial cohort. This could be the result of logistics related to radiotherapy treatment. For patients with a high level of co-morbidity the ability and/or inclination to answer questionnaires might be reduced. This was, however, true for both groups and the authors suggest that the lack of difference between laparoscopic and open surgery is valid.

the result of logistics related to radiotherapy treatment. For patients with a high level of co-morbidity the ability and/or inclination to answer questionnaires might be reduced. This was, however, true for both groups and the authors suggest that the lack of difference between laparoscopic and open surgery is valid. There is no obvious explanation for the difference in compliance between the laparoscopic and open groups at baseline (Fig. 1). It is also intriguing that the compliance varied for the different instruments as they were sent out as a complete booklet at each time point. In particular, compliance in completion of EQ-5D™ at baseline differed, with lower compliance in the open group. The trial was not blinded so the patients were aware of which technique they had been randomized to. It could be speculated that, having agreed to participate in a randomized trial testing a new and presumably less invasive surgical technique, patients would be more ‘positive’ to the new technique and so those randomized to laparoscopy would also comply with the demands of this substudy. Baseline clinical data in the two groups were similar and, if the difference in compliance had represented a systematic difference in recruitment, differences in the results would have been expected. It is therefore argued that this difference most probably arose by chance.

lso comply with the demands of this substudy. Baseline clinical data in the two groups were similar and, if the difference in compliance had represented a systematic difference in recruitment, differences in the results would have been expected. It is therefore argued that this difference most probably arose by chance. HRQL assessment is important when evaluating new treatments. Patients today have a longer life expectancy, and the overall improved results of rectal cancer treatment, with 5-year survival rates of more than 60 per cent, indicate that there will be many survivors. The present results are therefore of interest as they reflect patients' experience after rectal cancer surgery. As the surgical technique resulted in no difference in HRQL, other factors, such as reduction in the risk of small bowel obstruction18,19 or the amount of perioperative bleeding or postoperative pain20, may influence the choice of surgical technique for rectal cancer.

t patients' experience after rectal cancer surgery. As the surgical technique resulted in no difference in HRQL, other factors, such as reduction in the risk of small bowel obstruction18,19 or the amount of perioperative bleeding or postoperative pain20, may influence the choice of surgical technique for rectal cancer. The fact that HRQL after rectal cancer surgery is substantially reduced for a prolonged period is noteworthy, indicating the need for a high level of healthcare support for several months after operation. This is in agreement with the finding of Wilson and co-workers21, who reported that HRQL was impaired for up to 6 months after rectal cancer surgery. The present study showed clinically meaningful changes at 4 weeks after surgery, regardless of the surgical technique and for most functional scales, but these returned to, or were close to, preoperative values by 6 months. The findings in this HRQL study do not mirror the improved short-term clinical outcomes reported after laparoscopic colonic surgery, such as reduced pain and earlier restoration of bowel function. This could possibly be explained by the time points chosen for HRQL measurements, the first of the questionnaires being completed at 4 weeks after operation. A previous study of patients who had surgery for inflammatory bowel disease found that body image was rated more highly after laparoscopic than open surgery22. This was not demonstrated here and, although speculative, body image may have been less important to the older patients in this trial.

The fact that HRQL after rectal cancer surgery is substantially reduced for a prolonged period is noteworthy, indicating the need for a high level of healthcare support for several months after operation. This is in agreement with the finding of Wilson and co-workers21, who reported that HRQL was impaired for up to 6 months after rectal cancer surgery. The present study showed clinically meaningful changes at 4 weeks after surgery, regardless of the surgical technique and for most functional scales, but these returned to, or were close to, preoperative values by 6 months. The findings in this HRQL study do not mirror the improved short-term clinical outcomes reported after laparoscopic colonic surgery, such as reduced pain and earlier restoration of bowel function. This could possibly be explained by the time points chosen for HRQL measurements, the first of the questionnaires being completed at 4 weeks after operation. A previous study of patients who had surgery for inflammatory bowel disease found that body image was rated more highly after laparoscopic than open surgery22. This was not demonstrated here and, although speculative, body image may have been less important to the older patients in this trial. The authors are grateful to K. Inglis, E. Lindholm, K. Druhan (data manager), S. Skullman, Z. Läckberg, G. Kurlberg and M. Cuesta for their help with the study.

A previous study of patients who had surgery for inflammatory bowel disease found that body image was rated more highly after laparoscopic than open surgery22. This was not demonstrated here and, although speculative, body image may have been less important to the older patients in this trial. The authors are grateful to K. Inglis, E. Lindholm, K. Druhan (data manager), S. Skullman, Z. Läckberg, G. Kurlberg and M. Cuesta for their help with the study. The HRQL study was supported by research grants from the Swedish Cancer Foundation (2010/593), Region Västra Götaland, Sahlgrenska University Hospital (ALF grant 138751, agreement concerning research and education of doctors) and the Alice Swenson Foundation. Ethicon EndoSurgery provided financial support for the administration of the COLOR II trial. Disclosure: The authors declare no other conflict of interest.

Introduction Historically, the majority of colorectal cancer resections were open operations. There is growing enthusiasm for laparoscopic colorectal cancer surgery, with short-term advantages and no negative oncological consequences. Laparoscopic surgery can be challenging technically and associated with a long learning curve1–3. Although randomized trials provide evidence, their results may not always be directly transferable to the general population1,4. Monitoring the introduction and outcomes of laparoscopic colorectal cancer surgery ensures that patients receive quality care in a cost-effective manner. This study assessed the early introduction and outcomes of laparoscopic colorectal cancer surgery in the English health system. Methods The National Cancer Data Repository (NCDR) contains information about every patient diagnosed with cancer in England and allows their treatment pathway to be mapped from diagnosis to cure or death. It consists of linked cancer registry, Hospital Episode Statistics (HES) and National Bowel Cancer Audit Project (NBOCAP) data.

ational Cancer Data Repository (NCDR) contains information about every patient diagnosed with cancer in England and allows their treatment pathway to be mapped from diagnosis to cure or death. It consists of linked cancer registry, Hospital Episode Statistics (HES) and National Bowel Cancer Audit Project (NBOCAP) data. Information was extracted from the NCDR on all individuals who underwent a major resection for primary colorectal cancer (International Classification of Diseases 10th revision C18–C20) diagnosed between 1 January 2006 and 31 December 2008. Information on age, sex, tumour site, date of diagnosis, Index of Multiple Deprivation (IMD) income category (based on postcode at diagnosis) and modified Dukes' stage at diagnosis were extracted from the registry data component of the NCDR. Modified Dukes' stage was used as, over the time period of this study, this was the only staging information captured both by English cancer registries and by the NBOCAP. Information about patient management, including operation type, approach to surgery and hospital of treatment, was derived from HES. If data on modified Dukes' stage at diagnosis or approach to surgery were missing from the HES and cancer registry data in the NCDR, this information was taken from the NBOCAP data set. Standard methods were used to identify whether each patient underwent a major resection for colorectal cancer up to 1 month before or 12 months after the date of diagnosis5,6.

osis or approach to surgery were missing from the HES and cancer registry data in the NCDR, this information was taken from the NBOCAP data set. Standard methods were used to identify whether each patient underwent a major resection for colorectal cancer up to 1 month before or 12 months after the date of diagnosis5,6. Patients undergoing laparoscopic operations were identified as those with Classification of Interventions and Procedures version 4 (OPCS-4) codes indicating minimal access to abdominal cavity (Y75) or other specified approach to abdominal cavity (Y508) recorded on the same date as the major resection. Converted laparoscopic operations were identified as those with an OPCS-4 code indicating failed minimal access approach converted to open (Y714). Information on approach to surgery was also incorporated from the NBOCAP data set. A Charlson co-morbidity score7 was calculated for each individual based on diagnostic codes (excluding cancer) recorded for any hospital admission in the year before diagnosis of the colorectal tumour, excluding any admission spanning the date of diagnosis. The cancer component of the Charlson index was derived from the cancer registry information in the NCDR. Patients were grouped into Charlson score categories of 0, 1, 2 and at least 3, higher scores indicating greater co-morbidity.

gnosis of the colorectal tumour, excluding any admission spanning the date of diagnosis. The cancer component of the Charlson index was derived from the cancer registry information in the NCDR. Patients were grouped into Charlson score categories of 0, 1, 2 and at least 3, higher scores indicating greater co-morbidity. The urgency of surgery is known to have a strong prognostic impact on outcomes, but this information is not recorded routinely in HES. The method of admission is, however, available. Patients who were admitted as an emergency and underwent surgery within 2 days of admission were deemed to have undergone emergency surgery.

The urgency of surgery is known to have a strong prognostic impact on outcomes, but this information is not recorded routinely in HES. The method of admission is, however, available. Patients who were admitted as an emergency and underwent surgery within 2 days of admission were deemed to have undergone emergency surgery. Statistical analysis The proportion of procedures performed via open, laparoscopic or converted surgery were examined in relation to patient age, sex, year of diagnosis, modified Dukes' stage of disease at diagnosis, tumour location, IMD category, operation type and Charlson co-morbidity score. Factors associated with the use of laparoscopic surgery were also investigated using a hierarchical random-effects binary logistic regression model, fitted using Stata® Statistical Software Release 11 (StataCorp LP, College Station, Texas, USA). The model was built with a hierarchy of patients clustered within hospitals (level 2), so allowing for correlations between patient outcomes. Co-variables (explanatory variables) in the risk-adjusted model included age (per 10-year increase), sex, tumour site, IMD income category, year of diagnosis, stage at diagnosis, Charlson co-morbidity score, operation type (elective or emergency) and operative approach. Approach to surgery was categorized as open or laparoscopic; converted operations were included in the laparoscopic group on an intention-to-treat basis. Some case-mix information (such as stage of disease and socioeconomic deprivation category) was missing from the NCDR as it was not recorded routinely in the database.

pproach to surgery was categorized as open or laparoscopic; converted operations were included in the laparoscopic group on an intention-to-treat basis. Some case-mix information (such as stage of disease and socioeconomic deprivation category) was missing from the NCDR as it was not recorded routinely in the database. Analyses restricted to patients with complete data would not have allowed the overall outcome to be assessed. Missing data were, therefore, imputed deterministically using the ICE command8, with passive and substitute options and ordered logistic regression for ten imputations and ten cycles of regression switching. It was assumed that the data were ‘missing at random’. The imputation model consisted of 30-day postoperative mortality, survival time, length of hospital stay, age at diagnosis, sex, median annual workload of the hospital, modified Dukes' stage, IMD income category, operation type (elective or emergency), admission type (elective or emergency), year of diagnosis, year of operation, method of access (open, laparoscopic completed, laparoscopic converted), Charlson co-morbidity score, tumour site, hospital and cancer registry. For comparative purposes the models were built using both the imputed data set and a data set restricted to cases with complete data.

y), year of diagnosis, year of operation, method of access (open, laparoscopic completed, laparoscopic converted), Charlson co-morbidity score, tumour site, hospital and cancer registry. For comparative purposes the models were built using both the imputed data set and a data set restricted to cases with complete data. To investigate the relationship between laparoscopic treatment and the outcomes postoperative mortality, long-term survival and postoperative length of hospital stay, logistic, Cox and linear regression hierarchical random-effects models were fitted. Thirty-day postoperative mortality was defined as death within 30 days of major resection. Survival time was calculated from the date of major resection to the date of death or when censored (30 June 2010). Length of stay was defined as the number of days from major resection to the end of the associated hospital stay (calculated taking into account transfers between different hospitals). Length of stay was log-transformed before analysis with estimates back-transformed and interpreted as length of stay ratios. Length of stay values of less than 1·00 indicate a shorter stay, values greater than 1·00 indicate a longer stay, and values of 1·00 indicate no change in the duration of hospital stay due to the variable of interest.

-transformed before analysis with estimates back-transformed and interpreted as length of stay ratios. Length of stay values of less than 1·00 indicate a shorter stay, values greater than 1·00 indicate a longer stay, and values of 1·00 indicate no change in the duration of hospital stay due to the variable of interest. Results Between 2006 and 2008, 58 135 major colorectal cancer resections were performed, of which 10 955 (18·8 per cent) were attempted laparoscopically. In total, 9303 (84·9 per cent) of these were completed laparoscopically and 1652 (15·1 per cent) were converted to open procedures. Use of the laparoscopic approach increased from 10·0 (95 per cent c.i. 8·1 to 12·0) to 28·4 (25·4 to 31·4) per cent over the study period. The proportion of patients in whom laparoscopy was attempted ranged from 0 to 65·6 per cent of the major resections in each hospital. Similarly, conversion rates varied from 0 to 46·2 per cent of all operations attempted laparoscopically. Patient characteristics are summarized in Table 1, and Table 2 shows the results of multivariable analyses investigating the use of laparoscopic surgery. Table 1 Characteristics of the study population Open Laparoscopic attempted

Results Between 2006 and 2008, 58 135 major colorectal cancer resections were performed, of which 10 955 (18·8 per cent) were attempted laparoscopically. In total, 9303 (84·9 per cent) of these were completed laparoscopically and 1652 (15·1 per cent) were converted to open procedures. Use of the laparoscopic approach increased from 10·0 (95 per cent c.i. 8·1 to 12·0) to 28·4 (25·4 to 31·4) per cent over the study period. The proportion of patients in whom laparoscopy was attempted ranged from 0 to 65·6 per cent of the major resections in each hospital. Similarly, conversion rates varied from 0 to 46·2 per cent of all operations attempted laparoscopically. Patient characteristics are summarized in Table 1, and Table 2 shows the results of multivariable analyses investigating the use of laparoscopic surgery. Table 1 Characteristics of the study population Open Laparoscopic attempted Total n* Multilevel imputed† n* Multilevel imputed† Age at diagnosis (years) < 60 10 305 8426 (81·8) 81·6 (79·1, 84·0) 1879 (18·2) 18·4 (16·0, 20·9) 60–69 16 076 12 859 (80·0) 79·4 (76·9, 81·9) 3217 (20·0) 20·6 (18·1, 23·1) 70–79 19 709 15 982 (81·1) 81·0 (78·6, 83·4) 3727 (18·9) 19·0 (16·6, 21·4) ≥ 80 12 045 9913 (82·3) 82·3 (79·9, 84·7) 2132 (17·7) 17·7 (15·3, 20·1) Year of diagnosis 2006 18 841 16 964 (90·0) 90·0 (88·0, 91·9) 1877 (10·0) 10·0 (8·1, 12·0) 2007 19 336 15 839 (81·9) 81·6 (78·9, 84·3) 3497 (18·1) 18·4 (15·7, 21·1) 2008 19 958 14 377 (72·0) 71·6 (68·6, 74·6) 5581 (28·0) 28·4 (25·4, 31·4) Sex M 32 361 26 400 (81·6) 81·3 (78·9, 83·6) 5961 (18·4) 18·7 (16·4, 21·1) F 25 774 20 780 (80·6) 80·6 (78·1, 83·0) 4994 (19·4 19·4 (17·0, 21·9) Operation type Elective 51 530 40 865 (79·3) 79·1 (76·6, 81·7) 10 665 (20·7) 20·9 (18·3, 23·4) Emergency 6605 6315 (95·6) 95·2 (94·2, 96·2) 290 (4·4) 4·8 (3·8, 5·8) Modified Dukes' stage at diagnosis A 7583 5661 (74·7) 75·0 (72·0, 78·0) 1922 (25·3) 25·0 (22·0, 28·0) B 20 982 16 942 (80·7) 80·4 (77·9, 82·8) 4040 (19·3) 19·6 (17·2, 22·1) C 20 370 16 770 (82·3) 82·1 (79·8, 84·5) 3600 (17·7) 17·9 (15·5, 20·2) ‘D’ 4992 4317 (86·5) 86·6 (84·6, 88·7) 675 (13·5) 13·4 (11·3, 15·4) Unknown 4208 3490 (82·9) – 718 (17·1) IMD income category 1 (most affluent) 12 161 9753 (80·2) 79·8 (77·2, 82·4) 2408 (19·8) 20·2 (17·6, 22·8) 2 12 745 10 290 (80·7) 80·4 (78·0, 82·9) 2455 (19·3) 19·6 (17·1, 22·0) 3 12 535 10 113 (80·7) 80·8 (78·3, 83·2) 2422 (19·3) 19·2 (16·8, 21·7) 4 10 994 9019 (82·0) 81·3 (78·7, 83·8) 1975 (18·0) 18·7 (16·2, 21·3) 5 (most deprived) 8784 7298 (83·1) 81·8 (79·3, 84·4) 1486 (16·9) 18·2 (15·6, 20·7) Unknown 916 707 (77·2) – 209 (22·8) – Cancer site Colon 42 814 34 576 (80·8) 80·6 (78·3, 82·9) 8238 (19·2) 19·4 (17·1, 21·7) Rectum 15 321 12 604 (82·3 81·7 (79·0, 84·5) 2717 (17·7) 18·3 (15·5, 21·0) Charlson co-morbidity score 0 46 957 37 717 (80·3) 80·2 (77·8, 82·0) 9240 (19·7) 19·8 (17·4, 22·2) 1 7325 6177 (84·3) 83·9 (81·6, 86·2) 1148 (15·7) 16·1 (13·8, 18·4) 2 2513 2139 (85·1) 84·9 (82·7, 87·1) 374 (14·9) 15·1 (12·9, 17·3) ≥ 3 1340 1147 (85·6) 85·4 (82·9, 87·9) 193 (14·4) 14·6 (12·1, 17·1) Values in parentheses are *perc

-morbidity score 0 46 957 37 717 (80·3) 80·2 (77·8, 82·0) 9240 (19·7) 19·8 (17·4, 22·2) 1 7325 6177 (84·3) 83·9 (81·6, 86·2) 1148 (15·7) 16·1 (13·8, 18·4) 2 2513 2139 (85·1) 84·9 (82·7, 87·1) 374 (14·9) 15·1 (12·9, 17·3) ≥ 3 1340 1147 (85·6) 85·4 (82·9, 87·9) 193 (14·4) 14·6 (12·1, 17·1) Values in parentheses are *perc entages and †95 per cent confidence intervals. IMD, Index of Multiple Deprivation. Table 2 Odds of use of an attempted laparoscopic approach Complete case Multiple imputation Odds ratio P Odds ratio P Age at diagnosis (per 10 years) 0·99 (0·96, 1·01) 0·217 0·98 (0·96, 1·00) 0·059 Year of diagnosis 2·04 (1·97, 2·10) < 0·001 2·06 (2·00, 2·12) < 0·001 Sex 0·228 0·090 M 1·00 – 1·00 F 1·03 (0·98, 1·08) 1·04 (0·99, 1·09) Operation type < 0·001 < 0·001 Elective 1·00 1·00 Emergency 0·14 (0·12, 0·16) 0·15 (0·13, 0·17) Modified Dukes' stage at diagnosis < 0·001 < 0·001 A 1·00 1·00 B 0·73 (0·68, 0·78) 0·74 (0·69, 0·79) C 0·66 (0·61, 0·71) 0·66 (0·62, 0·71) ‘D’ 0·43 (0·39, 0·48) 0·45 (0·40, 0·50) IMD income category 0·001 0·002 1 (most affluent) 1·00 1·00 2 0·98 (0·91, 1·05) 0·96 (0·90, 1·04) 3 0·97 (0·90, 1·04) 0·96 (0·89, 1·03) 4 0·91 (0·84, 0·98) 0·89 (0·83, 0·97) 5 (most deprived) 0·84 (0·77, 0·92) 0·85 (0·78, 0·93) Cancer site < 0·001 < 0·001 Colon 1·00 1·00 Rectum 0·72 (0·68, 0·76) 0·71 (0·67, 0·75) Charlson co-morbidity score < 0·001 < 0·001 0 1·00 1·00 1 0·77 (0·71, 0·83) 0·77 (0·72, 0·83) 2 0·76 (0·67, 0·87) 0·74 (0·66, 0·84) ≥ 3 0·68 (0·57, 0·81) 0·69 (0·58, 0·82) Values in parentheses are 95 per cent confidence intervals. IMD, Index of Multiple Deprivation.

1·00 1·00 Rectum 0·72 (0·68, 0·76) 0·71 (0·67, 0·75) Charlson co-morbidity score < 0·001 < 0·001 0 1·00 1·00 1 0·77 (0·71, 0·83) 0·77 (0·72, 0·83) 2 0·76 (0·67, 0·87) 0·74 (0·66, 0·84) ≥ 3 0·68 (0·57, 0·81) 0·69 (0·58, 0·82) Values in parentheses are 95 per cent confidence intervals. IMD, Index of Multiple Deprivation. Not all procedures that were attempted laparoscopically could be completed by this route and Table 3 describes the features of patients whose procedure was converted to an open operation. The conversion rate varied in relation to various patient factors; the multivariable analyses investigating the odds of a laparoscopically attempted operation being converted are shown in Table 4. Year of diagnosis and patient age had no impact on the odds of conversion of an attempted laparoscopic procedure. The odds of conversion was reduced in women, but increased with advanced tumour stage, socioeconomic deprivation, and with rectal rather than colonic tumours. Table 3 Characteristics of patients in whom laparoscopic surgery was completed and those in whom it was converted to an open procedure Laparoscopic completed Converted from laparoscopic

Not all procedures that were attempted laparoscopically could be completed by this route and Table 3 describes the features of patients whose procedure was converted to an open operation. The conversion rate varied in relation to various patient factors; the multivariable analyses investigating the odds of a laparoscopically attempted operation being converted are shown in Table 4. Year of diagnosis and patient age had no impact on the odds of conversion of an attempted laparoscopic procedure. The odds of conversion was reduced in women, but increased with advanced tumour stage, socioeconomic deprivation, and with rectal rather than colonic tumours. Table 3 Characteristics of patients in whom laparoscopic surgery was completed and those in whom it was converted to an open procedure Laparoscopic completed Converted from laparoscopic n* Multilevel imputed† n* Multilevel imputed† Age at diagnosis (years) < 60 1628 (86·6) 86·5 (84·6, 88·4) 251 (13·4) 13·5 (11·6, 15·4) 60–69 2711 (84·3) 84·3 (82·6, 86·0) 506 (15·7) 15·7 (14·0, 17·4) 70–79 3133 (84·1) 83·6 (81·9, 85·3) 594 (15·9) 16·4 (14·7, 18·1) ≥ 80 1831 (85·9) 85·6 (83·7, 87·4) 301 (14·1) 14·4 (12·6, 16·3) Year of diagnosis 2006 1629 (86·8) 86·4 (84·3, 88·5) 248 (13·2) 13·6 (11·5, 15·7) 2007 2943 (84·2) 84·1 (82·5, 85·8) 554 (15·8) 15·9 (14·2, 17·5) 2008 4731 (84·8) 84·5 (82·9, 86·1) 850 (15·2) 15·5 (13·9, 17·1) Sex M 4919 (82·5) 82·2 (80·6, 83·8) 1042 (17·5) 17·8 (16·2, 19·4) F 4384 (87·8) 87·6 (86·3, 88·9) 610 (12·2) 12·4 (11·1, 13·7) Operation type Elective 9083 (85·2) 84·9 (83·7, 86·2) 1582 (14·8) 15·1 (13·8, 16·3) Emergency 220 (75·9) 75·6 (69·6, 81·6) 70 (24·1) 24·4 (18·4, 30·4) Modified Dukes' stage at diagnosis A 1677 (87·3) 87·3 (85·3, 89·2) 245 (12·7) 12·7 (10·8, 14·7) B 3423 (84·7) 84·5 (83·1, 86·0) 617 (15·3) 15·5 (14·0, 16·9) C 3033 (84·3) 84·1 (82·4, 85·8) 567 (15·7) 15·9 (14·2, 17·6) ‘D’ 554 (82·1) 82·4 (79·0, 85·9) 121 (17·9) 17·6 (14·1, 21·0) Unknown 616 (85·8) – 102 (14·2) – IMD income category 1 (most affluent) 2084 (86·5) 86·3 (84·5, 88·0) 324 (13·5) 13·7 (12·0, 15·5) 2 2114 (86·1) 86·0 (84·2, 87·8) 341 (13·9) 14·0 (12·2, 15·8) 3 2057 (84·9) 85·1 (83·2, 87·0) 365 (15·1) 14·9 (13·0, 16·8) 4 1650 (83·5) 83·5 (81·7, 85·4) 325 (16·5) 16·5 (14·6, 18·3) 5 (most deprived) 1219 (82·0) 82·1 (79·6, 84·6) 267 (18·0) 17·9 (15·4, 20·4) Unknown 179 (85·6) – 30 (14·4) – Cancer site Colon 7062 (85·7) 85·5 (84·3, 86·7) 1176 (14·3) 14·5 (13·3, 15·7) Rectum 2241 (82·5) 82·2 (80·2, 84·3) 476 (17·5) 17·8 (15·7, 19·8) Charlson co-morbidity score 0 7851 (85·0) 84·7 (83·5, 86·0) 1389 (15·0) 15·3 (14·0, 16·5) 1 968 (84·3) 84·3 (82·0, 86·6) 180 (15·7) 15·7 (13·4, 18·0) 2 316 (84·5) 84·4 (80·2, 88·6) 58 (15·5) 15·6 (11·4, 19·8) ≥ 3 168 (87·0) 87·0 (82·1, 92·0) 25 (13·0) 13·0 (8·0, 17·9) Values in parentheses are *percentages and †95 per cent confidence intervals. IMD, Index of Multiple Deprivation.

89 (15·0) 15·3 (14·0, 16·5) 1 968 (84·3) 84·3 (82·0, 86·6) 180 (15·7) 15·7 (13·4, 18·0) 2 316 (84·5) 84·4 (80·2, 88·6) 58 (15·5) 15·6 (11·4, 19·8) ≥ 3 168 (87·0) 87·0 (82·1, 92·0) 25 (13·0) 13·0 (8·0, 17·9) Values in parentheses are *percentages and †95 per cent confidence intervals. IMD, Index of Multiple Deprivation. Table 4 Odds of an attempted laparoscopic operation being converted to an open procedure Complete case Multiple imputation Odds ratio P Odds ratio P Age at diagnosis (per 10 years) 1·02 (0·96, 1·07) 0·553 1·04 (0·99, 1·09) 0·152 Year of diagnosis 1·05 (0·97, 1·14) 0·198 1·05 (0·98, 1·13) 0·189 Sex < 0·001 < 0·001 M 1·00 1·00 F 0·63 (0·56, 0·71) 0·65 (0·58, 0·73) Operation type < 0·001 < 0·001 Elective 1·00 1·00 Emergency 2·05 (1·51, 2·79) 2·06 (1·54, 2·76) Modified Dukes' stage at diagnosis 0·002 0·002 A 1·00 1·00 B 1·29 (1·09, 1·52) 1·28 (1·08, 1·51) C 1·28 (1·08, 1·51) 1·30 (1·10, 1·54) ‘D’ 1·55 (1·20, 2·00) 1·56 (1·20, 2·03) IMD income category 0·001 0·001 1 (most affluent) 1·00 1·00 2 1·03 (0·86, 1·23) 1·02 (0·86, 1·21) 3 1·12 (0·94, 1·34) 1·12 (0·94, 1·33) 4 1·23 (1·02, 1·48) 1·25 (1·05, 1·50) 5 (most deprived) 1·47 (1·21, 1·80) 1·42 (1·17, 1·72) Cancer site < 0·001 < 0·001 Colon 1·00 1·00 Rectum 1·29 (1·13, 1·47) 1·29 (1·14, 1·46) Charlson co-morbidity score 0·720 0·755 0 1·00 1·00 1 1·05 (0·87, 1·26) 1·04 (0·87, 1·24) 2 1·00 (0·74, 1·36) 1·00 (0·74, 1·34) ≥ 3 0·79 (0·49, 1·25) 0·81 (0·52, 1·25) Values in parentheses are 95 per cent confidence intervals. IMD, Index of Multiple Deprivation.

lon 1·00 1·00 Rectum 1·29 (1·13, 1·47) 1·29 (1·14, 1·46) Charlson co-morbidity score 0·720 0·755 0 1·00 1·00 1 1·05 (0·87, 1·26) 1·04 (0·87, 1·24) 2 1·00 (0·74, 1·36) 1·00 (0·74, 1·34) ≥ 3 0·79 (0·49, 1·25) 0·81 (0·52, 1·25) Values in parentheses are 95 per cent confidence intervals. IMD, Index of Multiple Deprivation. Analyses investigating how the approach to surgery influenced outcomes are shown in Table 5. Length of stay and 30-day postoperative mortality were lower in patients in whom laparoscopic surgery was attempted. The effects were greatest among those in whom the operation was completed laparoscopically. Individuals in whom laparoscopic surgery was completed had a 40 per cent reduced risk of death within 1 year compared with those who had open surgery. Table 5 Results of a multivariable regression model investigating outcomes in relation to approach to surgery Measure of effect

Analyses investigating how the approach to surgery influenced outcomes are shown in Table 5. Length of stay and 30-day postoperative mortality were lower in patients in whom laparoscopic surgery was attempted. The effects were greatest among those in whom the operation was completed laparoscopically. Individuals in whom laparoscopic surgery was completed had a 40 per cent reduced risk of death within 1 year compared with those who had open surgery. Table 5 Results of a multivariable regression model investigating outcomes in relation to approach to surgery Measure of effect Approach to surgery Complete case Multiple imputation Postoperative length of stay Open 1·00 1·00 Laparoscopic completed 0·65 (0·64, 0·66) 0·65 (0·64, 0·66) Conversion 0·92 (0·89, 0·95) 0·93 (0·89, 0·96) 30-day postoperative mortality Open 1·00 1·00 Laparoscopic completed 0·57 (0·49, 0·66) 0·55 (0·48, 0·64) Conversion 0·67 (0·50, 0·90) 0·68 (0·52, 0·90) 1-year survival Open 1·00 1·00 Laparoscopic completed 0·61 (0·56, 0·66) 0·60 (0·55, 0·65) Conversion 0·86 (0·72, 1·03) 0·84 (0·71, 1·00) Values in parentheses are 95 per cent confidence intervals. The model was adjusted for sex, age at diagnosis, Index of Multiple Deprivation (IMD) income category, year of operation, tumour site (colon/rectum), modified Dukes' stage, operation type (elective or emergency) and presence of co-morbidity.

0·84 (0·71, 1·00) Values in parentheses are 95 per cent confidence intervals. The model was adjusted for sex, age at diagnosis, Index of Multiple Deprivation (IMD) income category, year of operation, tumour site (colon/rectum), modified Dukes' stage, operation type (elective or emergency) and presence of co-morbidity. Discussion This retrospective population-based study has provided a national perspective on the adoption of laparoscopic colorectal cancer surgery and its outcomes in England. Laparoscopic surgery was attempted more frequently in patients with a better prognosis (such as elective presentation of early-stage tumours). The odds of conversion were greater in individuals with more advanced disease and those who posed a greater operative risk. Laparoscopic surgery was associated with a shorter hospital stay, a lower 30-day postoperative mortality rate and improved long-term survival. The increased trend for laparoscopic surgery has been demonstrated in other studies9,10, but they examined only operations recorded as being completed laparoscopically and the coding of such procedures is often inaccurate in routine data sets4. In the present study the total number of operations attempted laparoscopically was calculated by including all procedures coded as laparoscopically converted. This approach has confirmed a rapid increase in the adoption of the techniques and provided a more complete picture.

s is often inaccurate in routine data sets4. In the present study the total number of operations attempted laparoscopically was calculated by including all procedures coded as laparoscopically converted. This approach has confirmed a rapid increase in the adoption of the techniques and provided a more complete picture. Differences existed between the populations selected for each surgical approach, with laparoscopically treated patients tending to have elective admissions for early-stage disease. Local guidance states that minimal access surgery is not appropriate for all patients but should be available as an option under favourable conditions, for example in individuals with a body mass index (BMI) below 30 kg/m2, no history of major abdominal surgery, tumours category T3 or less, rectal cancers not requiring a total mesorectal excision (TME) and in the absence of clinical or radiological signs of obstruction11. This study has provided indirect evidence indicating that these recommendations are being implemented. These recommendations are not absolute, however, and the authors appreciate that many experienced surgeons and units routinely offer laparoscopic surgery to more complex cases (for example patients with a BMI exceeding 30 kg/m2 or who require TME). Unfortunately, data items that would allow identification of the more challenging laparoscopic cases (such as those with an increased BMI or advanced tumour category) are not yet available in the NCDR. Their inclusion would further increase the utility of the resource.

with a BMI exceeding 30 kg/m2 or who require TME). Unfortunately, data items that would allow identification of the more challenging laparoscopic cases (such as those with an increased BMI or advanced tumour category) are not yet available in the NCDR. Their inclusion would further increase the utility of the resource. In the present study one in six laparoscopic procedures was converted and this proportion changed little over the course of the study. Laparoscopic experts would view this rate as too high as many units now report conversion rates below 5 per cent. This emphasizes the need for continued efforts in education and training to reduce the rate further1. The clinical factors that may make a conversion more likely have been documented previously12–14. The present analysis has confirmed that advanced stage of disease and co-morbidity12–16 consistently increase the likelihood of a conversion. Randomized trials have demonstrated oncological equivalence of open and laparoscopic techniques, whereas case series have reported better outcomes in laparoscopically treated patients17–19 to the extent that one group recommended that laparoscopy should be considered routine20. However, patients who require conversion to open operation may have had poorer postoperative outcomes in some series15,21,22. The present study found that 30-day postoperative mortality, length of hospital stay and 1-year survival was better in laparoscopically treated patients (irrespective of whether a procedure was converted or not).

Introduction Rectum-preserving surgery has been proposed with gradual acceptance for early-stage tumours and for those downstaged by neoadjuvant therapy1. Local excision, either alone or combined with radiotherapy, cures the majority of cases that appear confined to the bowel wall. However, even in this select group, recurrence rates as high as 30 per cent following organ-preserving treatment have been described2. Optimized organ preservation approaches would benefit from the identification of high-risk tumour characteristics other than the somewhat crude morphometric, radiological and histological stratification currently available3–5. Prediction of nodal metastasis by size or imaging characteristics is an imprecise science, particularly following neoadjuvant therapy6. The probability of local recurrence after transanal endoscopic microsurgery (TEMS) for early rectal cancer may be predicted based upon tumour diameter, depth of invasion and adverse histological features7. This model provides practical reassurance for individuals with a predicted low risk of recurrence (less than 5 per cent), where total mesorectal excision (TME) has little to offer. As risk increases (10–25 per cent), patients face a dilemma and must trade the prospect of undertreatment or overtreatment. In addition, these determinants are not applicable to preoperative samples. Hence, molecular signatures of good (or indeed poor) prognosis accessible via tumour biopsies would be useful biomarkers to stratify risk and inform organ-preserving decisions.

mma and must trade the prospect of undertreatment or overtreatment. In addition, these determinants are not applicable to preoperative samples. Hence, molecular signatures of good (or indeed poor) prognosis accessible via tumour biopsies would be useful biomarkers to stratify risk and inform organ-preserving decisions. Early translational research8,9 has shown that patients with tumours expressing particular molecular profiles have a poor prognosis. Epigenetic differences (for instance DNA methylation) predict recurrence of early-stage rectal cancer10. Two patterns of abnormal DNA methylation are observed in colorectal cancer: genome-wide hypomethylation or localized hypermethylation at or near tumour suppressor gene promoters11. It has been suggested12 that the loss of function of classical tumour suppressor genes by promoter hypermethylation is more common than by mutation. Various methylation-related mechanisms that trigger genetic changes and contribute to tumorigenesis have been described13–16. Biomarkers of tumour progression show promise in distinguishing indolent and aggressive cancer in other organs17,18. Prognostic models incorporating multiple biomarkers have been described in breast19 and hepatocellular20 cancers. The present authors have shown previously21 that hypermethylation of two or more genes in rectal tumours can be associated with early-stage disease. The objective of this retrospective study was to refine the prognostic utility of rectal cancer biomarkers in order to aid selection of appropriate patients for organ-preserving strategies1,22.

ve shown previously21 that hypermethylation of two or more genes in rectal tumours can be associated with early-stage disease. The objective of this retrospective study was to refine the prognostic utility of rectal cancer biomarkers in order to aid selection of appropriate patients for organ-preserving strategies1,22. Methods Samples Rectal tumour samples were taken from two patient groups treated with TME at the University Hospitals of Birmingham. The first group was obtained from a tissue bank of colorectal cancer specimens from patients treated between 2001 and 2004. Samples from the core of the resected tumour had been flash-frozen by the operating surgeon immediately after surgery. Histologically confirmed tumour-free samples (5 cm or more from the tumour edge) were frozen for paired analysis. The second group was obtained from histopathology archives of formalin-fixed, paraffin-embedded rectal tumour sections from patients treated between 2005 and 2010. Patients who had received neoadjuvant chemoradiotherapy, or had a family or personal history of inflammatory or malignant bowel disease were excluded. Ethical approval was obtained from the Black Country Research Ethics Committee.

ed, paraffin-embedded rectal tumour sections from patients treated between 2005 and 2010. Patients who had received neoadjuvant chemoradiotherapy, or had a family or personal history of inflammatory or malignant bowel disease were excluded. Ethical approval was obtained from the Black Country Research Ethics Committee. Macrodissection, DNA extraction and bisulphite treatment Genomic DNA extraction of fresh-frozen tissues was performed using the DNeasy® Blood & Tissue Kit (QIAGEN, Crawley, UK) according to the manufacturer's instructions. Approximately 2 mm3 tissue was used, and purified genomic DNA was made up to a final volume of 200 µl. DNA extracts were quantified and assessed for quality using a NanoDrop 1000 Spectrophotometer (Thermo Scientific, Loughborough, UK). Bisulphite modification of 1 µg genomic DNA was performed using the EpiTect® Bisulfite Kit (QIAGEN), according to the manufacturer's instructions. Paraffin-embedded blocks were sectioned at 5 µm thickness and mounted on glass slides. A reference slide for each sample was stained with haematoxylin and eosin. Areas containing more than 80 per cent of tumour tissue were marked to guide sampling from adjacent, non-haematoxylin and eosin-stained sections. DNA extraction and bisulphite treatment of formalin-fixed, paraffin-embedded tissues was performed using the EpiTect® Plus Bisulfite Kit (QIAGEN), according to the manufacturer's instructions.

ore than 80 per cent of tumour tissue were marked to guide sampling from adjacent, non-haematoxylin and eosin-stained sections. DNA extraction and bisulphite treatment of formalin-fixed, paraffin-embedded tissues was performed using the EpiTect® Plus Bisulfite Kit (QIAGEN), according to the manufacturer's instructions. Bisulphite pyrosequencing Oestrogen receptor 1 gene (ESR1) and unc-5 homologue C gene (UNC5C) assays were purchased from QIAGEN. The glutathione S-transferase pi 1 gene (GSTP1) primer sequences and polymerase chain reaction (PCR) conditions are available at the PyroMark™ assay database (http://techsupport.pyrosequencing.com). Retinoic acid receptor β (RARB), long interspersed nucleotide element 1 (LINE-1) and adenomatous polyposis coli (APC) gene assays have been described previously21,23,24. Primers for the remaining nine assays were designed using the MethPrimer software (http://www.urogene.org/methprimer/index1.html). Primer sequences and the PCR annealing temperatures (Tm), and the number of cytosine–guanine nucleotide (CpG) sites examined are shown in Table S1 (supporting information). PCR for assays designed in-house was performed using 1 µl bisulphite-treated DNA, 5 pmol forward and reverse primers, 5 µl 10 × buffer, 1.5 µl 50-mmol/l magnesium chloride, 2·5 mmol of each dNTP and 1·5 units IMMOLASE™ DNA polymerase (Bioline, London, UK), made up to 50 µl with sterile distilled water. The PCR conditions were: 95°C for 10 min; 45 cycles of 95°C for 20 s, Tm for 20 s and 72°C for 30 s; and final extension at 72°C for 5 min. PCR for the remaining assays was performed according to either the manufacturer's instructions or published protocols (LINE-1). Negative water controls were included to ensure no contamination.

tions were: 95°C for 10 min; 45 cycles of 95°C for 20 s, Tm for 20 s and 72°C for 30 s; and final extension at 72°C for 5 min. PCR for the remaining assays was performed according to either the manufacturer's instructions or published protocols (LINE-1). Negative water controls were included to ensure no contamination. Some 5 µl PCR products were analysed on a 1 per cent agarose gel before pyrosequencing. Remaining PCR products were captured on streptavidin-coated beads, denatured and washed, followed by the addition of sequencing primer. Pyrosequencing was performed using PyroMark™ Gold Q96 reagents on a PyroMark™ Q96 ID machine (QIAGEN). Samples were repeated on different days to assess reproducibility. Internal validation was performed using bisulphite-treated unmethylated and methylated genomic DNA (CpGenome™ Universal DNA; Merck–Millipore, Watford, UK). The instrument software (Pyro Q-CpG™; QIAGEN) automatically calculates the percentage methylation at each CpG site in the assay by quantifying the relative peak heights of thymine/cytosine. Methylation for each gene was expressed as the median percentage methylation across all CpG sites.

Merck–Millipore, Watford, UK). The instrument software (Pyro Q-CpG™; QIAGEN) automatically calculates the percentage methylation at each CpG site in the assay by quantifying the relative peak heights of thymine/cytosine. Methylation for each gene was expressed as the median percentage methylation across all CpG sites. Analysis of KRAS mutation KRAS mutational analysis at codons 12 and 13 was performed using pyrosequencing, as described previously25,26. PCR reactions of 50 µl were set up using a standard protocol of 40 cycles at Tm of 59°C using 0.3 µl MyTaq™ (Bioline) enzyme. PCR products were processed and run on the PyroMark™ Q96 ID machine. The software automatically generates percentage values for codon 12 and 13 single-nucleotide polymorphisms. Nucleotide substitutions greater than 2·5 per cent were considered indicative of KRAS mutation. Analysis of microsatellite instability status Microsatellite instability (MSI) status was analysed using a refined method of the ‘Bethesda panel’ of markers27. The modified panel consists of mononucleotide repeat markers NR-21, NR-24, NR-27, BAT-25 and BAT-26. Tumours with instability at two or more markers were defined as MSI-high, whereas tumours with single-marker instability were designated as MSI-low. Relevant marker sections were amplified in a 25-µl multiplex PCR reaction including 1 µl sample DNA (35 cycles, Tm 55°C, 0·2 µl MyTaq™ enzyme). Fragment analysis was performed on an ABI 3730 machine (Life Technologies, Paisley, UK). Microsatellite marker sizes were determined using Peak Scanner™ software (Life Technologies).

nt marker sections were amplified in a 25-µl multiplex PCR reaction including 1 µl sample DNA (35 cycles, Tm 55°C, 0·2 µl MyTaq™ enzyme). Fragment analysis was performed on an ABI 3730 machine (Life Technologies, Paisley, UK). Microsatellite marker sizes were determined using Peak Scanner™ software (Life Technologies). Statistical analysis Statistical significance between groups was determined using a Wilcoxon signed rank test for paired variables, Mann–Whitney U test for unpaired continuous variables, and Fisher's exact test for categorical variables. P < 0·050 was considered statistically significant.

nt marker sections were amplified in a 25-µl multiplex PCR reaction including 1 µl sample DNA (35 cycles, Tm 55°C, 0·2 µl MyTaq™ enzyme). Fragment analysis was performed on an ABI 3730 machine (Life Technologies, Paisley, UK). Microsatellite marker sizes were determined using Peak Scanner™ software (Life Technologies). Statistical analysis Statistical significance between groups was determined using a Wilcoxon signed rank test for paired variables, Mann–Whitney U test for unpaired continuous variables, and Fisher's exact test for categorical variables. P < 0·050 was considered statistically significant. To design a predictive model using percentage methylation of individual CpGs for each gene of interest, a two-step logistic regression model was undertaken using Stata® version 12.1 (StataCorp LP, College Station, Texas, USA). This was done to predict lymph node status, lymphovascular invasion (LVI) and distant metastasis – histopathological parameters considered to be associated with adverse patient outcomes. In the first step, a reverse stepwise logistic regression model was performed using CpGs in each gene as independent variables. The threshold for significance was set at P < 0·100; significant CpGs were taken forward into the next stage, where these CpGs were entered into a reverse stepwise logistic regression model as independent variables, with the significance threshold for removal from the model set at P < 0·100. MSI status and KRAS mutation status were also entered into the model. The fit of the model was modelled using the area under the curve (AUC) statistic and a model was considered a good predictor when the AUC was above 0·70. Goodness-of-fit testing was carried out using standard residuals plots and the Hosmer–Lemeshow test (with P < 0·100 indicating an inadequate model). Model threshold for prediction was set pragmatically, aiming to maximize sensitivity as close as possible to 90 per cent or above, and keeping specificity close to 50 per cent or above, on the basis that high-frequency detection of adverse features was more important than inclusion of patients without these features within the cohort.

for prediction was set pragmatically, aiming to maximize sensitivity as close as possible to 90 per cent or above, and keeping specificity close to 50 per cent or above, on the basis that high-frequency detection of adverse features was more important than inclusion of patients without these features within the cohort. Results Clinical and pathological data are summarized in Table  1. Gene-specific hypermethylation was observed in rectal cancer, compared with matched adjacent normal mucosa (Figs 1 and 2; Fig. S1, supporting information). As expected, the methylation level of LINE-1, a marker of global methylation, was lower in rectal cancer (Fig. 2). Median percentage methylation was calculated for each of 15 genes for analysis of association with tumour stage and pathology variables (Figs 6). Early-stage (pathological tumour (pT) 1–2) and node-negative tumours had higher median percentage methylation for RARB (Figs 3 and 4). Larger tumours (median diameter 40 mm or more) had higher methylation levels of checkpoint with forkhead and ring finger gene, CHFR (Fig. 6), as did more advanced lesions (pT3–4). Node-negative tumours also exhibited increased methylation levels of chemokine ligand 12 gene (CXCL12) and death-associated protein kinase 1 gene (DAPK1) compared with node-positive tumours (Fig. 4). Tumours with either LVI or distant metastasis were associated with lower methylation values of RARB (Figs 5 and 6). In addition, tumours with organ secondaries were associated with less methylation of cadherin 13 gene, CDH13, and CXCL12 (Fig. 5).

otein kinase 1 gene (DAPK1) compared with node-positive tumours (Fig. 4). Tumours with either LVI or distant metastasis were associated with lower methylation values of RARB (Figs 5 and 6). In addition, tumours with organ secondaries were associated with less methylation of cadherin 13 gene, CDH13, and CXCL12 (Fig. 5). Table 1 Patient demographics and clinicopathological characteristics of samples used Adjacent tissue(n = 64) Rectal cancer(n = 133) Age (years)* 71 (33–89)   70 (31–92) Sex ratio (M : F) 42 : 22 78 : 55 TNM stage I 32 (24·1) II 43 (32·3) III 41 (30·8) IV 17 (12·8) Tumour size (mm)*   40 (9–110) Vascular invasion No 87 (65·4) Yes 46 (34·6) Degree of differentiation Well 12 (9·0) Moderate 115 (86·5) Poor 6 (4·5) Values in parentheses are percentages unless indicated otherwise * values are median (range). TNM, tumour node metastasis classification. Fig 1 Methylation levels of a APC, b CDH13, c CHFR, d RARB and e ESR1 genes in matched adjacent tissues and rectal cancers. Horizontal bars represent median methylation levels. a–e P < 0·001 (2-tailed Wilcoxon signed rank test) Fig 2 Methylation levels of a CXCL12, b DAPK1, c UNC5C, d MINT3, e MINT17 and f LINE-1 genes in matched adjacent tissues and rectal cancers. Horizontal bars represent median methylation levels. a,c–f P < 0·001, b P = 0·007 (2-tailed Wilcoxon signed rank test) Fig 3 Methylation levels of a RARB and b CHFR genes, stratified according to depth of invasion. Horizontal bars represent median methylation levels. a P < 0·001, b P = 0·005 (2-tailed Mann–Whitney U test)

Fig 2 Methylation levels of a CXCL12, b DAPK1, c UNC5C, d MINT3, e MINT17 and f LINE-1 genes in matched adjacent tissues and rectal cancers. Horizontal bars represent median methylation levels. a,c–f P < 0·001, b P = 0·007 (2-tailed Wilcoxon signed rank test) Fig 3 Methylation levels of a RARB and b CHFR genes, stratified according to depth of invasion. Horizontal bars represent median methylation levels. a P < 0·001, b P = 0·005 (2-tailed Mann–Whitney U test) Fig 4 Methylation levels of a RARB, b CXCL12 and c DAPK1 genes, stratified according to nodal metastasis: N−, nodal metastasis absent; N+, nodal metastasis present. Horizontal bars represent median methylation levels. a P = 0·008, b P = 0·021, c P = 0·022 (2-tailed Mann–Whitney U test) Fig 5 Methylation levels of a RARB, b CDH13 and c CXCL12 genes, stratified according to distant metastasis: M−, distant metastasis absent; M+, distant metastasis present. Horizontal bars represent median methylation levels. a P < 0·001, b P = 0·027, c P = 0·018 (2-tailed Mann–Whitney U test) Fig 6 Methylation levels of a RARB and b CHFR genes, stratified according to a lymphovascular invasion (LVI; absent or present) and b tumour size (less than, or greater than or equal to the median tumour size of 40 mm). Horizontal bars represent median methylation levels. a P = 0·038, b P = 0·011 (2-tailed Mann–Whitney U test)

thylation levels of a RARB and b CHFR genes, stratified according to a lymphovascular invasion (LVI; absent or present) and b tumour size (less than, or greater than or equal to the median tumour size of 40 mm). Horizontal bars represent median methylation levels. a P = 0·038, b P = 0·011 (2-tailed Mann–Whitney U test) One-third of rectal cancers (44 of 133) carried the KRAS mutation at codon 12 or 13. In contrast, only six rectal tumours were found to be microsatellite unstable (3 MSI-high and 3 MSI-low). Neither KRAS mutation nor MSI status was associated with advanced disease in univariable analysis. Construction of a multivariable biomarker model to predict disease progression in rectal cancer Bisulphite pyrosequencing can quantify methylation at multiple sequential CpGs, a feature that is absent in other methylation platforms. Interrogation of individual CpGs, rather than the mean across a locus, may identify the most important or useful sites. A multilevel reverse stepwise logistic regression model was constructed using individual CpGs to identify genes associated with disease progression: LVI, lymph node metastasis and distant metastasis (Table S2, supporting information).

rather than the mean across a locus, may identify the most important or useful sites. A multilevel reverse stepwise logistic regression model was constructed using individual CpGs to identify genes associated with disease progression: LVI, lymph node metastasis and distant metastasis (Table S2, supporting information). For LVI, the model identified seven CpGs from three genes as the most informative variables: CDH1 (sites 1, 2 and 4), CDH13 (sites 2 and 5) and methylated-in-tumour 3, MINT3 (sites 6 and 9) (Table 2). A model score was developed by summating the percentage methylation of each CpG, maintaining the direction of the coefficient. The AUC for this model was 0·76 (95 per cent confidence interval (c.i.) 0·68 to 0·84), and setting a cut-off score of −20 gave a sensitivity of 85·0 per cent, specificity of 45·3 per cent with a positive predictive value of (PPV) of 62·5 per cent and a negative predictive value (NPV) of 74·0 per cent (Fig. 7). Table 2 Biomarker model of genes associated with histopathological features of disease progression Histopathological feature Genes Lymphovascular invasion CDH1, CDH13, MINT3 Lymph node metastasis CDH1, CDH13, MINT3, CXCL12, RARB, APC Distant metastasis CDH1, MINT3, CXCL12, RARB, ESR1, CHFR Fig 7 Receiver operating characteristic (ROC) curves constructed for a lymphovascular invasion, b lymph node metastasis and c distant metastasis. Area under ROC curve: a 0·76, b 0·76, c 0·82

CDH13, MINT3 Lymph node metastasis CDH1, CDH13, MINT3, CXCL12, RARB, APC Distant metastasis CDH1, MINT3, CXCL12, RARB, ESR1, CHFR Fig 7 Receiver operating characteristic (ROC) curves constructed for a lymphovascular invasion, b lymph node metastasis and c distant metastasis. Area under ROC curve: a 0·76, b 0·76, c 0·82 For lymph node metastasis, ten CpGs from six genes were found to be significant: CDH1 (site 10), CDH13 (sites 7 and 8), MINT3 (sites 3 and 4), CXCL12 (sites 1 and 3), RARB (site 6) and APC (sites 1 and 6) (Table 2). The AUC for the model was found to be 0·76 (95 per cent c.i. 0·68 to 0·84), and setting a cut-off score of −43 gave a sensitivity of 91·1 per cent, specificity of 55·3 per cent, PPV of 60·0 per cent and NPV of 89·4 per cent (Fig. 7). For distant metastasis, nine CpGs from six genes were selected from the model: CDH1 (sites 2 and 6), MINT3 (sites 6 and 8), CXCL12 (sites 1 and 3), RARB (site 5), ESR1 (site 4) and CHFR (site 6) (Table 2). The AUC was 0·82 (95 per cent c.i. 0·73 to 0·91), and use of a cut-off score of −40 gave a sensitivity of 100 per cent and specificity of 51·3 per cent, with a PPV of 100 per cent and a NPV of 90·6 per cent (Fig. 7).

INT3 (sites 6 and 8), CXCL12 (sites 1 and 3), RARB (site 5), ESR1 (site 4) and CHFR (site 6) (Table 2). The AUC was 0·82 (95 per cent c.i. 0·73 to 0·91), and use of a cut-off score of −40 gave a sensitivity of 100 per cent and specificity of 51·3 per cent, with a PPV of 100 per cent and a NPV of 90·6 per cent (Fig. 7). Discussion Clinical outcomes for early rectal cancer are heterogeneous and it is becoming clear that genetic events play a role in determining such variation. Following on from previous work21 showing an association between gene methylation and pathological indices of disease progression, the present study explored its feasibility as a predictive biomarker. The results suggest that a multimarker methylation model is needed to achieve high sensitivity for the prediction of disease progression. Single-gene testing is unlikely to be clinically useful for prognosis; although MSI immunohistochemistry is becoming routine and may confer prognostic significance in certain scenarios28, it is notable that multiple mutations are tested simultaneously.

achieve high sensitivity for the prediction of disease progression. Single-gene testing is unlikely to be clinically useful for prognosis; although MSI immunohistochemistry is becoming routine and may confer prognostic significance in certain scenarios28, it is notable that multiple mutations are tested simultaneously. The majority of early rectal cancers are node-negative, and routine removal of the mesorectum may be unnecessary. Deciding which patient is suitable for an organ-preserving approach requires more refined prognostic testing than is currently available. A pragmatic approach was undertaken when determining the sensitivity and specificity thresholds for the present predictive model. The aim was to achieve a high sensitivity (90 per cent or above) for patients at risk of disease progression with a reasonable specificity (close to 50 per cent). By stratifying 50 per cent of patients as high risk and 50 per cent as low risk, organ-preserving treatments could be offered to the latter according to the profile of a small subset of genes. Analysis of the methylome using next-generation sequencing or gene array technologies may reveal additional markers to stratify risk further and tailor treatment options (radiotherapy, local excision or radical surgery).

ving treatments could be offered to the latter according to the profile of a small subset of genes. Analysis of the methylome using next-generation sequencing or gene array technologies may reveal additional markers to stratify risk further and tailor treatment options (radiotherapy, local excision or radical surgery). The present work has provided a step towards accurate discrimination of indolent and aggressive rectal cancer subtypes. The next stage is to evaluate the present biomarker model in a prospective study. It can be tested as part of the Transanal Endoscopic Microsurgery and Radiotherapy in Early Rectal Cancer (TREC) study, which is already evaluating the feasibility of randomization between conventional surgery and a new organ-preserving protocol of short-course preoperative radiotherapy with a 10-week interval to TEMS29. Correlation of the molecular analysis with clinical outcome data will be required to determine the predictive accuracy of the present biomarker model. The authors are grateful to G. Caldwell, L. Tannahill and C. Jones for their advice on the experiments performed. This work was funded by the Medical Research Council and supported by the National Institute for Health Research Experimental Cancer Medicine Centre and the Queen Elizabeth Hospital Birmingham Charity. A.B. is supported by the Wellcome Trust Postdoctoral Fellowship. Disclosure: The authors declare no conflict of interest. Supporting information Additional supporting information may be found in the online version of this article

The authors are grateful to G. Caldwell, L. Tannahill and C. Jones for their advice on the experiments performed. This work was funded by the Medical Research Council and supported by the National Institute for Health Research Experimental Cancer Medicine Centre and the Queen Elizabeth Hospital Birmingham Charity. A.B. is supported by the Wellcome Trust Postdoctoral Fellowship. Disclosure: The authors declare no conflict of interest. Supporting information Additional supporting information may be found in the online version of this article Table S1 Primer sequences, PCR annealing temperatures of pyrosequencing assays and number of CpG sites examined (Word document) Table S2 Significant genes and their corresponding CpG sites for the prediction of lymph node involvement, lymphovascular invasion and distant metastasis in rectal cancer (Word document) Fig S1 Methylation levels of CDH1, GSTP1, CASP8 and TIMP3 genes in adjacent tissues and rectal cancers (TIFF file)

Introduction The field of molecular biology has undergone rapid advancement in the past 5 years, with exciting consequences for the diagnosis, treatment and follow‐up of surgical patients. A series of enabling technologies and projects have expanded the knowledge of how basic molecular biology can assist in the management of surgical disease. The first, and most important, was the Human Genome Project, established in 1990 by the US National Institutes of Health and the UK Sanger Centre1. This established the reference human genome by carrying out sequencing of multiple fragments of a reference human genome using the dye‐terminator technique described by Sanger and colleagues2. A consequence of this technology is that the project took 10 years to produce a single genome and cost over US $3 billion to complete. The development of microarray technology and next‐generation sequencing (NGS) within the past 5 years has led to a step‐change in the implementation of genomic technologies; before this, the bulk of genetic research was carried out on DNA microarrays.

A series of enabling technologies and projects have expanded the knowledge of how basic molecular biology can assist in the management of surgical disease. The first, and most important, was the Human Genome Project, established in 1990 by the US National Institutes of Health and the UK Sanger Centre1. This established the reference human genome by carrying out sequencing of multiple fragments of a reference human genome using the dye‐terminator technique described by Sanger and colleagues2. A consequence of this technology is that the project took 10 years to produce a single genome and cost over US $3 billion to complete. The development of microarray technology and next‐generation sequencing (NGS) within the past 5 years has led to a step‐change in the implementation of genomic technologies; before this, the bulk of genetic research was carried out on DNA microarrays. Genome‐wide association studies DNA microarrays are available from a variety of manufacturers (Illumina, Affymetrix and Agilent) and consist of silicon or glass slides with oligonucleotides complementary to the DNA sequence being studied, which are annealed to their surface. This allows cheap, mass production of microarrays that can be used for large population‐based studies. Typically these microarrays have between 500 000 and 1·5 million genomic markers, usually single‐nucleotide polymorphisms (SNPs). SNPs are single‐nucleotide changes within a gene that lead to protein change and subsequent change in the function of that gene. When scanned with a laser, each individual oligonucleotide fluoresces a specific colour, depending on the bound oligonucleotide fragment (Figs 1 and 2).

ingle‐nucleotide polymorphisms (SNPs). SNPs are single‐nucleotide changes within a gene that lead to protein change and subsequent change in the function of that gene. When scanned with a laser, each individual oligonucleotide fluoresces a specific colour, depending on the bound oligonucleotide fragment (Figs 1 and 2). Figure 1 Image of scanned oligonucleotide array (from Wikimedia Commons) BJS-9722-FIG-0001-cFigure 2 Affymetrix microarray chip from Wikimedia Commons BJS-9722-FIG-0002-cA variety of projects have been undertaken using DNA microarrays, typically taking the form of the genome‐wide association study (GWAS). These are usually case–control studies with cases enriched for the disease of interest. SNPs of interest are identified and taken forward to validation in larger cohorts, giving insights into the disease process being studied. Examples of GWASs include the COGENT (COlorectal cancer GENeTics) Consortium, and the Wellcome Trust Case Control Consortium 1/2 (WTCCC 1/2) examining colorectal cancer, Crohn's disease, diabetes, ischaemic heart disease, and several other common diseases and pathologies.

nsights into the disease process being studied. Examples of GWASs include the COGENT (COlorectal cancer GENeTics) Consortium, and the Wellcome Trust Case Control Consortium 1/2 (WTCCC 1/2) examining colorectal cancer, Crohn's disease, diabetes, ischaemic heart disease, and several other common diseases and pathologies. Next‐generation sequencing The most widely used NGS technology, sequencing by synthesis (Illumina, San Diego, California, USA) allows entire human genomes to be sequenced within 24 h at low cost, with the sub $1000 genome barrier (X‐prize) being achieved earlier this year. Other innovative technologies include single molecular real‐time sequencing (PacBio® SMRT™; Pacific Biosciences, Menlo Park, California, USA), semiconductor sequencing (Ion Torrent™; Life Technologies, Paisley, UK) and nanopore sequencing (Oxford Nanopore, Oxford, UK). Until recently, NGS was limited to small studies on a few samples owing to cost constraints, but because of the rapid fall in price‐per‐sample (Fig. 3), large studies are now in progress. The Cancer Genome Atlas (TCGA) project is sequencing the cancer genome of 33 different cancer types in the US population; the 100 000 Genomes Project in the UK is undertaking to sequence 50 000 cancer genomes and 50 000 rare disease genomes over the next 5 years. Figure 3 Cost in US dollars per genome sequenced in 2001–2014, at 5‐month intervals (data from http://www.genome.gov/sequencingcosts/)

Next‐generation sequencing The most widely used NGS technology, sequencing by synthesis (Illumina, San Diego, California, USA) allows entire human genomes to be sequenced within 24 h at low cost, with the sub $1000 genome barrier (X‐prize) being achieved earlier this year. Other innovative technologies include single molecular real‐time sequencing (PacBio® SMRT™; Pacific Biosciences, Menlo Park, California, USA), semiconductor sequencing (Ion Torrent™; Life Technologies, Paisley, UK) and nanopore sequencing (Oxford Nanopore, Oxford, UK). Until recently, NGS was limited to small studies on a few samples owing to cost constraints, but because of the rapid fall in price‐per‐sample (Fig. 3), large studies are now in progress. The Cancer Genome Atlas (TCGA) project is sequencing the cancer genome of 33 different cancer types in the US population; the 100 000 Genomes Project in the UK is undertaking to sequence 50 000 cancer genomes and 50 000 rare disease genomes over the next 5 years. Figure 3 Cost in US dollars per genome sequenced in 2001–2014, at 5‐month intervals (data from http://www.genome.gov/sequencingcosts/) BJS-9722-FIG-0003-cTo carry out NGS, an inherently massively parallel technique, several steps are required. All technologies require capture of DNA or RNA into sequencing libraries. Illumina Solexa™ then uses this captured DNA/RNA to generate clusters, which are amplified. Cluster amplification is the process whereby target DNA is immobilized on to spatially separated template sites, allowing sequencing reactions to occur in parallel. Sequencing is then carried out on the clusters present on the glass slide. The slide is portioned into eight channels, enabling independent samples to be run simultaneously. Typically, reads of between 75 and 100 base pairs are possible, and the nucleotide incorporation cycle is repeated until sufficient depth is covered for all targeted regions. Data generated are then aligned with a reference sequence, and variants are called after comparing the sequencing data to a reference. This allows a tumour specimen to be compared with its paired normal tissue sample.

nucleotide incorporation cycle is repeated until sufficient depth is covered for all targeted regions. Data generated are then aligned with a reference sequence, and variants are called after comparing the sequencing data to a reference. This allows a tumour specimen to be compared with its paired normal tissue sample. A critical difference in analysis of human cancers is the concept of germline and somatic mutations. Germline mutations are the constitutive DNA that the patient is born with, and variation within the germline confers an increased (or decreased) risk of cancer. Somatic mutations occur as a consequence of tumour development, although they may initiate tumour development by occurring spontaneously as a result of external factors such as ionizing radiation or carcinogens. Microarray and sequencing technologies also allow analysis of other types of genetic information, such as DNA methylation (epigenetics), which acts as a switch in the regulation of gene expression. NGS also allows analysis of gene expression by NGS of mRNA (RNA‐seq) and other genetic modifications, such as sequencing of chromatin‐immunoprecipitated DNA (ChIP‐seq). Non‐coding small RNAs that may affect gene function, such as long non‐coding RNA (lncRNA), small interfering RNA (siRNA) and small nucleolar RNA (snoRNA), can also be analysed by NGS.

expression by NGS of mRNA (RNA‐seq) and other genetic modifications, such as sequencing of chromatin‐immunoprecipitated DNA (ChIP‐seq). Non‐coding small RNAs that may affect gene function, such as long non‐coding RNA (lncRNA), small interfering RNA (siRNA) and small nucleolar RNA (snoRNA), can also be analysed by NGS. Other 'omics technologies Other technologies are emerging as potential methods for the downstream analysis and stratification of patient samples in surgical disease. Two examples of these technologies are metabolomics and proteomics. Metabolomics uses either nuclear magnetic resonance or mass spectrometry to ascertain the presence of metabolites in surgical specimens. The patterns and relative abundances of the metabolites observed can give clues as to the underlying biological processes at work in the tissues studied3. Proteomics uses mass spectrometry to understand the structure of proteins. Several methods exist to allow proteins to be studied in the ionized form without fragmentation or damage including matrix‐assisted laser desorption ionization (MALDI)4 and electrospray ionization5. Because proteins are modified after they are produced by transcription (post‐translational modification), study of both the protein and RNA involved in tissues allows a fuller appreciation of the changes that may be occurring in a particular disease.

ed laser desorption ionization (MALDI)4 and electrospray ionization5. Because proteins are modified after they are produced by transcription (post‐translational modification), study of both the protein and RNA involved in tissues allows a fuller appreciation of the changes that may be occurring in a particular disease. These technologies have allowed advances in understanding of the initiation and progression of multiple surgical diseases, and in adjuvant therapies for surgical disease such as chemotherapy and radiotherapy. They also allow the possibility of population screening of asymptomatic carriers.

ed laser desorption ionization (MALDI)4 and electrospray ionization5. Because proteins are modified after they are produced by transcription (post‐translational modification), study of both the protein and RNA involved in tissues allows a fuller appreciation of the changes that may be occurring in a particular disease. These technologies have allowed advances in understanding of the initiation and progression of multiple surgical diseases, and in adjuvant therapies for surgical disease such as chemotherapy and radiotherapy. They also allow the possibility of population screening of asymptomatic carriers. Lower gastrointestinal tract: colorectal cancer and inflammatory bowel disease Genetic predisposition The predominant surgically relevant disease types studied in the lower gastrointestinal (GI) tract have been colorectal cancer, Crohn's disease and ulcerative colitis. The COGENT Consortium6 has undertaken multiple GWASs of patients with colorectal cancer associated with a strong family history or extreme phenotype (such as young age of onset), identifying ten SNPs associated with colorectal cancer at population‐wide significance. Although inheritance of one disease association SNP confers a population risk of odds ratio approximately 1·05 (Table 1, taken from Tenesa and Dunlop7), inheritance of multiple SNPs (more than 10) confers a cumulative threefold risk of cancer. Other successes from the GWAS/NGS approach have been identification of the GREM‐associated duplication in hereditary mixed polyposis syndrome8 and mutations in the DNA polymerase genes POLE and POLD as a cause of hereditary polyposis9. Identification of these mutations allows familial testing, enhanced surveillance and reduction in the risk of developing colorectal cancer.

entification of the GREM‐associated duplication in hereditary mixed polyposis syndrome8 and mutations in the DNA polymerase genes POLE and POLD as a cause of hereditary polyposis9. Identification of these mutations allows familial testing, enhanced surveillance and reduction in the risk of developing colorectal cancer. Table 1 Frequency of identified single‐nucleotide polymorphisms in colorectal cancer, and their effect sizes (from Tenesa and Dunlop7) Gene/locus Chromosome SNP Effect size (odds ratio) Allele frequency Population attributable risk (%) – 8q24 rs6983267 1·21 (1·15, 1·27) 0·51 9·7 GREM1 15q13 rs4779584 1·26 (1·19, 1·34) 0·18 4·5 SMAD7 18q21 rs4939827 1·18 (1·12, 1·23) 0·52 8·6 – 11q23 rs3802842 1·12 (1·07, 1·17) 0·29 3·4 EIF3H 8q23 rs16892766 1·25 (1·19, 1·32) 0·07 1·7 — 10p14 rs10795668 1·12 (1·10, 1·16) 0·67 7·4 BMP4 14q21 rs4444235 1·11 (1·08, 1·15) 0·46 4·8 CDH1 16q22 rs9929218 1·10 (1·06, 1·12) 0·71 6·6 RHPN2 19q13 rs10411210 1·15 (1·10, 1·20) 0·90 11·9 BMP2 20q12 rs961253 1·12 (1·08, 1·16) 0·35 4·0 Values in parentheses are 95 per cent c.i. SNP, single‐nucleotide polymorphism. In Crohn's disease, multiple large‐population GWAS studies10 have been undertaken identifying multiple SNPs of predisposition, suggesting that Crohn's disease has a strong heritable component. In total, more than 73 SNPs have been identified, with the strongest association in the NOD2 gene, which plays an important role in immunity. In total, these loci make up about 20 per cent of the observed inheritability of Crohn's disease.

redisposition, suggesting that Crohn's disease has a strong heritable component. In total, more than 73 SNPs have been identified, with the strongest association in the NOD2 gene, which plays an important role in immunity. In total, these loci make up about 20 per cent of the observed inheritability of Crohn's disease. Comparatively less research has been undertaken in germline susceptibility to ulcerative colitis; several large population GWAS studies11, 12, 13 have demonstrated over 30 associated SNPs. These SNPs are in a variety of genes, but are associated predominantly with immune system and immunity‐related genes. In addition, approximately 50 per cent of identified loci overlap with those of Crohn's disease. Genomic analysis of colorectal cancer The colorectal cancer TCGA project14 has carried out exome sequencing (sequencing of the protein coding regions of the genome), RNA‐seq, genome‐wide methylation analysis and protein expression (via reverse‐phase protein arrays; RPPAs) of, at the time of writing, 461 colorectal tumours. This group has confirmed recurrent driver mutations in APC, TP53, SMAD4, PIK3CA and KRAS, but also found novel therapeutic targets in ARID1A, SOX9 and FAM123B. Another study15 observed gene fusions (merging of two genes, which causes abnormal function) in R‐spondin. The mutations observed in ARID1A are particularly exciting as they present a potential therapeutic target16. These data sets provide a wealth of information about colorectal cancer, and linkage to a clinical data set provides opportunities for future biomarker studies.

nes, which causes abnormal function) in R‐spondin. The mutations observed in ARID1A are particularly exciting as they present a potential therapeutic target16. These data sets provide a wealth of information about colorectal cancer, and linkage to a clinical data set provides opportunities for future biomarker studies. Recent work has examined the role of integration of multiple 'omics data sets to produce classifiers of disease17, also known as endotypes. These are based on mutation, expression and immunological data sets. The Colorectal Cancer Subtyping Consortium found four distinct Colorectal cancer Molecular Subtypes (CMSs) (Table 2). The classifiers identified provide insight into the biology of the distinct types. CMS1 consisted of microsatellite‐unstable, immunologically active tumours occurring mainly on the right side in the elderly, whereas CMS2 (the most frequent endotype) consisted of chromosomally unstable, microsatellite‐stable tumours. Further study of these classifiers may permit finer stratification, allowing precisely targeted therapy. Table 2 The Colorectal Cancer Subtyping Consortium classification of colorectal cancer

Recent work has examined the role of integration of multiple 'omics data sets to produce classifiers of disease17, also known as endotypes. These are based on mutation, expression and immunological data sets. The Colorectal Cancer Subtyping Consortium found four distinct Colorectal cancer Molecular Subtypes (CMSs) (Table 2). The classifiers identified provide insight into the biology of the distinct types. CMS1 consisted of microsatellite‐unstable, immunologically active tumours occurring mainly on the right side in the elderly, whereas CMS2 (the most frequent endotype) consisted of chromosomally unstable, microsatellite‐stable tumours. Further study of these classifiers may permit finer stratification, allowing precisely targeted therapy. Table 2 The Colorectal Cancer Subtyping Consortium classification of colorectal cancer Classifier Frequency (%) Characteristics CMS1 14 MSI, immune pathway activation/expression, right‐side tumours, older age at diagnosis, females, hypermutation, BRAF mutation, intermediate survival CMS2 41 High CIN, MSS, strong Wnt/Myc pathway activation, left‐side tumours, TP53 mutation, EGFR amplification/overexpression, better survival CMS3 8 Low CIN, moderate Wnt/Myc pathway activation, KRAS mutation, PIK3CA mutation, IGFBP2 overexpression, intermediate survival CMS4 20 CIN/MSI heterogeneous, mesenchymal/TGF‐β activation, younger age at diagnosis, NOTCH3/VEGFR2 overexpression, worse survival CMS, Colorectal cancer Molecular Subtype; MSI, microsatellite instability; CIN, chromosomal instability; MSS, microsatellite stable; TGF, transforming growth factor.

intermediate survival CMS4 20 CIN/MSI heterogeneous, mesenchymal/TGF‐β activation, younger age at diagnosis, NOTCH3/VEGFR2 overexpression, worse survival CMS, Colorectal cancer Molecular Subtype; MSI, microsatellite instability; CIN, chromosomal instability; MSS, microsatellite stable; TGF, transforming growth factor. Screening biomarkers A wide variety of biomarkers have been examined18 in colorectal cancer, as both markers of screening and of prognosis. The ideal biomarker would be easily detectable in either stool or blood, cheap and highly accurate. Unfortunately, no current biomarker fits these criteria precisely owing to the molecular heterogeneity associated with colorectal cancer. The most promising markers seem to be associated with abnormal DNA methylation. For example, in colorectal cancer, differential methylation of the septin 9 gene has been shown to have 72 per cent sensitivity and 90 per cent specificity for the detection of malignancy. However, many biomarker studies across all cancer types are plagued by poor study design, insufficient power and non‐hypothesis‐driven marker selection. Currently, a well designed, UK‐based trial of methylated biomarkers, the ENDCaP‐C study (Enhanced Neoplasia Detection and Cancer Prevention in Chronic Colitis) is under way19, examining their value in the detection of dysplasia in a screened population of patients with ulcerative colitis.

hesis‐driven marker selection. Currently, a well designed, UK‐based trial of methylated biomarkers, the ENDCaP‐C study (Enhanced Neoplasia Detection and Cancer Prevention in Chronic Colitis) is under way19, examining their value in the detection of dysplasia in a screened population of patients with ulcerative colitis. Another rich field of developing interest in colorectal cancer is sequencing of the microbial genomes that exist within the colon. Experimental murine models seem to indicate that the microbiome within the colon alters the risk of colorectal cancer by modulating inflammation20. This has also been demonstrated to be the case in patients with inflammatory bowel disease, for both ulcerative colitis and Crohn's disease21. Metabolomic techniques also show promise in acting as screening biomarkers. Mirnezami and co‐workers22 undertook high‐resolution magic‐angle spinning nuclear magnetic resonance spectroscopy in 44 tumour–normal pairs, finding cancer‐specific metabolite patterns allowing differentiation of cancer from normal tissues, in addition to finding changes in the metabolome as the tumour progressed. A number of proteomic studies have also been carried out in colorectal cancer23, although these all suffer from lack of validation, and the variety of different markers identified undoubtedly reflects the varying populations from which they were sampled.

on to finding changes in the metabolome as the tumour progressed. A number of proteomic studies have also been carried out in colorectal cancer23, although these all suffer from lack of validation, and the variety of different markers identified undoubtedly reflects the varying populations from which they were sampled. Targeted therapies, prognostic and predictive biomarkers The discovery that the epidermal growth factor (EGF) pathway in colorectal cancer was sensitive to inhibition by anti‐EGF receptor (EGFR) monoclonal antibodies (mAbs) led to the rapid development of panitumumab and cetuximab. However, it was found in initial trials that the antibodies seemed to have no clinical effect against colorectal cancer; although a proportion of patients seemed to benefit from therapy, the majority did not24. It was found subsequently that mutations in the EGFR pathway genes (KRAS, BRAF, NRAS and PIK3CA) conferred resistance to anti‐EGFR mAbs due to hyperactivation of the pathway independent of the EGFR25. The CRYSTAL trial26 compared patients with a KRAS mutation and those without, finding a clear response and survival benefit for anti‐EGFR mAbs in patients without mutation. A number of other pathway‐specific inhibitors exist for colorectal cancer, including the antivascular endothelial growth factor (VEGF) mAb bevacizumab, MEK inhibitors that target EGF pathway mutated cancers and cancer vaccines. The FOCUS4 trial27 is currently recruiting patients for a molecularly stratified trial of metastatic colorectal cancer therapy: patients are selected for a specific therapy when they possess a mutation specific to that cancer. This raises the intriguing possibility of molecular‐targeted therapy for primary, non‐metastatic tumours as neoadjuvant therapy before surgery.

s for a molecularly stratified trial of metastatic colorectal cancer therapy: patients are selected for a specific therapy when they possess a mutation specific to that cancer. This raises the intriguing possibility of molecular‐targeted therapy for primary, non‐metastatic tumours as neoadjuvant therapy before surgery. A commercially available test exists for prediction of recurrence and benefit for 5‐fluorouracil chemotherapy (Oncotype DX®; Genomic Health, Redwood City, California, USA), based on a multigene panel of RNA expression from formalin‐fixed paraffin‐embedded tissues28. This was developed by screening 761 candidate genes against a large cohort of 1851 patients undergoing surgery for colorectal cancer, with or without adjuvant 5‐fluorouracil therapy. Validation was carried out on the QUASAR study29, which could stratify prognosis, but did not correlate with benefit from chemotherapy.

. This was developed by screening 761 candidate genes against a large cohort of 1851 patients undergoing surgery for colorectal cancer, with or without adjuvant 5‐fluorouracil therapy. Validation was carried out on the QUASAR study29, which could stratify prognosis, but did not correlate with benefit from chemotherapy. Another intriguing possibility is the use of immune‐based stratification to estimate colorectal cancer prognosis. Lal et al.30 used the TCGA expression data set to identify four different immune classifiers based on the expression of immune system‐related genes. Immunogenicity is thought to be related to survival, as tumours that are more visible to the immune system are more likely to undergo destruction by the immune system. One of the mechanisms that may occur in highly mutated tumours, such as tumours with a POLE mutation or those with microsatellite instability, is where the large number of mutations causes a variety of frameshift mutations. These frameshift mutations drive the production of neoantigens31 caused by the alternative splicing of multiple genes, which increases the visibility of the tumour to the immune system. Upper gastrointestinal tract: oesophagogastric cancer Genetic predisposition Research into predisposition to gastro‐oesophageal cancer is complicated by the fact that it can arise in two histologically different epithelial types: squamous cell carcinoma (SCC) and adenocarcinoma. A significant proportion of the risk is likely to comprise lifestyle factors such as smoking, gastro‐oesophageal reflux and diet.

o predisposition to gastro‐oesophageal cancer is complicated by the fact that it can arise in two histologically different epithelial types: squamous cell carcinoma (SCC) and adenocarcinoma. A significant proportion of the risk is likely to comprise lifestyle factors such as smoking, gastro‐oesophageal reflux and diet. In adenocarcinoma, a GWAS of the premalignant stage of oesophageal cancer32, Barrett's oesophagus, demonstrated associations between the major histocompatibility locus and a gene associated with oesophageal development (FOXF1). It was also found that the predisposition to Barrett's oesophagus was made up of multiple common variants of small effect, rather than a single genetic driver. A further GWAS of oesophageal adenocarcinoma32 demonstrated associations with transcription factors (CRTC1, FOXP1, BARX1). In light of these results it is difficult to highlight SNPs that may act as markers for increased disease risk or act as molecular targets for therapy. Several GWAS studies have been undertaken, predominantly in Chinese populations at high risk of oesophageal SCC33, 34. They highlighted SNPs in the riboflavin transporter C20orf54, and a cell growth and differentiation gene, PLCE1. Riboflavin deficiency was identified before this study as a risk factor35 for oesophageal SCC. The PLCE1 variant was further found to interact specifically with tobacco smoke exposure36.

sophageal SCC33, 34. They highlighted SNPs in the riboflavin transporter C20orf54, and a cell growth and differentiation gene, PLCE1. Riboflavin deficiency was identified before this study as a risk factor35 for oesophageal SCC. The PLCE1 variant was further found to interact specifically with tobacco smoke exposure36. Genomic analysis of oesophagogastric cancer Both the Broad Institute (Cambridge, Massachusetts, USA) and the OCCAMS (Oesophageal Cancer Clinical and Molecular Stratification) Consortium have studied oesophageal adenocarcinoma. The Broad Institute project37 undertook whole‐exome and whole‐genome sequencing in 149 tumour–normal pairs, verifying previously identified mutations in TP53, CDKN2A, SMAD4, ARID1A and PIK3CA. Previously unidentified mutations in SPG20, TLR4, ELMO1 and DOCK2 were also found, and a possible role for the RAC1 pathway (a modulator of epithelial–mesenchymal transition) was identified. The OCCAMS Consortium38 examined whole‐genome sequencing of oesophageal adenocarcinoma, as well as targeted sequencing of never‐dysplastic Barrett's oesophagus (which did not progress to malignancy) and high‐grade dysplasia. They found that TP53 was the dominant mutation seen in adenocarcinoma, in over 80 per cent of samples, but that these mutations were also present in biopsies from never‐dysplastic patients, contrary to what was expected based on the known oncological progression of these lesions. The only stage‐specific mutations seen in high‐grade dysplasia and adenocarcinoma were in TP53 and SMAD4.

oma, in over 80 per cent of samples, but that these mutations were also present in biopsies from never‐dysplastic patients, contrary to what was expected based on the known oncological progression of these lesions. The only stage‐specific mutations seen in high‐grade dysplasia and adenocarcinoma were in TP53 and SMAD4. Whole‐genome and whole‐exome sequencing of oesophageal SCC has also been performed39. Whole‐genome sequencing of 17 tumour–normal pairs and whole‐exome sequencing in a further 71 tumour–normal pairs identified recurrent mutations in TP53, RB1, CDKN2A, PIK3CA, NOTCH1 and NFE2L2, as well as ADAM29 and FAM135B. The genomic landscape of oesophageal SCC was significantly different from that of oesophageal adenocarcinoma, highlighting the different therapeutic strategies that are needed in this disease.

our–normal pairs identified recurrent mutations in TP53, RB1, CDKN2A, PIK3CA, NOTCH1 and NFE2L2, as well as ADAM29 and FAM135B. The genomic landscape of oesophageal SCC was significantly different from that of oesophageal adenocarcinoma, highlighting the different therapeutic strategies that are needed in this disease. Biomarkers of predisposition/sensitivity and screening Given the above OCCAMS findings, it is difficult to use TP53 mutation as a biomarker of adenocarcinoma, as it has also been identified in biopsies from never‐dysplastic patients, who should not progress to adenocarcinoma. As a precursor to the OCCAMS study38, the same group undertook combined‐array CGH (comparative genomic hybridization, a type of microarray analysis looking at chromosomal abnormalities) in tumour samples and gene expression via microarray40. They found a pattern of copy number alterations and associated expression change that could identify poor‐prognosis oesophageal adenocarcinoma. The OCCAMS group also examined gene expression changes in adenocarcinoma using RNA microarrays41, finding a four‐gene expression panel of DCK, PAPSS2, SIRT2 and TRIM44 that were independently predictive of survival.

erations and associated expression change that could identify poor‐prognosis oesophageal adenocarcinoma. The OCCAMS group also examined gene expression changes in adenocarcinoma using RNA microarrays41, finding a four‐gene expression panel of DCK, PAPSS2, SIRT2 and TRIM44 that were independently predictive of survival. There have been a number of attempts at developing methylated biomarkers in oesophageal adenocarcinoma, both as screening biomarkers and to identify high‐risk Barrett's oesophagus. The genes studied include CDKN2A, vimentin, P14ARF, CDX2, SOCS1/3, SFRP1/2/4/5 and WIF1 42, 43, 44. Unfortunately no consistent marker can be identified that successfully differentiates adenocarcinoma from normal oesophagus and high‐grade Barrett's dysplasia. These types of focused biomarker will become increasingly important to stratify therapy. Attempts have been made using proteomics45 to distinguish oesophagogastric cancer from benign disease, to allow screening. MALDI mass spectrometry was used to compare the differences between oesophageal cancer and normal mucosa, and gastric cancer and normal mucosa. It was found that, although there were clear differences between cancer and normal tissue, there was a wide variety of changes that varied between different tumours, making determination of a specific biomarker difficult.

he differences between oesophageal cancer and normal mucosa, and gastric cancer and normal mucosa. It was found that, although there were clear differences between cancer and normal tissue, there was a wide variety of changes that varied between different tumours, making determination of a specific biomarker difficult. Targeted therapies Despite the recent advances in discovery science in gastro‐oesophageal cancer, few therapeutic targets currently exist. A small proportion of cancers overexpress the human EGFR 2 (HER2) protein46, and clinical trials of trastuzumab, a mAb against the HER–receptor complex are ongoing. The ToGA (Trastuzumab for Gastric Cancer) study47 investigated the addition of trastuzumab to standard chemotherapy, and demonstrated a small survival benefit in HER2‐positive cancer. Targeting by anti‐VEGF therapy is also undergoing clinical trials48; however, results have been mixed with no improvement in overall survival, but improvements in response rates and progression‐free survival.

ion of trastuzumab to standard chemotherapy, and demonstrated a small survival benefit in HER2‐positive cancer. Targeting by anti‐VEGF therapy is also undergoing clinical trials48; however, results have been mixed with no improvement in overall survival, but improvements in response rates and progression‐free survival. Breast cancer Genetic predisposition The role of inherited variability in breast cancer has been investigated extensively. A very strong signal for variants in the fibroblast growth factor receptor 2 gene (FGFR2) have been found in GWAS studies across multiple populations49, 50, 51. Carriers of the two low‐risk alleles at FGFR2 (frequency 38 per cent of the population) have a relative risk of breast cancer of 0·83 compared with the general population52; carriers of one high‐risk and one low‐risk allele (47 per cent) have a relative risk of 1·05; and carriers of two high‐risk alleles (14 per cent) have a relative risk of 1·2653. FGFR2 mutations are of particular interest as they may represent a therapeutic target in breast cancer54. A recent GWAS55 in oestrogen receptor (ER)‐negative breast cancer demonstrated four variants that reached genome‐wide significance in MDM4, LGR6, FTO and a SNP within the 2p24·1 region. These SNPs were present only in ER‐negative breast cancers, in contrast to the findings in a combined ER‐positive/ER‐negative GWAS56. This found variants located with PTHLH, known to have a role in breast development, and NRIP1, a co‐factor of the ER.

significance in MDM4, LGR6, FTO and a SNP within the 2p24·1 region. These SNPs were present only in ER‐negative breast cancers, in contrast to the findings in a combined ER‐positive/ER‐negative GWAS56. This found variants located with PTHLH, known to have a role in breast development, and NRIP1, a co‐factor of the ER. Next‐generation sequencing A number of NGS projects have examined the mutation spectrum in breast cancer. The most comprehensive of these is from the TCGA project57, which carried out exome sequencing, RNA‐seq, methylation array analysis and RPPA of 463 patients. Recurrent mutations were found in TP53, PIK3CA and GATA3 at frequencies of over 10 per cent, reinforcing their role as driver mutations in breast cancer, as well as mutations in several dozen genes previously identified in breast cancer. Study of the role of expression subtypes in breast cancer demonstrated four separate subtypes (luminal A/B, basal and HER2E), with the mutational burden and spectrum varying in each. The HER2E subtype demonstrated a relatively low mutational frequency, whereas the luminal A subtype demonstrated large numbers of significantly mutated genes, the most frequent mutation being in PIK3CA.

rated four separate subtypes (luminal A/B, basal and HER2E), with the mutational burden and spectrum varying in each. The HER2E subtype demonstrated a relatively low mutational frequency, whereas the luminal A subtype demonstrated large numbers of significantly mutated genes, the most frequent mutation being in PIK3CA. In common with colorectal cancer, methylation array analysis of breast cancer revealed a hypermethylator phenotype, associated with the luminal B expression subtype, and a hypomethylated phenotype associated with a basal expression subtype and comparatively higher frequency of TP53 mutation. Copy number analysis was also performed, with the previously identified amplifications in HER2 and EGFR being identified, as well as novel amplifications in PIK3CA and FOXA1 and deletions in RB1 and PTEN. A striking finding throughout the study was the detection of genetic heterogeneity both within tumours and between samples, highlighting the diverse nature of this disease.

sly identified amplifications in HER2 and EGFR being identified, as well as novel amplifications in PIK3CA and FOXA1 and deletions in RB1 and PTEN. A striking finding throughout the study was the detection of genetic heterogeneity both within tumours and between samples, highlighting the diverse nature of this disease. Shah and colleagues58 examined the mutational spectrum in triple‐negative breast cancer (ER/progesterone receptor/HER‐negative), again finding that TP53 and PIK3CA mutations were clonally dominant, but also finding a wide variety of mutation spectra, including tumours with few driving mutations and tumours with extremely complex mutational spectra (the hypermutated phenotype). A separate study59 identified recurrent mutations in the transcription factor genes CBFB and RUNX1, as well as a gene fusion seen only in triple‐negative breast cancers, the MAGI3–AKT3 fusion transcript. An exome sequencing study60 of 100 breast cancers at the Wellcome Trust Sanger Institute (Cambridge, UK) identified more than 40 driver mutations in a breast cancer cohort, including genes now known to be important therapeutic targets, such as AKT1/2, ARID1B, CASP8 and MAP3K1.

cancers, the MAGI3–AKT3 fusion transcript. An exome sequencing study60 of 100 breast cancers at the Wellcome Trust Sanger Institute (Cambridge, UK) identified more than 40 driver mutations in a breast cancer cohort, including genes now known to be important therapeutic targets, such as AKT1/2, ARID1B, CASP8 and MAP3K1. NGS also has applications in the monitoring of metastatic disease or in recurrence. Dawson et al.61 used a combination of targeted NGS, digital PCR and whole‐genome sequencing to examine DNA circulating in the bloodstream that is shed from metastatic tumours. They found that increasing amounts of circulating DNA correlated with poorer overall survival, but also that, by comparing mutations in the primary tumour with the cell‐free DNA obtained, recurrence of the primary tumour could be detected by the presence of the same somatic mutations in the circulating DNA. This technology has applications for the detection of recurrence in multiple tumour types across the disease spectrum.

t, by comparing mutations in the primary tumour with the cell‐free DNA obtained, recurrence of the primary tumour could be detected by the presence of the same somatic mutations in the circulating DNA. This technology has applications for the detection of recurrence in multiple tumour types across the disease spectrum. Biomarkers of predisposition/sensitivity and screening One‐step nucleic acid amplification (OSNA) is an example of the use of genetic technologies to stratify patients. It works by the detection of raised copy number of the cytokeratin 19 gene (CK19) as a surrogate marker of the presence of breast cancer within sentinel lymph nodes62, and is used as a proxy marker for axillary lymph node positivity in breast cancer. OSNA has been approved as a technological solution to determine sentinel lymph node positivity by the UK National Institute for Health and Care Excellence (NICE), and has been shown to be equivalent to both radioisotope‐ and dye‐based technologies for mapping lymph nodes63. Another recent development is the DNA damage response assay for the prediction of response to anthracycline/cyclophosphamide‐based chemotherapy in breast cancer64. The study examined RNA expression in patients with a DNA damage response deficiency and developed a 44‐gene RNA expression panel that could predict response to chemotherapy in sporadic breast cancer.

e response assay for the prediction of response to anthracycline/cyclophosphamide‐based chemotherapy in breast cancer64. The study examined RNA expression in patients with a DNA damage response deficiency and developed a 44‐gene RNA expression panel that could predict response to chemotherapy in sporadic breast cancer. Targeted therapies A rich variety of targeted therapies exist for breast cancer, based on the underlying biology of the tumour, gained by the extensive molecular research carried out on breast cancer. Aromatase inhibitors (such as anastrazole) and targeted ER modulators (such as tamoxifen) have been used extensively for many years, based on the observation that a proportion of breast cancers are ER‐positive. Identification of the overexpression of HER2 in breast cancer has allowed targeting with both mAbs (trastuzumab) and small‐molecule tyrosine kinase inhibitors such as lapatanib, with appreciable survival benefits. More recently, several novel pathways have been identified as being dysregulated in breast cancer: the insulin‐like growth factor (IGF) 1 pathway; the BRCA‐associated double‐strand break repair protein, poly(ADP‐ribose) polymerase (PARP) 1; and the phosphatidylinositol‐3‐kinase (PIK3)–Akt–mammalian target of rapamycin (mTOR) pathway. Immunohistochemistry of IGF‐1 has revealed that it is overexpressed in more than 50 per cent of breast cancers, and a mAb (cixutumumab) that targets these pathways is currently in phase 1 studies65.

bose) polymerase (PARP) 1; and the phosphatidylinositol‐3‐kinase (PIK3)–Akt–mammalian target of rapamycin (mTOR) pathway. Immunohistochemistry of IGF‐1 has revealed that it is overexpressed in more than 50 per cent of breast cancers, and a mAb (cixutumumab) that targets these pathways is currently in phase 1 studies65. PARP inhibitors were identified from work targeting BRCA mutant breast cancer66 as a therapeutic strategy to treat patients with BRCA mutations. A small molecular PARP inhibitor, olaparib, has been used to treat germline BRCA mutant breast cancer67, with a clinical trial (the OlympiAD trial) currently under way. PARP inhibitors also show promise in triple‐negative breast cancer, as studies have shown that a proportion of these women possess BRCA mutations68, 69 and may respond to PARP inhibitors. mTOR pathway inhibitors, such as everolimus, should theoretically be of benefit in breast cancer as a result of the dysregulation of this pathway demonstrated by molecular studies, although initial results have been disappointing70. The future The union of basic biological research with surgery has allowed the field of translational surgical biology to develop, utilizing modern molecular technologies to stratify surgical disease based on therapeutic and outcome response.

PARP inhibitors were identified from work targeting BRCA mutant breast cancer66 as a therapeutic strategy to treat patients with BRCA mutations. A small molecular PARP inhibitor, olaparib, has been used to treat germline BRCA mutant breast cancer67, with a clinical trial (the OlympiAD trial) currently under way. PARP inhibitors also show promise in triple‐negative breast cancer, as studies have shown that a proportion of these women possess BRCA mutations68, 69 and may respond to PARP inhibitors. mTOR pathway inhibitors, such as everolimus, should theoretically be of benefit in breast cancer as a result of the dysregulation of this pathway demonstrated by molecular studies, although initial results have been disappointing70. The future The union of basic biological research with surgery has allowed the field of translational surgical biology to develop, utilizing modern molecular technologies to stratify surgical disease based on therapeutic and outcome response. A number of possibilities exist for future research. First, whole‐genome studies using NGS technologies of prospectively collected samples will allow identification of biomarkers for response (such as radiotherapy in rectal cancer) and treatment (for instance, molecularly targeted therapies). An example of current clinical issues that could be answered using this technology is the response of rectal cancer to radiotherapy. Around 5–10 per cent of patients with advanced rectal cancer given preoperative radiotherapy have a complete response71. Research is currently under way to understand what makes these tumours particularly sensitive (the so‐called extreme responders) to radiation, and it is likely that a highly focused approach using NGS will identify responsible pathways.

advanced rectal cancer given preoperative radiotherapy have a complete response71. Research is currently under way to understand what makes these tumours particularly sensitive (the so‐called extreme responders) to radiation, and it is likely that a highly focused approach using NGS will identify responsible pathways. Stratification of premalignant lesions is also an important focus of research. In colorectal, upper GI and breast cancer, few biomarkers exist to predict which premalignant lesions will progress to invasive cancer. Typically less than 1 per cent of patients with Barrett's oesophagus will progress to invasive oesophageal cancer, but no markers reliably predict this. The use of retrospective cohorts of patients with progressive Barrett's oesophagus and NGS analysis might identify genetic markers. The concept of disease classifiers will play an increasingly important role and, as more NGS data sets become available, the granularity of classifiers will improve to the extent whereby a more precise understanding of the biology underlying a tumour will be available. This will be enhanced by the integration of multiplatform (NGS, metabolomics, proteomics) data into these classifiers, and will allow tailoring of therapy to the underlying disease.

arity of classifiers will improve to the extent whereby a more precise understanding of the biology underlying a tumour will be available. This will be enhanced by the integration of multiplatform (NGS, metabolomics, proteomics) data into these classifiers, and will allow tailoring of therapy to the underlying disease. The availability of NGS data sets for the surgical population will be enhanced by the commissioning of the UK 100 000 Genomes Project72. This exciting project will carry out whole‐genome sequencing of 50 000 tumour–normal pairs from patients with a range of cancers, but concentrating primarily on colorectal, breast, lung and prostate cancer. The data made available by this project will allow a highly detailed examination of the drivers in these cancer types; this is the largest project of its type in the world. Although the technological advances are numerous, they are not without their own challenges. The recent identification of intratumoral heterogeneity, long hypothesized but only recently identified definitively in renal cell cancer73 and in preneoplastic lesions74, provides unique challenges for stratification. It is likely that multiple subclones of tumour exist within a primary tumour, each with differing characteristics. These will dictate prognosis and response to therapy, and further investigation is needed to understand the extent and consequences of this phenomenon.

ons74, provides unique challenges for stratification. It is likely that multiple subclones of tumour exist within a primary tumour, each with differing characteristics. These will dictate prognosis and response to therapy, and further investigation is needed to understand the extent and consequences of this phenomenon. Technological leaps in molecular biology will enable selection of the right therapy for the right patient at the right time, and further build on the surgical and anaesthetic improvements achieved in the past century to provide maximal benefit for the patient. Acknowledgements A.D.B. is supported by a Wellcome Trust Post‐Doctoral Fellowship for Clinician Scientists (102732/Z/13/Z). A.D.B. undertakes collaborative research with Illumina UK into colorectal cancer; sequencing was provided free of charge by Illumina UK as part of the research project. Disclosure: The authors declare no other conflict of interest.

Introduction The incidence of obesity is rising rapidly across high‐income countries, with the current prevalence in the USA (36 per cent) and UK (26 per cent) expected to double by 20501. It is estimated that up to 66 per cent of patients undergoing surgery in the UK are overweight2. Current evidence is conflicting regarding the impact of obesity on postoperative complications after major surgery. Contemporary multicentre studies3, 4, 5, 6, 7 in specific patient groups from Japan, Denmark, Switzerland and the USA have associated obesity with worse or neutral short‐term postoperative outcomes. A recent risk‐adjusted analysis from the US National Surgical Quality Improvement Program (NSQIP) identified an ‘obesity paradox’ in non‐bariatric surgery, whereby overweight and obese patients had a lower adjusted risk of postoperative mortality8. With at least 600 000 major gastrointestinal operations being carried out each year in the UK, knowing whether obesity increases postoperative complication rates is important for patients, doctors and commissioners9. If increasing body mass is associated with worse outcomes, patients may benefit from perioperative optimization. Depending on the type and timing of surgery, this may include nutritional optimization before surgery, the application of beneficial technology (such as minimally invasive surgery) and access to high‐dependency postoperative care. Firm evidence would provide justification for research to assess these programmes, as there may be unintended consequences, including malnutrition, associated with weight loss.

before surgery, the application of beneficial technology (such as minimally invasive surgery) and access to high‐dependency postoperative care. Firm evidence would provide justification for research to assess these programmes, as there may be unintended consequences, including malnutrition, associated with weight loss. As the variable of interest is body mass index (BMI), randomized trials assigning patients to subgroups of interest (normal weight, overweight and obese) are not possible. Current evidence is thus based on analysis of observational data. A dedicated, prospective analysis in a broad group of patients undergoing elective and emergency surgery with a preplanned, detailed risk adjustment strategy is lacking. This prospective study aimed to determine associations between BMI and postoperative morbidity following elective and emergency major gastrointestinal surgery in the UK and Ireland.

is in a broad group of patients undergoing elective and emergency surgery with a preplanned, detailed risk adjustment strategy is lacking. This prospective study aimed to determine associations between BMI and postoperative morbidity following elective and emergency major gastrointestinal surgery in the UK and Ireland. Methods The protocol for this multicentre prospective cohort study was disseminated through a multinational medical student and trainee surgical collaborative network (with coverage in the UK and Republic of Ireland)10. This network has been described in detail elsewhere11, 12, 13. Briefly, teams of medical students with senior registrar and consultant oversight collected data on consecutive patients across 2‐week periods. Results are reported according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) and Statistical Analyses and Methods in the Published Literature (SAMPL) guidelines14, 15. National Research Ethics Service review of the protocol deemed that full ethical review was not required owing to the observational and anonymous nature of this study. Each participating centre was responsible for local registration as service evaluation or clinical audit. In the Republic of Ireland, each participating centre was responsible for completing the research ethics process at their centre, as required by local guidelines.

e observational and anonymous nature of this study. Each participating centre was responsible for local registration as service evaluation or clinical audit. In the Republic of Ireland, each participating centre was responsible for completing the research ethics process at their centre, as required by local guidelines. Eligibility criteria Consecutive adult patients (aged at least 18 years) undergoing gastrointestinal or hepatobiliary surgery were included in the study. At each centre a minimum 2‐week interval was selected for patient inclusion between 1 October and 12 November 2014; multiple non‐overlapping 2‐week periods per hospital were allowed. Eligible procedures were those involving surgery on any part of the gastrointestinal tract or biliary tree, including a hospital admission with an overnight stay. Both elective and emergency procedures performed using open, laparoscopic, laparoscopically assisted or robotic approaches were included. Patients undergoing day‐case, urological, gynaecological, vascular or transplant procedures were excluded.

act or biliary tree, including a hospital admission with an overnight stay. Both elective and emergency procedures performed using open, laparoscopic, laparoscopically assisted or robotic approaches were included. Patients undergoing day‐case, urological, gynaecological, vascular or transplant procedures were excluded. Outcome measures The Clavien–Dindo system was used to define postoperative complications, whereby complication severity is defined by the subsequent treatment required16. This classification system is a validated means to determine the severity of postoperative complications based on the treatment received. All postoperative events were included, even when there was no direct relationship with the surgery. Each patient's highest Clavien–Dindo grade complication was recorded. The primary outcome measure was the 30‐day major complication rate (Clavien–Dindo grade III–V), which included endoscopic, radiological or surgical reintervention (Clavien‐Dindo III), unexpected critical care admission (Clavien–Dindo IV) and death (Clavien–Dindo V). The secondary outcome was the surgical‐site infection (SSI) rate, chosen as it has been associated with obesity previously6, defined by the Centers for Disease Control definition17. Additional secondary outcome measures listed in the protocol included: outcomes for underweight patients, system‐specific complications, unplanned admission to the critical care unit, reoperation and readmission. For the purposes of clarity, these will be described in detail in further reports.

se Control definition17. Additional secondary outcome measures listed in the protocol included: outcomes for underweight patients, system‐specific complications, unplanned admission to the critical care unit, reoperation and readmission. For the purposes of clarity, these will be described in detail in further reports. Main explanatory variable The main explanatory variable was preoperative BMI, assessed either in the preoperative assessment clinic or on admission. This was calculated as weight (in kilograms) divided by height (in metres) squared. As the primary aim of this study was to assess the effect of being overweight or obese, patients were stratified by BMI according to groups defined by the World Health Organization (WHO): normal weight (BMI 18·5–24·9 kg/m2), overweight (BMI 25·0–29·9 kg/m2) and obese (BMI at least 30·0 kg/m2)18. Explanatory variables Explanatory variables were collected in order to provide a risk‐adjusted estimate. Variables were predefined and selected on the basis of clinical plausibility. The following patient‐ and operation‐level factors were selected. Patient

Main explanatory variable The main explanatory variable was preoperative BMI, assessed either in the preoperative assessment clinic or on admission. This was calculated as weight (in kilograms) divided by height (in metres) squared. As the primary aim of this study was to assess the effect of being overweight or obese, patients were stratified by BMI according to groups defined by the World Health Organization (WHO): normal weight (BMI 18·5–24·9 kg/m2), overweight (BMI 25·0–29·9 kg/m2) and obese (BMI at least 30·0 kg/m2)18. Explanatory variables Explanatory variables were collected in order to provide a risk‐adjusted estimate. Variables were predefined and selected on the basis of clinical plausibility. The following patient‐ and operation‐level factors were selected. Patient To account for co‐morbidities, the American Society of Anesthesiologists (ASA) fitness grade was recorded and the Revised Cardiac Risk Index (RCRI) calculated for each patient. The ASA grade takes into account disease severity and is a reliable metric for the measurement of postoperative mortality and complications19. The RCRI is used to estimate a patient's risk of perioperative cardiac complications, including cardiac death, non‐fatal myocardial infarction and non‐fatal cardiac arrest20. Age, sex and smoking status were also collected. Operation

To account for co‐morbidities, the American Society of Anesthesiologists (ASA) fitness grade was recorded and the Revised Cardiac Risk Index (RCRI) calculated for each patient. The ASA grade takes into account disease severity and is a reliable metric for the measurement of postoperative mortality and complications19. The RCRI is used to estimate a patient's risk of perioperative cardiac complications, including cardiac death, non‐fatal myocardial infarction and non‐fatal cardiac arrest20. Age, sex and smoking status were also collected. Operation A novel Hospital Episode Statistics (HES)‐based operative risk score was devised in order to determine the mortality risk associated with each specific operation. HES are summary, population‐level data available from National Health Service administrative records. From this aggregate data, a summary of mortality rates by procedure during 2009 and 2010 was obtained as the most recent available, procedure‐level data. Procedures included in the study were identified and classified according to their 30‐day mortality rate into predefined strata of low‐risk (less than 1 per cent), moderate‐risk (1– 9·9 per cent) and high‐risk (10 per cent or more) groups (Table S1, supporting information). Additionally, diagnosis (benign versus malignant), urgency of surgery (elective versus emergency) and operative approach (open versus laparoscopic) were included.

ned strata of low‐risk (less than 1 per cent), moderate‐risk (1– 9·9 per cent) and high‐risk (10 per cent or more) groups (Table S1, supporting information). Additionally, diagnosis (benign versus malignant), urgency of surgery (elective versus emergency) and operative approach (open versus laparoscopic) were included. Data accuracy Before data collection, all collaborators were invited to attend investigator meetings and complete a mandatory online training module. To ensure high data quality, only submitted data sets with over 95 per cent completeness for both case ascertainment and the study data fields were eligible for inclusion. If a large proportion of data was missing, data would be imputed using Markov chain Monte Carlo equations. In addition, a process of data validation was performed by independent collaborators. Ten per cent of all included patients were validated independently for accuracy. Twelve key predefined data points were validated for each patient (age, sex, height, weight, Index of Multiple Deprivation decile, Malnutrition Universal Screening Tool score, timing of BMI measurement, urgency of operation, postoperative critical care admission, complications, return to theatre, readmission). A data point was defined as a single value in each of the data fields for an individual patient. The overall accuracy was calculated according to number of correct validated data points divided by the total number of validated data points. The case ascertainment rate was determined by independent review of theatre logbooks for eligible cases, cross‐referenced against the number of actual cases submitted.

patient. The overall accuracy was calculated according to number of correct validated data points divided by the total number of validated data points. The case ascertainment rate was determined by independent review of theatre logbooks for eligible cases, cross‐referenced against the number of actual cases submitted. Data handling Data collection was performed using the secure research electronic capture database (REDCap) system21. The submitted data were then checked centrally and, where missing data were identified, the local investigator was contacted and asked to complete the record. Once vetted, the record was accepted into the data set for analysis. Sample size This study was powered to detect a minimum significant difference between obese and normal‐weight patients, although no upper limit on patient numbers was set owing to the nature of the study. A minimum of 3550 normal‐weight and obese patients would provide 80 per cent power to detect a 2·8 per cent increase in major postoperative complication rate from 8 to 10·8 per cent (α = 0·05). This was based on previous estimates and a complication rate increase that would be deemed clinically significant11.

he study. A minimum of 3550 normal‐weight and obese patients would provide 80 per cent power to detect a 2·8 per cent increase in major postoperative complication rate from 8 to 10·8 per cent (α = 0·05). This was based on previous estimates and a complication rate increase that would be deemed clinically significant11. Statistical analysis There was likely to be considerable selection bias in routine practice that influenced the crude outcome of this study. This is related to individual patient risk factors and the risks of the operation they are undergoing (for example obese patients may have higher risk owing to co‐morbidity; equally they may have better outcome after lower‐risk bariatric surgery). Additionally, there may be bias in how patients are treated across hospitals. Multilevel models were used to determine unbiased distributions of fixed‐effect regression coefficients for the outcomes of major complication rate (primary outcome measure) and SSI (secondary outcome measure). The hospital was considered as a single, level 1 random effect, with individual patient and operative risks entered as level 2 fixed effects. Age was expressed as a continuous variable with corresponding odds ratios (ORs) relating to a per‐year increase. Explanatory variables contained within statistical models were selected based on clinical plausibility and independence, and model selection was informed using the Akaike information criterion. Effect estimates are presented as ORs and bootstrapped 95 per cent confidence intervals, and statistical significance expressed as P values. ORs were generated to describe the relationships between BMI groups and the outcome of interest. An OR greater than 1 represented increased likelihood in the experimental group (obese) versus the control group (normal BMI). The OR describes the relationship between an explanatory variable and an outcome in terms of odds of suffering a complication, rather than the risk. The OR was considered to be statistically significant at the P < 0·050 level. All relevant second‐order interactions were examined. Differences between categorical demographic groups were tested using the Kruskal–Wallis test or Welch's t test for continuous data or χ2 test for proportions. Two‐sided statistical significance was defined at the level of P < 0·050.

gnificant at the P < 0·050 level. All relevant second‐order interactions were examined. Differences between categorical demographic groups were tested using the Kruskal–Wallis test or Welch's t test for continuous data or χ2 test for proportions. Two‐sided statistical significance was defined at the level of P < 0·050. Data analysis was undertaken using R Foundation Statistical software (R 3.2.1) with the Hmisc, ggplot2, plyr, lme4, reshape2, RCurl, splines and stringr packages (R Foundation for Statistical Computing, Vienna, Austria). Results Of 10 272 records submitted from 163 centres, 9264 eligible records were submitted to the final analysis (Fig. 1). Of these, 7965 (86·0 per cent) had a preoperative BMI measurement; 2545 patients (32·0 per cent) were of normal weight, 2673 (33·6 per cent) were overweight and 2747 (34·5 per cent) were obese. Independent validation of 1008 patients with 12 096 data points showed that the accuracy was 98·0 per cent and the case ascertainment rate 92·2 per cent. Figure 1 Flow chart of patient inclusion

Results Of 10 272 records submitted from 163 centres, 9264 eligible records were submitted to the final analysis (Fig. 1). Of these, 7965 (86·0 per cent) had a preoperative BMI measurement; 2545 patients (32·0 per cent) were of normal weight, 2673 (33·6 per cent) were overweight and 2747 (34·5 per cent) were obese. Independent validation of 1008 patients with 12 096 data points showed that the accuracy was 98·0 per cent and the case ascertainment rate 92·2 per cent. Figure 1 Flow chart of patient inclusion BJS-10203-FIG-0001-cDemographics There were significant differences in patient characteristics across BMI groups (Table 1). Patients in the overweight group were older and patients in obese group were younger than those of normal weight; both groups were more likely to undergo lower‐risk surgery and more likely to be operated on using a laparoscopic approach. Obese and overweight patients had higher ASA and RCRI scores, reflecting greater co‐morbidity, but normal‐weight patients underwent proportionally more emergency and higher‐risk procedures. The missing data were minimal and therefore imputation was not required. Table 1 Baseline demographics and operative details by body mass index group Normal (n = 2545) Overweight (n = 2673) Obese (n = 2747) P † Age (years)* 55·3(28·2) 57·8(17·7) 53·3(16·2) < 0·001‡

BJS-10203-FIG-0001-cDemographics There were significant differences in patient characteristics across BMI groups (Table 1). Patients in the overweight group were older and patients in obese group were younger than those of normal weight; both groups were more likely to undergo lower‐risk surgery and more likely to be operated on using a laparoscopic approach. Obese and overweight patients had higher ASA and RCRI scores, reflecting greater co‐morbidity, but normal‐weight patients underwent proportionally more emergency and higher‐risk procedures. The missing data were minimal and therefore imputation was not required. Table 1 Baseline demographics and operative details by body mass index group Normal (n = 2545) Overweight (n = 2673) Obese (n = 2747) P † Age (years)* 55·3(28·2) 57·8(17·7) 53·3(16·2) < 0·001‡ Sex < 0·001 M 1164 (45·7) 1451 (54·3) 1035 (37·7) F 1381 (54·3) 1222 (45·7) 1712 (62·3) ASA fitness grade < 0·001 I 793 (31·2) 780 (29·2) 549 (20·0) II 1106 (43·5) 1282 (48·0) 1441 (52·5) III 522 (20·5) 514 (19·2) 681 (24·8) IV 91 (3·6) 74 (2·8) 57 (2·1) V 17 (0·7) 5 (0·2) 5 (0·2) Missing 16 (0·6) 18 (0·7) 14 (0·5) Smoking status < 0·001 Non‐smoker 1998 (78·5) 2245 (84·0) 2315 (84·3) Current smoker 547 (21·5) 428 (16·0) 432 (15·7) Revised Cardiac Risk Index < 0·001 0 2059 (80·9) 2006 (75·0) 1942 (70·7) I 366 (14·4) 488 (18·3) 607 (22·1) ≥ II 120 (4·7) 177 (6·6) 196 (7·1) Missing 0 (0) 2 (0·1) 2 (0·1) Operative risk class < 0·001 Low 966 (38·0) 1100 (41·2) 1490 (54·2) Moderate 848 (33·3) 883 (33·0) 675 (24·6) High 731 (28·7) 690 (25·8) 582 (21·2) Diagnosis < 0·001 Benign 1803 (70·8) 1843 (68·9) 1796 (65·4) Malignant 742 (29·2) 830 (31·1) 557 (20·3) Bariatric 0 (0) 0 (0) 394 (14·3) Urgency of surgery < 0·001 Elective 1364 (53·6) 1660 (62·1) 1901 (69·2) Emergency 1180 (46·4) 1012 (37·9) 846 (30·8) Missing 1 (0·0) 1 (0·0) 0 (0) Operative approach < 0·001 Open 1163 (45·7) 1094 (40·9) 834 (30·4) Laparoscopic 1379 (54·2) 1578 (59·0) 1911 (69·6) Missing 3 (0·1) 1 (0·0) 2 (0·1) Values in parentheses are percentages unless indicated otherwise;

(53·6) 1660 (62·1) 1901 (69·2) Emergency 1180 (46·4) 1012 (37·9) 846 (30·8) Missing 1 (0·0) 1 (0·0) 0 (0) Operative approach < 0·001 Open 1163 (45·7) 1094 (40·9) 834 (30·4) Laparoscopic 1379 (54·2) 1578 (59·0) 1911 (69·6) Missing 3 (0·1) 1 (0·0) 2 (0·1) Values in parentheses are percentages unless indicated otherwise; * values are mean(s.d.). ASA, American Society of Anesthesiologists. † χ2 test, except ‡ Kruskal–Wallis test. Case mix Overall, malignancy was the most common diagnosis, followed by appendicitis and gallstone disease (Table 2). In the overweight and obese groups, more procedures were undertaken for gallstone disease than in the normal‐weight group. Of the obese patients, 14·3 per cent underwent bariatric surgery. Obese patients were more likely to have laparoscopic surgery (69·6 versus 54·2 per cent of normal‐weight patients), which corresponded with overweight and obese patients undergoing less risky procedures than normal‐weight patients. However, obese patients were still subjected to high‐risk surgery, with 30·8 per cent undergoing emergency surgery and 45·8 per cent a moderate‐ or high‐ risk operative procedure (Table 1). Table 2 Summary of diagnoses by body mass index group

Case mix Overall, malignancy was the most common diagnosis, followed by appendicitis and gallstone disease (Table 2). In the overweight and obese groups, more procedures were undertaken for gallstone disease than in the normal‐weight group. Of the obese patients, 14·3 per cent underwent bariatric surgery. Obese patients were more likely to have laparoscopic surgery (69·6 versus 54·2 per cent of normal‐weight patients), which corresponded with overweight and obese patients undergoing less risky procedures than normal‐weight patients. However, obese patients were still subjected to high‐risk surgery, with 30·8 per cent undergoing emergency surgery and 45·8 per cent a moderate‐ or high‐ risk operative procedure (Table 1). Table 2 Summary of diagnoses by body mass index group Normal (n = 2545) Overweight (n = 2673) Obese (n = 2747) Malignancy 742 (29·2) 830 (31·1) 557 (20·3) Appendicitis 623 (24·5) 508 (19·0) 385 (14·0) All other indications 406 (16·0) 332 (12·4) 245 (8·9) Cholecystitis 299 (11·7) 558 (20·9) 791 (28·8) Inflammatory bowel disease 207 (8·1) 123 (4·6) 77 (2·8) Diverticulitis 47 (1·8) 75 (2·8) 65 (2·4) Peptic ulcer disease 46 (1·8) 30 (1·1) 17 (0·6) Ischaemic bowel 39 (1·5) 29 (1·1) 18 (0·7) Other liver or pancreatic disease 29 (1·1) 47 (1·8) 56 (2·0) Hernia 25 (1·0) 31 (1·2) 38 (1·4) Pancreatitis 24 (0·9) 27 (1·0) 34 (1·2) Faecal perforation 21 (0·8) 18 (0·7) 14 (0·5) Gastro‐oesophageal reflux 20 (0·8) 48 (1·8) 39 (1·4) Fistula 17 (0·7) 17 (0·6) 17 (0·6) Bariatric indication 0 (0) 0 (0) 394 (14·3) Values in parentheses are percentages.

se 29 (1·1) 47 (1·8) 56 (2·0) Hernia 25 (1·0) 31 (1·2) 38 (1·4) Pancreatitis 24 (0·9) 27 (1·0) 34 (1·2) Faecal perforation 21 (0·8) 18 (0·7) 14 (0·5) Gastro‐oesophageal reflux 20 (0·8) 48 (1·8) 39 (1·4) Fistula 17 (0·7) 17 (0·6) 17 (0·6) Bariatric indication 0 (0) 0 (0) 394 (14·3) Values in parentheses are percentages. Outcomes For all patients, the overall 30‐day major (Clavien–Dindo III–V) complication rate was 11·4 per cent (908 of 7965 patients). The overall SSI rate was 5·9 per cent (472 of 7965), increasing to 11·1 per cent (90 of 808) in those undergoing emergency colorectal resection. An unadjusted breakdown of these outcomes by BMI group is shown in Table 3. Table 3 Unadjusted 30‐day complication rates by body mass index group Normal (n = 2545) Overweight (n = 2673) Obese (n = 2747) P * Major complications (Clavien–Dindo III–V) 0·040 No 2238 (87·9) 2351 (88·0) 2468 (89·8) Yes 307 (12·1) 322 (12·0) 279 (10·2) Surgical‐site infection 0·274 No 2410 (94·7) 2507 (93·8) 2576 (93·8) Yes 135 (5·3) 166 (6·2) 171 (6·2) Values in parentheses are percentages. * χ2 test. A significant interaction was found between BMI and malignant/benign diagnosis for major complications (Table 4) and for SSI (Table 5). Taking these interactions into account produced adjusted ORs for each group (Fig. 2). Overweight and obese patients with malignancy were at greater risk of major complications. Obese patients with malignancy were at greater risk of SSI. There were no associations between any BMI groups with a benign diagnosis and either outcome.

se interactions into account produced adjusted ORs for each group (Fig. 2). Overweight and obese patients with malignancy were at greater risk of major complications. Obese patients with malignancy were at greater risk of SSI. There were no associations between any BMI groups with a benign diagnosis and either outcome. Table 4 Univariable and multilevel logistic regression analyses to determine association of 30‐day major postoperative complications (Clavien–Dindo III–V) with patient and operative factors Univariable analysis Multilevel analysis Odds ratio P Odds ratio P

se interactions into account produced adjusted ORs for each group (Fig. 2). Overweight and obese patients with malignancy were at greater risk of major complications. Obese patients with malignancy were at greater risk of SSI. There were no associations between any BMI groups with a benign diagnosis and either outcome. Table 4 Univariable and multilevel logistic regression analyses to determine association of 30‐day major postoperative complications (Clavien–Dindo III–V) with patient and operative factors Univariable analysis Multilevel analysis Odds ratio P Odds ratio P BMI Normal 1·00 (reference) – 1·00 (reference) – Overweight 1·00 (0·85, 1·18) 0·985 0·89 (0·71, 1·12)* 0·329 Obese 0·82 (0·69, 0·98) 0·027 0·84 (0·66, 1·06)* 0·147 Age (per year) 1·02 (1·02, 1·03) < 0·001 1·01 (1·00, 1·01) < 0·001 ASA fitness grade I–II 1·00 (reference) – 1·00 (reference) – III–V 2·82 (2·44, 3·25) < 0·001 1·97 (1·66, 2·33) < 0·001 Diagnosis Benign 1·00 (reference) – 1·00 (reference) – Malignant 1·89 (1·64, 2·18) < 0·001 0·77 (0·58, 1·02) 0·064 Revised Cardiac Risk Index 0 1·00 (reference) – 1·00 (reference) – I 1·68 (1·42, 1·98) < 0·001 1·08 (0·90, 1·31) 0·421 ≥ II 2·21 (1·73, 2·79) < 0·001 1·14 (0·85, 1·46) 0·346 Operative risk class Low 1·00 (reference) – 1·00 (reference) – Moderate 3·67 (3·06, 4·41) < 0·001 3·05 (2·43, 3·79) < 0·001 High 3·52 (2·92, 4·27) < 0·001 2·74 (2·24, 3·40) < 0·001 Sex M 1·00 (reference) – 1·00 (reference) – F 0·66 (0·57, 0·76) < 0·001 0·77 (0·67, 0·89) < 0·001 Smoking status Non‐smoker 1·00 (reference) – 1·00 (reference) – Current smoker 1·00 (0·83, 1·19) 0·971 1·06 (0·86, 1·30) 0·548 Timing of surgery Elective 1·00 (reference) – 1·00 (reference) – Emergency 1·13 (0·98, 1·30) 0·087 1·64 (1·39, 1·94) < 0·001 Interaction variables BMI group by diagnosis Overweight by malignancy – – 1·59 (1·12, 2·29)† 0·008 Obese by malignancy – – 1·91 (1·31, 2·83)† 0·002 Values in parentheses are 95 per cent confidence intervals. An unexpected interaction was identified between body mass index (BMI) and diagnosis. This is included in the multilevel logistic regression model. The interaction can be interpreted as follows:

12, 2·29)† 0·008 Obese by malignancy – – 1·91 (1·31, 2·83)† 0·002 Values in parentheses are 95 per cent confidence intervals. An unexpected interaction was identified between body mass index (BMI) and diagnosis. This is included in the multilevel logistic regression model. The interaction can be interpreted as follows: * the exponentiated coefficients for BMI categories overweight and obese given in the model including the interaction term represent the odds ratios where the diagnosis is benign. † The exponentiated coefficients for the interaction terms can be thought of as a ratio of odds ratios. Akaike information criterion 5164·07. ASA, American Society of Anesthesiologists. Table 5 Univariable and multilevel logistic regression analyses to determine association of surgical‐site infection with patient and operative factors Univariable analysis Multilevel analysis Odds ratio P Odds ratio P

† The exponentiated coefficients for the interaction terms can be thought of as a ratio of odds ratios. Akaike information criterion 5164·07. ASA, American Society of Anesthesiologists. Table 5 Univariable and multilevel logistic regression analyses to determine association of surgical‐site infection with patient and operative factors Univariable analysis Multilevel analysis Odds ratio P Odds ratio P BMI Normal 1·00 (reference) – 1·00 (reference) – Overweight 1·18 (0·94, 1·50) 0·161 1·21 (0·91, 1·57)* 0·155 Obese 1·19 (0·94, 1·50) 0·152 1·19 (0·89, 1·58)* 0·168 Age (per year) 1·01 (1·00, 1·01) 0·007 1·00 (1·00, 1·01) 0·496 ASA fitness grade I–II 1·00 (reference) – 1·00 (reference) – III–V 1·36 (1·10, 1·66) 0·003 0·98 (0·79, 1·22) 0·903 Diagnosis Benign 1·00 (reference) – 1·00 (reference) – Malignant 1·51 (1·24, 1·84) < 0·001 0·88 (0·60, 1·30) 0·430 Revised Cardiac Risk Index 0 1·00 (reference) – 1·00 (reference) – I 1·42 (1·13, 1·78) 0·002 1·24 (0·97, 1·57) 0·083 ≥ II 1·75 (1·24, 2·40) 0·001 1·42 (0·96, 1·96) 0·081 Operative risk class Low 1·00 (reference) – 1·00 (reference) – Moderate 2·23 (1·77, 2·83) < 0·001 2·42 (1·82, 3·21) < 0·001 High 2·48 (1·96, 3·15) < 0·001 2·56 (1·98, 3·31) < 0·001 Sex M 1·00 (reference) – 1·00 (reference) – F 0·76 (0·63, 0·92) 0·005 0·86 (0·72, 1·04) 0·132 Smoking status Non‐smoker 1·00 (reference) – 1·00 (reference) – Current smoker 1·22 (0·96, 1·53) 0·091 1·29 (1·00, 1·63) 0·047 Timing of surgery Elective 1·00 (reference) – 1·00 (reference) – Emergency 1·09 (0·90, 1·31) 0·383 1·47 (1·20, 1·80) < 0·001 Interaction variables BMI group by diagnosis Overweight by malignancy – – 1·04 (0·64, 1·66)† 0·768 Obese by malignancy – – 1·75 (1·05, 2·69)† 0·023 Values in parentheses are 95 per cent confidence intervals. An unexpected interaction was identified between body mass index (BMI) and diagnosis. This is included in the multilevel logistic regression model. The interaction can be interpreted as follows:

64, 1·66)† 0·768 Obese by malignancy – – 1·75 (1·05, 2·69)† 0·023 Values in parentheses are 95 per cent confidence intervals. An unexpected interaction was identified between body mass index (BMI) and diagnosis. This is included in the multilevel logistic regression model. The interaction can be interpreted as follows: * the exponentiated coefficients for BMI categories overweight and obese given in the model including the interaction term represent the odds ratios where the diagnosis is benign. † The exponentiated coefficients for the interaction terms can be thought of as a ratio of odds ratios. Akaike information criterion 3454·58. ASA, American Society of Anesthesiologists. Figure 2 Analysis of the interaction between body mass index (BMI) and diagnosis using recoded variables. Odd ratios are shown with 95 per cent confidence intervals

† The exponentiated coefficients for the interaction terms can be thought of as a ratio of odds ratios. Akaike information criterion 3454·58. ASA, American Society of Anesthesiologists. Figure 2 Analysis of the interaction between body mass index (BMI) and diagnosis using recoded variables. Odd ratios are shown with 95 per cent confidence intervals BJS-10203-FIG-0002-cTo explore the interaction between diagnosis and obesity, a further analysis of obese patients alone was performed (Table 6). This showed that obese patients undergoing surgery for malignancy were subject to higher patient‐ and operation‐level risks than obese patients undergoing surgery for benign conditions. This included older age, higher ASA grades, higher RCRI scores, more open surgery, and a greater proportion of patients undergoing high‐risk surgical procedures (18·3 per cent benign versus 32·7 per cent malignant). A comparison of patients undergoing surgery for malignancy only revealed that those who were obese had greater risk in terms of ASA grade and RCRI score than normal‐weight or overweight patients (Table S2, supporting information). Obese patients undergoing laparoscopic surgery for malignancy were almost twice as likely to require conversion to an open operation than normal‐weight and overweight patients, and obese patients having surgery for a benign condition. Patient characteristics and outcomes within the obese group according to WHO subgroups are described in Table S3 (supporting information). The operative risk class stratification predicted increasing mortality across all three BMI categories (Table S4, supporting information).

nts having surgery for a benign condition. Patient characteristics and outcomes within the obese group according to WHO subgroups are described in Table S3 (supporting information). The operative risk class stratification predicted increasing mortality across all three BMI categories (Table S4, supporting information). Table 6 Characteristics and outcomes of obese patients by benign or malignant diagnosis Benign (n = 2190) Malignant (n = 557) P † Age (years)* 50·2(15·8) 65·4(11·5) < 0·001‡ Sex < 0·001 M 701 (32·0) 334 (60·0) F 1489 (68·0) 223 (40·0) ASA fitness grade < 0·001 I 505 (23·1) 44 (7·9) II 1143 (52·2) 298 (53·5) III 485 (22·1) 196 (35·2) IV 40 (1·8) 17 (3·1) V 5 (0·2) 0 (0·0) Missing 12 (0·5) 2 (0·4) Smoking status 0·007 Non‐smoker 1825 (83·3) 490 (88·0) Current smoker 365 (16·7) 67 (12·0) Revised Cardiac Risk Index < 0·001 0 1624 (74·2) 318 (57·1) I 429 (19·6) 178 (32·0) ≥ II 135 (6·2) 61 (11·0) Missing 2 (0·1) 0 (0) Operative risk class < 0·001 Low 1464 (66·8) 26 (4·7) Moderate 326 (14·9) 349 (62·7) High 400 (18·3) 182 (32·7) Urgency of surgery < 0·001 Elective 1385 (63·2) 516 (92·6) Emergency 805 (36·8) 41 (7·4) Operative approach < 0·001 Open 365 (16·7) 287 (51·5) Open, laparoscopy‐assisted 33 (1·5) 34 (6·1) Laparoscopy 1665 (76·0) 179 (32·1) Laparoscopy converted to open 126 (5·8) 56 (10·1) Missing 1 (0·0) 1 (0·2) Outcomes Major complications (Clavien–Dindo III–V) < 0·001 No 2019 (92·2) 449 (80·6) Yes 171 (7·8) 108 (19·4) Surgical‐site infection < 0·001 No 2081 (95·0) 495 (88·9) Yes 109 (5·0) 62 (11·1) Values in parentheses are percentages unless indicated otherwise;

converted to open 126 (5·8) 56 (10·1) Missing 1 (0·0) 1 (0·2) Outcomes Major complications (Clavien–Dindo III–V) < 0·001 No 2019 (92·2) 449 (80·6) Yes 171 (7·8) 108 (19·4) Surgical‐site infection < 0·001 No 2081 (95·0) 495 (88·9) Yes 109 (5·0) 62 (11·1) Values in parentheses are percentages unless indicated otherwise; * values are mean(s.d.). ASA, American Society of Anesthesiologists. † χ2 test, except ‡ Welch's t test. Discussion This large, prospective study found that overweight and obese patients undergoing surgery for gastrointestinal malignancy were at increased risk of major postoperative complications compared with normal‐weight patients. Overweight and obese patients undergoing surgery for benign conditions did not have higher risks of complications, nor was body mass associated with adverse outcomes in this group. Obese patients were at higher risk of SSI overall, with a weak association for overweight patients.

s compared with normal‐weight patients. Overweight and obese patients undergoing surgery for benign conditions did not have higher risks of complications, nor was body mass associated with adverse outcomes in this group. Obese patients were at higher risk of SSI overall, with a weak association for overweight patients. These findings may be explained by differences in the characteristics of patients undergoing surgery for either a benign or malignant diagnosis. Because of the urgency of their treatment, patients having surgery for malignancy are not readily selected or operations delayed based on fitness. However, those undergoing surgery for benign conditions are likely to be subject to a selection bias of fitter patients for generally lower‐risk procedures. Obese patients underwent far more bariatric and gallbladder operations, leading to a higher number of laparoscopic and elective procedures. In malignant disease, overweight and obese patients had more co‐morbidity than normal‐weight patients, which may have contributed to their increased risk. Less use of laparoscopy and higher rates of conversion in obese patients with malignancy also suggests technical difficulties. Although these risks were adjusted for in the present models, it is possible that the variables included did not fully adjust for the risk to which these patients were subjected.

ncreased risk. Less use of laparoscopy and higher rates of conversion in obese patients with malignancy also suggests technical difficulties. Although these risks were adjusted for in the present models, it is possible that the variables included did not fully adjust for the risk to which these patients were subjected. The main strength of this study is the accurate and comprehensive risk adjustment from a validated prospective data set, which did not rely on administrative or retrospective data. One of the most important clinical factors affecting mortality risk is the magnitude of the operation performed and its timing. To account for this, a novel HES‐based operative risk score was developed, which can be used by future investigators to stratify their own patient groups.

administrative or retrospective data. One of the most important clinical factors affecting mortality risk is the magnitude of the operation performed and its timing. To account for this, a novel HES‐based operative risk score was developed, which can be used by future investigators to stratify their own patient groups. The main limitation of this study is the presence of selection bias between indication for surgery and obesity. Here, morbidly obese patients with benign conditions and co‐morbidities might not have been treated with surgery, as clinicians may have opted for an alternative management plan or weight loss owing to the perceived risk of higher complications. Thus surgery for benign conditions still needs to be balanced against the risks when co‐morbidities are present. Furthermore, patients who received a benign diagnosis may in fact have had malignant disease. In the elective setting, this is highly unlikely, as malignant disease is staged before operation and confirmed histologically before surgery. The group of patients undergoing emergency surgery may have been at higher risk of a misclassified diagnosis as they may not have received preoperative histopathology, but were likely to have had cross‐sectional imaging. Nevertheless, as histopathology reports are usually available within 30 days of operation and therefore the follow‐up period, the risk of this influencing the results was low. Complications occurring after discharge that did not re‐present to the same hospital may not have been detected, but are of uncertain consequence. The 30‐day follow‐up period was selected to ensure that the study was logistically feasible compared with a longer follow‐up period which may have captured additional complications. Despite this, 30‐day follow‐up remains a sensitive measure and is known to correlate well with 90‐day outcomes22. Finally, an independent validation process proved the data to be both highly accurate and complete.

tically feasible compared with a longer follow‐up period which may have captured additional complications. Despite this, 30‐day follow‐up remains a sensitive measure and is known to correlate well with 90‐day outcomes22. Finally, an independent validation process proved the data to be both highly accurate and complete. Previous population‐level studies have suggested an ‘obesity paradox’, whereby overweight and obese patients have lower postoperative complication rates than those of normal weight8. The present findings refute this, as obese patients with malignant disease were at significantly higher risk of major complications. There are four possible explanations for this difference. It may reflect different statistical risk adjustment strategies. Here, the risk of type I errors was limited by fitting models according to prespecified clinically relevant factors. The novel HES‐based risk score provides a high level of quality control for impact of operation, which was lacking in other studies. Other factors were adjusted for including: age, disease severity (ASA grade), urgency of operation and smoking status. A further reason why the present finding has not been observed in previous studies is that there may have been selection bias in other populations, with obese patients being denied surgery or having surgery delayed. It is still reasonable to suggest that there may be other physiological or lifestyle differences between this UK and Irish population compared with previously described North American populations, which remain unaccounted for. Different preoperative optimization practices may also exist, especially with the depth of community care from general practitioners in the UK. Finally, treatment differences in UK and Irish practice that were not captured in this study may be associated with different complications, such as the use of neoadjuvant therapies for cancer.

ative optimization practices may also exist, especially with the depth of community care from general practitioners in the UK. Finally, treatment differences in UK and Irish practice that were not captured in this study may be associated with different complications, such as the use of neoadjuvant therapies for cancer. This large study from the UK and Ireland is globally relevant, with its findings potentially affecting hundreds of thousands of patients undergoing major gastrointestinal surgery every year. Within this context, increased use of laparoscopy may reduce complications and SSI. Therefore, a laparoscopic‐first approach should be adopted routinely for all overweight and obese patients where possible23.

s potentially affecting hundreds of thousands of patients undergoing major gastrointestinal surgery every year. Within this context, increased use of laparoscopy may reduce complications and SSI. Therefore, a laparoscopic‐first approach should be adopted routinely for all overweight and obese patients where possible23. Further research is required to determine cost‐effective methods to reduce perioperative complications in overweight and obese patients with malignancy. In the present study, patients undergoing surgery for malignancy were at an increased risk of complications; therefore this is a specifically high‐risk group. New ways of preoperative optimization of patients to reduce complications, such as preoperative cardiopulmonary exercise, have shown promise in pilot studies. Within the obese and overweight patient population, improving fitness for surgery may prove a useful way of reducing major complications24, 25. In light of the present findings, obese and overweight patients should be a key stakeholder group for the development of these studies. Preoperative weight‐loss programmes may not be an ideal solution, as a poor catabolic nutritional state is associated with worse postoperative outcomes and heightened mortality26. During the operation, the physiological differences found between normal and obese patients should be considered carefully, particularly with regard to fluid management and drug doses tailored toward achieving sufficient tissue concentrations. Intraoperative hypothermia has been shown to increase the risk of SSI27; obese or overweight patients should therefore receive perioperative warming because of their greater surface area. Finally, adequate fluid management in obese or overweight patients is difficult to achieve, and studies investigating fluid management strategies (such as goal‐directed fluid therapy) should assess the effects of BMI on outcomes28. Intensive postoperative care, early mobilization, physiotherapy and novel methods to reduce SSI, above and beyond laparoscopy, should be investigated. Previous research has associated alterations in tissue oxygen tension, the pharmacokinetic distribution of prophylactic antibiotics and impaired immunological response in obese patients with a heightened risk of SSI29, 30, 31. There are considerable additional costs associated with SSI, adding considerable length and costs to inpatient stay32.

ociated alterations in tissue oxygen tension, the pharmacokinetic distribution of prophylactic antibiotics and impaired immunological response in obese patients with a heightened risk of SSI29, 30, 31. There are considerable additional costs associated with SSI, adding considerable length and costs to inpatient stay32. Education, novel wound devices and antibiotic delivery methods may be beneficial in reducing SSI in obese patients with cancer. Collaborators Writing and Steering Group (asterisk indicates joint first author): T. M. Drake*, D. Nepogodiev*, S. J. Chapman*, J. C. Glasbey*, C. Khatri*, C. Y. Kong*, H. A. Claireaux, M. F. Bath, M. Mohan, L. McNamee, M. Kelly, H. Mitchell, J. E. Fitzgerald, E. M. Harrison, A. Bhangu (overall guarantor). Statistical analysis: T. M. Drake, A. Bhangu, E. M. Harrison (statistical guarantor).

Collaborators Writing and Steering Group (asterisk indicates joint first author): T. M. Drake*, D. Nepogodiev*, S. J. Chapman*, J. C. Glasbey*, C. Khatri*, C. Y. Kong*, H. A. Claireaux, M. F. Bath, M. Mohan, L. McNamee, M. Kelly, H. Mitchell, J. E. Fitzgerald, E. M. Harrison, A. Bhangu (overall guarantor). Statistical analysis: T. M. Drake, A. Bhangu, E. M. Harrison (statistical guarantor). Local leads: H. A. Claireaux (University of Bristol, Bristol); I. Antoniou (Hull York Medical School, Hull); R. Dean (Leicester Medical School, Leicester); N. Davies (University of East Anglia, Norwich); S. Trecarten, I. Henderson (University of Nottingham, Nottingham); C. Holmes (University of Sheffield, Sheffield); J. Wylie, R. H. Shuttleworth (Queen's University Belfast, Belfast); A. Jindal (University of Limerick, Limerick); F. Hughes, P. Gouda (National University of Ireland, Galway); L. McNamee (Royal College of Surgeons in Ireland, Dublin); R. Fleck (Trinity College, Dublin); M. Hanrahan (University College Cork, Cork); P. Karunakaran (University College Dublin, Dublin); J. H. Chen, M. C. Sykes (Imperial College, London); R. K. Sethi, S. Suresh (King's College, London); P. Patel, M. Patel (Queen Mary University, London); R. K. Varma, J. Mushtaq (St George's University of London, London); B. Gundogan (University College London, London); W. Bolton (University of Leeds, Leeds); M. Mohan, T. Khan (University of Liverpool, Liverpool); J. Burke, R. Morley (University of Manchester, Manchester); N. Favero (Newcastle University Medical School, Newcastle upon Tyne); R. Adams (University of Aberdeen, Aberdeen); V. Thirumal (University of Dundee, Dundee); E. D. Kennedy (University of Edinburgh, Edinburgh); K. K. Ong, Y. H. Tan (University of Glasgow, Glasgow); J. Gabriel (Brighton and Sussex Medical School, Brighton); A. Bakhsh, J. Y. L. Low (Peninsula, Exeter and Plymouth); A. Yener (Southampton Medical School, Southampton); V. Paraoan (University of Cambridge, Cambridge); R. Preece, T. W. Tilston (Cardiff University, Cardiff); E. Cumber (University of Oxford, Oxford); S. Dean (Swansea University, Swansea); T. Ross, E. McCance (University of Birmingham, Birmingham); H. Amin (Keele University, Keele); L. Satterthwaite (University of Warwick, Coventry).

University of Cambridge, Cambridge); R. Preece, T. W. Tilston (Cardiff University, Cardiff); E. Cumber (University of Oxford, Oxford); S. Dean (Swansea University, Swansea); T. Ross, E. McCance (University of Birmingham, Birmingham); H. Amin (Keele University, Keele); L. Satterthwaite (University of Warwick, Coventry). Other collaborators: K. D. Clement, R. Gratton, E. D. Mills, S. M. Chiu, G. Hung, N. M. Rafiq, J. D. B. Hayes, K. L. Robertson, K. Dynes (Aberdeen Royal Infirmary, Aberdeen); H. C. Huang, S. Assadullah, J. W. Duncumb, R. D. C. Moon, S. X. Poo, J. K. Mehta, K. R. Joshi, R. Callan, J. M. Norris, N. J. Chilvers, H. Keevil, P. Jull (Addenbrooke's Hospital, Cambridge); S. Mallick, D. Elf, L. Carr (Airedale General Hospital, Steeton); C. Player, E. C. Barton, A. L. Martin, S. G. Ratu, E. J. Roberts, P. N. Phan, A. R. Dyal, J. E. Rogers, A. D. Henson (Alexandra Hospital, Redditch); N. B. Reid, D. Burke, G. Culleton, S. Lynne, D. Burke (Antrim Area Hospital, Antrim); S. Mansoor, C. Brennan, R. Blessed, C. Holloway, A. Hill, T. Goldsmith, S. Mackin (Arrowe Park Hospital, Wirral); S. Kim, E. Woin, G. Brent, J. Coffin, O. Ziff (Barnet Hospital, Barnet); Z. Momoh, R. Debenham, M. Ahmed (Basingstoke and North Hampshire Hospital, Basingstoke); C. S. Yong, J. C. Wan, H. C. Copley, P. Raut, F. I. Chaudhry (Bedford Hospital, Bedford); R. H. Shuttleworth, G. Nixon, C. Dorman, R. Tan, S. Kanabar, N. Canning, M. Dolaghan, N. Bell, M. McMenamin (Belfast City Hospital, Belfast); A. Chhabra, K. Duke, L. Turner, T. Patel, L. S. Chew, M. Mirza, S. Lunawat, B. Oremule (Blackpool Victoria Hospital, Blackpool); N. Ward, M. Khan (Bolton Royal Hospital, Bolton); E. T. Tan, D. Maclennan, R. J. McGregor, E. G. Chisholm, E. J. Griffin, L. Bell (Borders General Hospital, Melrose); B. A. Hughes, J. Davies, H. Haq, H. Ahmed, N. Ungcharoen, C. Whacha, R. Thethi (Bradford Royal Infirmary, Bradford); R. M. Markham, A. H. Y. Lee, E. Batt, N. P. Bullock, C. T. Francescon, J. E. Davies, N. M. Shafiq (Bristol Royal Infirmary, Bristol); J. Zhao, S. Vivekanantham, I. Barai, J. L. Y. Allen, D. C. Marshall, C. J. McIntyre, H. C. P. Wilson, A. J. Ashton, C. Lek (Charing Cross Hospital, London); N. Behar, M. Davis‐Hall, N. Seneviratne, S. Kim, L. Esteve, M. Sirakaya, S. Ali, S. Pope, J. S. Ahn, A. Craig‐McQuaide (Chelsea and Westminster Hospital, London); W. A. Gatfield, S. Leong, A. M. Demetri, A. L. Kerr (Cheltenham General Hospital, Cheltenham); C. Rees, J. Loveday, S. Liu, M. Wijesekera, D. Maru, M.

Table S1 Derivation of the operative risk class by mortality rate Table S2 Comparison of normal‐weight, overweight and obese patients undergoing surgery for malignancy Table S3 Comparison across World Health Organization obesity subgroups I, II and III Table S4 Thirty‐day mortality by operative risk class across body mass index groups Click here for additional data file. Acknowledgements The authors thank the consultant surgeons and hospital Trusts who supported and advised STARSurg teams working in their departments. On behalf of the STARSurg group, the co‐authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study guarantor had full access to the data and had the ultimate right of decision when to submit for publication. STARSurg has received INSPIRE funding from the Academy of Medical Sciences and the Wellcome Trust; there was no direct involvement of the funder in this study. The Royal College of Surgeons (England) provided meeting facilities for a collaborator training day, and the Association of Surgeons in Training supported this meeting with a Regional Event Grant. Disclosure: The authors declare no conflict of interest.

ndon); N. Behar, M. Davis‐Hall, N. Seneviratne, S. Kim, L. Esteve, M. Sirakaya, S. Ali, S. Pope, J. S. Ahn, A. Craig‐McQuaide (Chelsea and Westminster Hospital, London); W. A. Gatfield, S. Leong, A. M. Demetri, A. L. Kerr (Cheltenham General Hospital, Cheltenham); C. Rees, J. Loveday, S. Liu, M. Wijesekera, D. Maru, M. Attalla, N. Smith (Chester Hospital, Chester); D. Brown, P. Sritharan, A. Shah, V. Charavanamuttu, G. Heppenstall‐Harris, K. Ng, T. Raghvani, N. Rajan, K. Hulley (Colchester Hospital, Colchester); N. Moody, M. Williams, A. Cotton (Conquest Hospital, Hastings); M. Sharifpour, K. N. Lwin, M. Bright, A. R. Chitnis, M. Abdelhadi, A. D. Semana (County Hospital, Stafford); F. Morgan, R. Reid, J. Dickson, L. Anderson, R. McMullan, J. Dickson, N. Ahern, A. Asmadi, L. B. Anderson (Craigavon Area Hospital, Craigavon); J. Lua Boon Xuan, L. Crozier, S. McAleer (Daisy Hill Hospital, Newry); D. M. Lees, A. A. Adebayo, M. Das, A. H. Amphlett (Derriford Hospital, Plymouth); A. Al‐Robeye, A. Valli, J. Khangura, A. Winarski, A. Ali, J. Khangura (Dewsbury Hospital, Dewsbury); H. Woodward, C. Gouldthrope, M. Turner, K. Sasapu (Diana Princess of Wales Hospital, Grimsby); M. Tonkins, J. R. L. Wild, M. Robinson, J. Hardie, R. Heminway, R. Narramore, N. Ramjeeawon, A. Hibberd (Doncaster Royal Infirmary, Doncaster); F. Winslow, W. Ho, B. F. Chong, K. Lim, S. Ho (Dumfries and Galloway Royal Infirmary, Dumfries); J. A. Crewdson, S. Singagireson, N. Kalra, F. Koumpa, H. Jhala, W. C. Soon, M. Karia, M. G. Rasiah, D. Xylas (Ealing Hospital, Southall); H. Gilbert, M. Sundar‐Singh, J. Wills (Eastbourne District General Hospital, Eastbourne); J. Mushtaq, S. Akhtar, S. Patel, L. Hu, C. Brathwaite‐Shirley, H. Nayee, O. Amin, T. Rangan, E. J. H. Turner (East Surrey Hospital, Redhill); C. McCrann, R. Shepherd, N. Patel, J. Prest‐Smith, E. Auyoung, A. Murtaza, A. Coates (Epsom Hospital, Epsom); O. Prys‐Jones, M. King, S. Gaffney, C. J. Dewdney, I. Nehikhare, J. Lavery (Forth Valley Royal Hospital, Larbert); J. Bassett, K. Davies, K. Ahmad, A. Collins, M. Acres, C. Egerton, T. Khan (Furness General Hospital, Barrow‐in‐Furness); K. Cheng, X. Chen, N. Chan, A. Sheldon, S. Khan, J. Empey, E. Ingram, A. Malik, M. Johnstone (Gartnavel General Hospital, Glasgow); R. Goodier, J. P. Shah, J. E. Giles, J. A. Sanders, S. W. McLure, S. Pal, A. Rangedara, A. N. Baker, C. A. Asbjoernsen (George Eliot Hospital, Nuneaton); C. Girling, L. Gray, L. Gauntlett, C. Joyner, S. Qureshi, S.

, A. Sheldon, S. Khan, J. Empey, E. Ingram, A. Malik, M. Johnstone (Gartnavel General Hospital, Glasgow); R. Goodier, J. P. Shah, J. E. Giles, J. A. Sanders, S. W. McLure, S. Pal, A. Rangedara, A. N. Baker, C. A. Asbjoernsen (George Eliot Hospital, Nuneaton); C. Girling, L. Gray, L. Gauntlett, C. Joyner, S. Qureshi, S. Dean (Glangwili General Hospital, Carmarthen); Y. P. Mogan, J. C. K. Ng, A. N. Kumar, J. H. Park, D. Tan, K. P. Choo, K. P. Raman, P. Buakuma, C. Xiao, S. Govinden (Glasgow Royal Infirmary, Glasgow); O. D. Thompson, M. A. Charalambos, E. Brown, R. B. Karsan, T. Dogra, L. M. Bullman, P. M. Dawson (Gloucestershire Royal Hospital, Gloucester); A. L. Frank, H. Abid, L. Tung, U. Qureshi, A. Tahmina, B. W. Matthews (Good Hope Hospital, Sutton Coldfield); R. T. Harris, A. O'Connor, K. Mazan, S. Iqbal, S. A. Stanger, J. D. Thompson (Great Western Hospital, Swindon); J. A. L. Sullivan, E. Uppal, A. MacAskill, F. A. Bamgbose, C. Neophytou, A. F. Carroll, C. W. Rookes, U. Datta, A. J. Dhutia (Hammersmith Hospital, Hammersmith); S. Rashid, N. Ahmed, T. Lo (Harrogate District Hospital, Harrogate); S. Bhanderi, C. D. Blore, S. Ahmed, H. Shaheen, S. Abburu, S. Majid, Z. Abbas, S. S. Talukdar, S. Ahmed (Heartlands Hospital, Birmingham); L. J. Burney, J. B. Patel, O. Al‐Obaedi, A. W. Roberts, O. Al‐Obaedi, S. Mahboob (Hereford County Hospital, Hereford); B. Singh, S. Sheth, P. Karia, A. Prabhudesai, K. Kow, K. Koysombat, S. Wang, P. Morrison, Y. Maheswaran, P. Keane (Hillingdon Hospital, Uxbridge); P. C. Copley, O. Brewster, G. X. Xu, P. Harries, C. Wall (Hinchingbrooke Hospital, Huntingdon); A. Al‐Mousawi, S. Bonsu, P. Cunha, T. Ward, J. Paul, K. Nadanakumaran, S. Tayeh, T. Ward, H. Holyoak, J. Remedios, K. Theodoropoulou, T. Ward (Homerton University Hospital, London); A. Luhishi, L. Jacob, F. Long, A. Atayi, S. Sarwar, O. Parker (Huddersfield Royal Infirmary, Huddersfield); J. Harvey, H. Ross, R. Rampal, G. Thomas, P. Vanmali, C. McGowan, J. Stein (Hull Royal Infirmary, Hull, and Castle Hill Hospital, Cottingham); V. Robertson, L. Carthew, V. Teng, J. Fong (Inverclyde Royal Hospital, Greenock); A. N. Street, C. E. Thakker (Ipswich Hospital, Ipswich); D. O'Reilly, M. Bravo, A. Pizzolato, H. A. Khokhar, M. Ryan, L. Cheskes, R. Carr, A. E. Salih (James Connolly Hospital, Dublin); S. Bassiony, R. Yuen, D. Chrastek, H. Rosen O'Sullivan, A. Amajuoyi, A. Wang, O. Sitta, J. Wye (James Paget University Hospital, Great Yarmouth); M. A. Qamar, C. Major, A.

Hospital, Ipswich); D. O'Reilly, M. Bravo, A. Pizzolato, H. A. Khokhar, M. Ryan, L. Cheskes, R. Carr, A. E. Salih (James Connolly Hospital, Dublin); S. Bassiony, R. Yuen, D. Chrastek, H. Rosen O'Sullivan, A. Amajuoyi, A. Wang, O. Sitta, J. Wye (James Paget University Hospital, Great Yarmouth); M. A. Qamar, C. Major, A. Kaushal (Kent and Canterbury Hospital, Canterbury); C. Morgan, M. Petrarca, R. Allot, K. Verma, S. Dutt, R. Allot, C. P. Chilima, S. Peroos, R. Allot (King George Hospital, Ilford); S. R. Kosasih, H. Chin, L. Ashken, R. J. Pearse R. A. O'Loughlin, A. Menon, K. Singh, J. Norton (King's College Hospital, London); R. Sagar, N. Jathanna, L. Rothwell, N. Watson, F. Harding, P. Dube (King's Mill Hospital, Sutton‐in‐Ashfield); H. Khalid, N. Punjabi, M. Sagmeister, P. Gill, S. Shahid, S. Hudson‐Phillips, D. George, J. Ashwood, T. Lewis (Kingston Hospital, Kingston upon Thames); M. Dhar, P. Sangal, I. A. Rhema, D. Kotecha, R. Dean, Z. Afzal, J. A. Syeed, E. Prakash, P. Jalota, R. Dean (Leicester Royal Infirmary, Leicester); J. Herron, L. Kimani, A. Delport, A. Shukla (Lincoln County Hospital, Lincoln); V. Agarwal, S. Parthiban, H. Thakur, W. Cymes, S. Rinkoff (Lister Hospital, Stevenage); J. A. Turnbull, M. Hayat, S. Darr, U. Khan, J. Lim, A. Higgins (Manchester Royal Infirmary, Manchester); G. Lakshmipathy, B. Forte, E. Canning, A. Jaitley, J. Lamont, E. Toner, A. Ghaffar, M. McDowell, D. Salmon (Mater Infirmorum Hospital, Belfast); P. Gouda, O. O'Carroll, A. Khan, M. E. Kelly, K. Clesham, C. Palmer, R. Lyons, M. E. Kelly, A. Bell, R. Chin, R. M. Waldron, M. E. Kelly (Mayo General Hospital, Castlebar); A. Trimble, S. E. Cox, U. Ashfaq, J. Campbell, R. B. S. Holliday, G. McCabe (Monklands Hospital, Airdrie); F. Morris, R. Priestland, S. Dean, O. K. Vernon, A. Ledsam, R. Vaughan (Morriston Hospital, Swansea); D. Lim, Z. R. Bakewell, R. K. Hughes (Musgrove Park Hospital, Taunton); R. M. Koshy, H. R. Jackson, P. Narayan, A. E. Cardwell, C. L. Jubainville, T. Arif, L. E. Elliott, V. Gupta, T. Arif (New Cross Hospital, Wolverhampton); G. Bhaskaran, K. Singh, A. Odeleye, F. Ahmed, R. Shah, A. Odeleye, J. Pickard, Y. N. Suleman, A. Odeleye (Newham University Hospital, London); A. S. North, L. F. McClymont, N. Hussain, I. Ibrahim, G. S. Ng, V. Wong, A. E. Lim, L. N. Harris, T. Tharmachandirar, D. Mittapalli (Ninewells Hospital, Dundee); V. Patel, M. Lakhani (Nobles Hospital, Isle of Man); N. Davies, H. Z. Bazeer, V. Narwani, K. K. Sandhu, L. R. Wingfield, S. Gentry, H. Adjei, M.

tal, London); A. S. North, L. F. McClymont, N. Hussain, I. Ibrahim, G. S. Ng, V. Wong, A. E. Lim, L. N. Harris, T. Tharmachandirar, D. Mittapalli (Ninewells Hospital, Dundee); V. Patel, M. Lakhani (Nobles Hospital, Isle of Man); N. Davies, H. Z. Bazeer, V. Narwani, K. K. Sandhu, L. R. Wingfield, S. Gentry, H. Adjei, M. Bhatti, L. Braganza (Norfolk and Norwich University Hospital, Norwich); J. Barnes, S. Mistry, G. Chillarge, S. Stokes, J. Cleere, S. Wadanamby, A. M. Bucko, J. Meek, N. Boxall, E. G. Heywood, J. J. Wiltshire, C. Toh, A. E. Ward, B. N. Shurovi, T. M. Drake (Northern General Hospital, Sheffield); D. Horth, B. Y. Patel, B. Ali, T. Spencer, T. Axelson, L. Kretzmer, C. Chhina (North Manchester General Hospital, Manchester); C. Anandarajah, T. Fautz, C. Horst (North Middlesex University Hospital NHS Trust, Edmonton); A. A. Thevathasan, J. Q. Ng, F. Hirst (North Tyneside General Hospital, North Shields); C. F. Brewer, A. E. Logan, J. W. Lockey, P. R. Forrest, N. Keelty, A. D. Wood, L. R. Springford, P. Avery, T. M. Schulz, T. P. Bemand, L. Howells (Northwick Park Hospital, Harrow); H. Collier, A. Khajuria, R. G. Tharakan, S. Parsons (Nottingham City Hospital, Nottingham); A. M. Buchan, R. J. McGalliard, J. D. Mason, O. J. Cundy, N. Li, N. A. Redgrave, R. P. Watson, T. P. Pezas, Y. F. Dennis, E. Segall, M. Hameed, A. S. Lynch (John Radcliffe and Churchill Hospitals, Oxford); M. Chamberlain, F. S. Peck, Y. N. Neo, G. Russell, M. Elseedawy, S. Lee, N. L. Foster, Y. H. Soo, L. Puan (Perth Royal Infirmary, Perth); R. Dennis, H. Goradia, A. Qureshi (Peterborough City Hospital, Peterborough); S. Osman, T. Reeves, L. Dinsmore, M. Marsden, Q. Lu, T. Pitts‐Tucker (Portsmouth Hospitals NHS Trust Queen Alexandra Hospital, Portsmouth); C. E. Dunn, R. A. Walford, E. Heathcote, R. Martin, A. Pericleous, K. Brzyska, K. G. Reid, M. R. Williams, N. Wetherall (Prince Charles Hospital, Merthyr Tydfil); E. McAleer, D. Thomas, R. Kiff, C. Gouldthrope (Princess of Wales Hospital, Bridgend); S. Milne, M. J. V. Holmes, S. Stokes, J. Bartlett, J. Lucas de Carvalho, T. Bloomfield (Princess Royal Hospital, Haywards Heath); F. Tongo, R. H. Bremner N. Yong, B. A. Atraszkiewicz, A. Mehdi, M. Tahir, G. X. J. Sherliker, A. K. Tear, A. Pandey (Princess Royal University Hospital, Bromley); A. Broyd, H. M. Omer, M. Raphael, W. W. Chaudhry, S. Shahidi, A. S. Jawad, C. K. Gill, I. Hindle Fisher, I. Adeleja, I. J. Clark, G. E. Aidoo‐Micah (Queen Elizabeth Hospital, Birmingham); P. W. Stather, G. J.

z, A. Mehdi, M. Tahir, G. X. J. Sherliker, A. K. Tear, A. Pandey (Princess Royal University Hospital, Bromley); A. Broyd, H. M. Omer, M. Raphael, W. W. Chaudhry, S. Shahidi, A. S. Jawad, C. K. Gill, I. Hindle Fisher, I. Adeleja, I. J. Clark, G. E. Aidoo‐Micah (Queen Elizabeth Hospital, Birmingham); P. W. Stather, G. J. Salam, T. E. Glover, G. Deas, N. K. Sim, R. D. Obute, W. M. Wynell‐Mayow (Queen Elizabeth Hospital, King's Lynn); M. S. Sait, N. Mitha, G. L. de Bernier, M. Siddiqui, R. Shaunak, A. Wali, G. Cuthbert (Queen Elizabeth Hospital, Woolwich); R. Bhudia, E. Webb, S. Shah, N. Ansari, M. Perera, N. Kelly (Queen's Hospital, Romford); R. McAllister, G. H. Stanley, C. P. Keane, V. Shatkar, C. Maxwell‐Armstrong (Queen's Medical Centre, Nottingham); L. A. Henderson, N. Maple, R. Manson, R. D. Adams, E. Brown, E. Semple (Raigmore Hospital, Inverness); M. Mills, A. Daoub, A. Marsh, A. Ramnarine, J. Hartley, M. Malaj (Rotherham General Hospital, Rotherham); P. D. Jewell, E. A. Whatling, N. Hitchen, M. Chen (Royal Berkshire Hospital, Reading); B. Goh, J. Fern, S. Rogers, L. Derbyshire (Royal Blackburn Hospital, Blackburn); D. T. Robertson, N. Abuhussein, P. Deekonda, A. Abid, A. Bakhsh (Royal Devon and Exeter Hospital, Exeter); P. L. M. Harrison, L. Aildasani, H. Turley (Royal Glamorgan Hospital, Ynysmaerdy); M. A. Sherif, G. Pandey, J. J. Filby, A. Johnston, E. Burke, M. Mohamud, K. Gohil, A. Y. Tsui, R. Singh (Royal Gwent Hospital, Newport); S. J. Lim, K. O'Sullivan, L. L. McKelvey, S. O'Neill, H. F. Roberts, F. S. Brown, Y. Cao, R. T. Buckle, Y. Liew, S. Sii, C. M. Ventre, C. J. Graham, T. Filipescu (Royal Infirmary of Edinburgh, Edinburgh); A. Yousif, R. Dawar, A. Wright, M. Peters, R. Varley, S. Owczarek, S. Hartley (Royal Lancaster Infirmary, Lancaster); M. Khattak, A. Iqbal, M. Ali, B. Durrani, Y. Narang, G. S. Bethell, L. Horne, R. Pinto (Royal Liverpool Hospital, Liverpool); K. Nicholls, I. Kisyov, H. D. Torrance, P. Patel, M. Patel, W. English, S. M. Lakhani, S. F. Ashraf, M. Venn (Royal London Hospital, London); V. Elangovan, Z. Kazmi, J. Brecher, S. Sukumar, A. Mastan, A. Mortimer, J. Parker, J. Boyle (Royal Preston Hospital, Preston); M. Elkawafi, J. Beckett, A. Mohite, A. Narain, E. Mazumdar, A. Sreh, A. Hague, D. Weinberg L. Fletcher (Royal Shrewsbury Hospital, Shrewsbury); M. Steel, H. Shufflebotham, M. Masood, Y. Sinha, H. Amin, C. Jenvey, H. Kitt, R. Slade, A. R. Craig, C. Deall, Y. Sinha (Royal Stoke University Hospital, Stoke‐on‐Trent); J. Gabriel, T.

w intervention(s). The By‐Band‐Sleeve study also includes a ‘non‐randomized’ cohort of eligible patients who consent to data collection and follow‐up but not to randomization. Analysis of this cohort will provide important insights into how the trial has been integrated into each clinical practice at the study centres. One challenge to overcome when a study changes from two to three groups is to ensure that trial processes are not disrupted during the transition and that patients continue to be recruited and followed up in accordance with the protocol. For example, some patients were consented to a two‐group trial but, by the time they were randomized, the trial had switched to three groups. It was not feasible to have a further consultation to discuss the trial again, so the randomization for these patients was handled centrally by the trials unit to avoid them being allocated to the new procedure.

fi, J. Beckett, A. Mohite, A. Narain, E. Mazumdar, A. Sreh, A. Hague, D. Weinberg L. Fletcher (Royal Shrewsbury Hospital, Shrewsbury); M. Steel, H. Shufflebotham, M. Masood, Y. Sinha, H. Amin, C. Jenvey, H. Kitt, R. Slade, A. R. Craig, C. Deall, Y. Sinha (Royal Stoke University Hospital, Stoke‐on‐Trent); J. Gabriel, T. Reakes, J. Chervenkoff, E. Strange, M. O'Bryan, C. Murkin, D. Joshi, E. Strange, T. Bergara, S. Naqib, D. Wylam, E. Strange (Royal Sussex County Hospital, Brighton); S. E. Scotcher, C. M. Hewitt, M. T. Stoddart, A. Kerai, A. J. Trist, S. J. Cole, C. L. Knight, S. Stevens, G. E. Cooper (Royal United Hospital, Bath); R. Ingham, J. Dobson, J. Wylie, A. O'Kane, J. Moradzadeh, A. Duffy, C. Henderson, S. Ashraf, C. McLaughin (Royal Victoria Hospital, Belfast); T. C. Hoskins, R. S. Reehal, L. R. Bookless, R. C. McLean, E. J. Stone (Royal Victoria Infirmary, Newcastle upon Tyne); E. V. Wright, H. R. Abdikadir, C. Roberts, O. Spence, M. Srikantharajah, M. Patel, E. M. Ruiz, J. H. Matthews, E. Gardner, C. Roberts (Russell's Hall Hospital, Dudley); E. Hester, P. Naran, R. Simpson, M. Minhas, E. Cornish, S. A. Semnani, D. Rojoa (Salford Royal Hospital, Salford); A. Radotra, J. Eraifej, K. Eparh, D. N. E. Smith, B. D. Mistry, S. L. Hickling, A. Bhangu (Sandwell General Hospital, West Bromwich); W. Din, C. Liu, P. Mithrakumar, V. Mirdavoudi (Scarborough Hospital, Scarborough); M. Rashid, C. Mcgenity, O. Hussain, M. Kadicheeni, H. Gardner, N. Anim‐Addo, J. Pearce, A. Aslanyan, C. Ntala, T. Sorah, J. Parkin, M. Alizadeh (Scunthorpe Hospital, Scunthorpe); A. White, F. Edozie, J. Johnston, A. Kahar, V. Navayogaarajah, B. Patel, D. Carter, P. Khonsari, A. Burgess, B. Patel (Southend Hospital, Westcliff‐on‐Sea); C. Kong, A. Ponweera, A. Cody, Y. Tan, A. Y. L. Ng, A. Croall, C. Allan, S. Ng, V. Raghuvir (Southern General Hospital, Glasgow); R. Telfer, A. D. Greenhalgh, C. N. McKerr, M. A. Edison, B. A. Patel, K. Dear, M. R. Hardy (Southmead Hospital, Bristol); P. Williams, S. Hassan, U. Sajjad (Southport Hospital, Southport); E. M. O'Neill, S. Lopes, L. Healy (South Tipperary Hospital, Tipperary); N. Jamal, S. Tan, D. Lazenby, S. B. Husnoo, S. Beecroft, T. Sarvanandan (Stepping Hill Hospital, Stockport); C. Weston, N. Bassam, S. Rabinthiran, U. Hayat, L. Ng, D. Varma (St George's Hospital, London); M. Sukkari, A. Mian, A. Coates, A. Omar, J. W. Kim, A. Coates, J. Sellathurai, J. Mahmood (St Helier Hospital, Carshalton); C. O'Connell, R. Bose, H. Heneghan, P. Lalor, J. Matheson, C.

n (Stepping Hill Hospital, Stockport); C. Weston, N. Bassam, S. Rabinthiran, U. Hayat, L. Ng, D. Varma (St George's Hospital, London); M. Sukkari, A. Mian, A. Coates, A. Omar, J. W. Kim, A. Coates, J. Sellathurai, J. Mahmood (St Helier Hospital, Carshalton); C. O'Connell, R. Bose, H. Heneghan, P. Lalor, J. Matheson, C. Doherty, C. Cullen, D. Cooper, S. Angelov, C. Drislane (St James' Hospital, Dublin); A. C. D. Smith, A. Kreibich, E. Palkhi, A. Durr, A. Lotfallah, D. Gold, E. Mckean, A. Durr, A. Dhanji, A. Anilkumar, A. Thacoor, A. Durr (St James's University Hospital, Leeds); Z. H. Siddiqui, S. Lim, A. Piquet, S. M. Anderson, A. Jindal, D. R. McCormack, J. Gulati, A. Ibrahim, A. Jindal, S. E. Murray, S. L. Walsh, A. McGrath (St Luke's Hospital, Limerick); P. Ziprin, E. Y. Chua, C. N. Lou, J. Bloomer, H. R. Paine, D. Osei‐Kuffour, C. J. White, A. Szczap, S. Gokani, K. Patel (St Mary's Hospital, London); M. K. Malys, A. Reed, G. E. Torlot, E. M. Cumber (Stoke Mandeville Hospital, Aylesbury, and High Wycombe Hospital, High Wycombe); A. Charania, S. Ahmad, N. Varma, H. Cheema, L. Austreng, H. Petra, M. Chaudhary (St Peter's Hospital, Chertsey); M. I. Zegeye, F. Cheung, D. Coffey, R. S. Heer, S. Singh, E. Seager, S. Cumming, R. S. Suresh, S. Verma, I. B. Ptacek, A. M. Gwozdz (St Thomas' Hospital, London); T. Yang, A. A. Khetarpal, S. Shumon, T. M. P. Fung, W. Leung, P. Kwang, L. Chew, W. Loke, A. Curran (Sunderland Royal Hospital, Sunderland); C. Chan (Tameside General Hospital, Ashton‐under‐Lyne); C. McGarrigle, K. Mohan, S. Cullen, E. Wong, C. Toale, D. Collins, N. Keane, B. P. Traynor, D. Shanahan (The Adelaide and Meath Hospital, Dublin); A. Yan, D. J. Jafree, C. Topham S. Mitrasinovic, S. Omara, B. Gundogan, G. Bingham, P. M. Lykoudis, B. H. Miranda (The Royal Free Hospital, London); K. Whitehurst, G. Kumaran, Y. Devabalan, H. Aziz, M. Shoa, S. Dindyal (The Whittington Hospital, London); J. A. Yates, I. Bernstein, G. Rattan, J. A. Yates (Tullamore Hospital, Tullamore); R. Coulson, S. Stezaker, A. Isaac, M. Salem, A. McBride, A. Isaac, H. McFarlane, L. Yow, J. MacDonald (Ulster Hospital, Belfast); R. D. Bartlett, S. Turaga, U. White, W. Liew, N. Yim (University College London Hospital, London); A. Ang, A. Simpson, D. McAuley, E. Craig, L. Murphy, P. Shepherd, J. Y. Kee, A. Abdulmajid, A. Chung (University Hospital Ayr, Ayr); H. L. Warwick, A. Livesey, P. Holton, M. D. Theodoreson, S. L. Jenkin, J. Turner, J. H. Entwisle, S. T. Marchal, S. O'Connor, H. K.

w, N. Yim (University College London Hospital, London); A. Ang, A. Simpson, D. McAuley, E. Craig, L. Murphy, P. Shepherd, J. Y. Kee, A. Abdulmajid, A. Chung (University Hospital Ayr, Ayr); H. L. Warwick, A. Livesey, P. Holton, M. D. Theodoreson, S. L. Jenkin, J. Turner, J. H. Entwisle, S. T. Marchal, S. O'Connor, H. K. Blege (University Hospital Coventry and Warwickshire, Coventry); J. M. Aithie, L. M. Sabine, G. E. Stewart, S. Jackson (University Hospital Crosshouse, Crosshouse); A. Kishore, C. M. Lankage, F. Acquaah, H. L. Joyce (University Hospital Lewisham, Lewisham); A. Jindal, K. L. McKevitt, C. J. Coffey, A. S. Fawaz, K. S. Dolbec, D. A. O'Sullivan, J. M. Geraghty (University Hospital Limerick, Limerick); E. Lim, L. Bolton, D. FitzPatrick, C. Robinson, T. Ramtoola, S. Collinson (University Hospital of South Manchester, Manchester); L. Grundy, P. M. McEnhill, G. S. Harbhajan Singh, D. Loughran, D. M. Golding, R. E. Keeling, R. P. Williams, R. D. J. Whitham, S. Yoganathan, R. Nachiappan, R. J. Egan (University Hospital of Wales, Cardiff); R. Owasil, M. L. Kwan, A. He, R. W. Goh, R. Bhome, H. Wilson, P. J. Teoh, K. Raji, T. Reeves, N. Jayakody, J. Matthams (University Hospital Southampton NHS Foundation Trust, Southampton); J. Chong, S. Tan, C. Y. Luk, R. J. Greig, M. Trail, G. Charalambous, V. Thirumal, A. S. Rocke, N. Gardiner (Victoria Hospital, Kirkcaldy); F. Bulley, N. Warren, E. Brennan, P. Fergurson, R. Wilson, H. Whittingham (Victoria Infirmary, Glasgow); E. J. Brown, R. Khanijau, K. Gandhi, S. Morris, A. J. Boulton, N. Chandan, A. E. Barthorpe, R. Maamari, S. Sandhu (Walsall Manor Hospital, Walsall); M. McCann, L. Higgs, V. Balian, C. Reeder, C. Diaper, V. Balian, T. Sale, H. Ali, V. Balian (Warrington Hospital, Warrington); C. H. Archer, A. K. Clarke, J. Heskin, P. C. Hurst, J. D. Farmer, L. D. O'Flynn, L. Doan, B. A. Shuker, G. D. Stott (Warwick Hospital, Warwick); N. A. Vithanage, K. A. Hoban, P. N. Nesargikar, H. R. Kennedy, C. M. Grossart, E. S. M. Tan, C. S. D. Roy, P. Sim, K. E. Leslie (Western General Hospital, Edinburgh); D. Sim, M. H. Abul, N. Cody, A. Y. Tay, E. Woon, S. Sng, J. Mah, J. Robson (Western Infirmary, Glasgow); E. Shakweh, V. C. Wing, H. Mills, M. M. Li, T. R. Barrow, S. Balaji, H. E. M. Jordan, C. Phillips, H. Naveed (West Middlesex Hospital, Isleworth); S. Hirani, A. Tai, R. Ratnakumaran, A. Sahathevan, A. M. A. Shafi, M. Seedat, R. Weaver (Whipps Cross Hospital, London); A. Batho, R. Punj, H. Selvachandran, N. Bhatt, S. Botchey, Z.

h, V. C. Wing, H. Mills, M. M. Li, T. R. Barrow, S. Balaji, H. E. M. Jordan, C. Phillips, H. Naveed (West Middlesex Hospital, Isleworth); S. Hirani, A. Tai, R. Ratnakumaran, A. Sahathevan, A. M. A. Shafi, M. Seedat, R. Weaver (Whipps Cross Hospital, London); A. Batho, R. Punj, H. Selvachandran, N. Bhatt, S. Botchey, Z. Khonat, K. Brennan (Whiston Hospital, Prescot); K. K. Ong, C. J. Morrison, E. Devlin, A. Linton, E. Galloway, S. McGarvie, N. Ramsay (Wishaw General Hospital, Wishaw); H. D. McRobbie, H. Whewell, W. Dean (Wrexham Maelor Hospital, Wrexham); S. Nelaj, M. Eragat, A. Mishra, T. Kane, M. Zuhair, M. Wells, D. Wilkinson, N. Woodcock (York Hospital, York); E. Sun, N. Aziz, M. K. Abd Ghaffar (Ysbyty Gwynedd, Bangor). Supporting information Additional supporting information may be found in the online version of this article: Table S1 Derivation of the operative risk class by mortality rate (Word document) Table S2 Comparison of normal‐weight, overweight and obese patients undergoing surgery for malignancy (Word document) Table S3 Comparison across World Health Organization obesity subgroups I, II and III (Word document) Table S4 Thirty‐day mortality by operative risk class across body mass index groups (Word document) Snapshot quiz 16/9 Supporting information Table S1 Derivation of the operative risk class by mortality rate Table S2 Comparison of normal‐weight, overweight and obese patients undergoing surgery for malignancy Table S3 Comparison across World Health Organization obesity subgroups I, II and III Table S4 Thirty‐day mortality by operative risk class across body mass index groups

Introduction Surgery has recently been recognized as a complex healthcare intervention1, 2, 3, 4. Complex interventions comprise multiple interacting components that may be accompanied by concomitant interventions (or co‐interventions), including anaesthesia and elements of preoperative and postoperative care2. This complexity can create challenges during the design of surgical interventions in RCTs, in terms of establishing standards of surgery and monitoring whether interventions are delivered as intended. This is exemplified in a recent systematic review5 of 80 RCTs, reporting details of 160 surgical interventions, which found that only 47 (29·4 per cent) were reported to be standardized in some way, and monitoring of adherence to the intervention was similarly poor. These issues have partly been addressed through the publication of the SPIRIT6 and TIDieR7 guidance. The SPIRIT statement provides a checklist of 33 items to be reported in trial protocols. Items relating to interventions (11a–d) recommend that the trial protocol provides information about ‘each group with sufficient detail to allow replication’ and ‘procedures for monitoring adherence to intervention protocols’. The TIDieR guidance – an extension of item 11 of SPIRIT – comprises 12 items relating to the description of all types of intervention, and recommends that the duration, dose and materials used in the intervention are provided.

il to allow replication’ and ‘procedures for monitoring adherence to intervention protocols’. The TIDieR guidance – an extension of item 11 of SPIRIT – comprises 12 items relating to the description of all types of intervention, and recommends that the duration, dose and materials used in the intervention are provided. Although SPIRIT and TIDieR represent important progress for the design and reporting of interventions within RCTs, these guidance documents are not specific for, or easily applicable to, surgical interventions. For example, there is no such thing as a ‘dose’ of a surgical intervention and surgery cannot typically be delivered in an identical manner multiple times. It is difficult, therefore, to know how these recommendations should be used and applied during the design of RCTs in surgery, meaning that the optimal way of describing surgical interventions remains uncertain. The aim of this study was to develop new methods for standardizing and monitoring surgical interventions within RCTs in surgery.

refore, to know how these recommendations should be used and applied during the design of RCTs in surgery, meaning that the optimal way of describing surgical interventions remains uncertain. The aim of this study was to develop new methods for standardizing and monitoring surgical interventions within RCTs in surgery. Methods Development of the typology The typology – defined as a system used for grouping or classifying items according to how they are similar – was informed by a detailed systematic review of how surgical interventions are described, standardized and monitored in published RCTs5. Briefly, RCTs in surgery, published between 2010 and 2011, were identified by hand‐searching online archives of the top six journals ranked by impact factor for each of general medicine and surgery8. Included were trials in any surgical specialty evaluating a surgical intervention. This was defined as trials that involve physically changing body tissues and organs through manual operation such as cutting, abrading, suturing or the use of lasers4. Where available, trial protocols were obtained for each study and analysed in the same way as full‐text articles.

ty evaluating a surgical intervention. This was defined as trials that involve physically changing body tissues and organs through manual operation such as cutting, abrading, suturing or the use of lasers4. Where available, trial protocols were obtained for each study and analysed in the same way as full‐text articles. A detailed analysis of the included RCTs and protocols was undertaken using a modified framework approach for the analysis of qualitative data9. Framework analysis is usually a deductive approach used to analyse content by assigning descriptive labels, which are outlined a priori and collectively comprise a framework. In this study, it was not possible to assign labels a priori, and the process was therefore modified such that a subset of ten papers were read and reread to understand the data and develop a preliminary framework. Remaining trial reports were read sequentially and, where required, existing categories were amended or removed, and additional elements added. This iterative process of testing the framework was repeated independently by two researchers until all included papers had been assessed. Once all studies and protocols had been reviewed, a meeting with the research team was held to discuss the proposed typology. Following this, the typology was modified (1 category was added and some existing categories were rephrased). Subsequently, all papers were reassessed with the updated typology to ensure that every description was accounted for. Results Typology for designing surgical interventions in RCTs

A detailed analysis of the included RCTs and protocols was undertaken using a modified framework approach for the analysis of qualitative data9. Framework analysis is usually a deductive approach used to analyse content by assigning descriptive labels, which are outlined a priori and collectively comprise a framework. In this study, it was not possible to assign labels a priori, and the process was therefore modified such that a subset of ten papers were read and reread to understand the data and develop a preliminary framework. Remaining trial reports were read sequentially and, where required, existing categories were amended or removed, and additional elements added. This iterative process of testing the framework was repeated independently by two researchers until all included papers had been assessed. Once all studies and protocols had been reviewed, a meeting with the research team was held to discuss the proposed typology. Following this, the typology was modified (1 category was added and some existing categories were rephrased). Subsequently, all papers were reassessed with the updated typology to ensure that every description was accounted for. Results Typology for designing surgical interventions in RCTs A total of 80 RCTs evaluating 160 interventions within a range of surgical subspecialties were identified (Fig. S1, supporting information10). Of the 160 interventions, at least some textual description of the surgical intervention (beyond its name) was provided for 118 (73·8 per cent) and this informed the typology. The typology is divided into three sections relating to the description, standardization and monitoring of surgical interventions (Fig. 1). The final part of the paper demonstrates how a trial protocol for surgical interventions can be designed using the typology, illustrating this with worked examples.

d the typology. The typology is divided into three sections relating to the description, standardization and monitoring of surgical interventions (Fig. 1). The final part of the paper demonstrates how a trial protocol for surgical interventions can be designed using the typology, illustrating this with worked examples. Figure 1 Overview of the typology of surgical interventions BJS-10254-FIG-0001-cSection 1: Intervention description Three ways of describing a surgical intervention were identified: the overall technical purpose of the intervention, the intervention components and the steps within each component. Overall technical purpose of the surgical intervention The purpose of a surgical intervention can be classified as being exploratory, a resection and/or a reconstruction. These three purposes are not mutually exclusive and some interventions may traverse more than one category. For example, resection of bowel may involve resection (removal of the diseased bowel) and reconstruction (reconnection of the bowel). Initial specification of the overall technical purpose of a surgical intervention facilitates the identification of the underlying components that require consideration. Identification of intervention components

The purpose of a surgical intervention can be classified as being exploratory, a resection and/or a reconstruction. These three purposes are not mutually exclusive and some interventions may traverse more than one category. For example, resection of bowel may involve resection (removal of the diseased bowel) and reconstruction (reconnection of the bowel). Initial specification of the overall technical purpose of a surgical intervention facilitates the identification of the underlying components that require consideration. Identification of intervention components Surgical interventions can be divided into components, that is constituent parts or elements of a larger whole. The term ‘component’ was selected (rather than element, part, phase or other similar words) because of its established use in surgical education and training, and within the TIDieR guidance itself 7, 11. A list of all potential components of surgical interventions, identified from the 80 RCTs used to develop the typology, is provided in Table 1. Some surgical procedures may include all of the components, whereas others may only include a few. The minimum components of a surgical intervention are the creation and closure of an incision (two components). Table 1 Definitions of the components of surgical interventions

Surgical interventions can be divided into components, that is constituent parts or elements of a larger whole. The term ‘component’ was selected (rather than element, part, phase or other similar words) because of its established use in surgical education and training, and within the TIDieR guidance itself 7, 11. A list of all potential components of surgical interventions, identified from the 80 RCTs used to develop the typology, is provided in Table 1. Some surgical procedures may include all of the components, whereas others may only include a few. The minimum components of a surgical intervention are the creation and closure of an incision (two components). Table 1 Definitions of the components of surgical interventions Components of the intervention Description Before skin incision Events associated with the surgical intervention itself, but occurring before the skin incision, e.g. patient positioning, skin preparation, hair removal, surgical scrub Incision(s) and access The cut(s) made into skin and deeper tissues. This may require consideration of access, i.e. the method used to approach the operation. Broadly this can be categorized as open or minimally invasive, and further subdivided into multiple‐port, single‐port, robotic or natural‐orifice approaches Dissection The process of exposing an organ, tissue or structure Resection Removal of all or part of an organ, tissue or structure Haemostasis The stopping of bleeding or arrest of blood circulation in an organ, tissue or structure Reconstruction The process of rebuilding, repairing or replacing an organ, tissue or structure. This component may include an anastomosis (connection between two structures) or the insertion of a surgical adjunct such as a mesh or prosthesis Closure The process of closing or sealing the incision(s). Several layers of closure may be required (e.g. skin, fascia) After skin closure Any event associated with the surgical intervention but undertaken after skin closure (e.g. application of dressings or bandages) Insertion of surgical adjunct This component relates to the insertion of surgical adjuncts that are not related directly to reconstruction, but are inserted at the time of the surgical procedure (e.g. drains or feeding tubes) Intraoperative diagnosis Further characterization of a disease process or anatomy during the surgical procedure itself (e.g. intraoperative cholangiography, blue dye tests or scintigraphy) Other Any other component not listed above

n, but are inserted at the time of the surgical procedure (e.g. drains or feeding tubes) Intraoperative diagnosis Further characterization of a disease process or anatomy during the surgical procedure itself (e.g. intraoperative cholangiography, blue dye tests or scintigraphy) Other Any other component not listed above Identification of individual steps of interventions Detailed analysis of the components of surgical interventions identified that there are steps within each component, representing the precise details within a component. For example, making an incision (1 component) involves several individual details including its location, length, direction and depth. The number and type of steps within any component may be large and wide‐ranging, and vary between interventions. It is therefore not possible to propose a uniform typology for the steps of surgical interventions. Steps can be identified for each intervention once the technical purpose and the constituent components have been established. It is recommended that descriptions of surgical interventions are considered at three levels in trial protocols: the overall intervention, its components, and steps within each component. Examples are provided in Table 2. Initially, establishing all three levels of description of each intervention is necessary. This detailed intervention description can then be used to consider how interventions (and their components and steps) might be standardized (section 2) and monitored (section 3) – if at all – within an RCT. Table 2 Levels of descriptions of surgical interventions

It is recommended that descriptions of surgical interventions are considered at three levels in trial protocols: the overall intervention, its components, and steps within each component. Examples are provided in Table 2. Initially, establishing all three levels of description of each intervention is necessary. This detailed intervention description can then be used to consider how interventions (and their components and steps) might be standardized (section 2) and monitored (section 3) – if at all – within an RCT. Table 2 Levels of descriptions of surgical interventions Level of description Example Entire intervention ‘The open tension‐free mesh hernioplasty was performed according to Lichtenstein’12 Component of intervention ‘Reconstruction consisted of replacement…with an artificial lumbar disc’13 Steps within component ‘Pneumoperitoneum was established by open access and maintained at 12–15 mmHg. Three 12‐mm ports were placed: in the midline above the umbilicus, in the epigastrium and in the ipsilateral iliac fossa. A 5‐mm port was placed in the flank’14

Component of intervention ‘Reconstruction consisted of replacement…with an artificial lumbar disc’13 Steps within component ‘Pneumoperitoneum was established by open access and maintained at 12–15 mmHg. Three 12‐mm ports were placed: in the midline above the umbilicus, in the epigastrium and in the ipsilateral iliac fossa. A 5‐mm port was placed in the flank’14 Section 2: Standardization of surgical interventions In an RCT, it is critical to decide whether a surgical intervention needs to be standardized, and how this should be done. Standardization refers to whether the trial protocol specifies exactly how an intervention should be delivered, and may inherently necessitate monitoring during the trial to establish whether centres and surgeons actually followed these instructions. There are several factors that might influence intervention standardization, such as the overall trial design (for example pragmatic versus explanatory) or the developmental stage of the intervention15. For surgical interventions, it is recommended that three aspects of standardization are considered for each component and step: the type of standardization, conditions relating to it, and the flexibility of delivery. These factors should all be set out clearly in the protocol to inform trial conduct, monitoring and reporting of what was delivered during the trial. Types of standardization

Section 2: Standardization of surgical interventions In an RCT, it is critical to decide whether a surgical intervention needs to be standardized, and how this should be done. Standardization refers to whether the trial protocol specifies exactly how an intervention should be delivered, and may inherently necessitate monitoring during the trial to establish whether centres and surgeons actually followed these instructions. There are several factors that might influence intervention standardization, such as the overall trial design (for example pragmatic versus explanatory) or the developmental stage of the intervention15. For surgical interventions, it is recommended that three aspects of standardization are considered for each component and step: the type of standardization, conditions relating to it, and the flexibility of delivery. These factors should all be set out clearly in the protocol to inform trial conduct, monitoring and reporting of what was delivered during the trial. Types of standardization The type of standardization required for each component and step of an intervention may be classified as mandated, prohibited or optional. A mandatory step, for example, would be essential to perform in all interventions (and if not performed constitutes a deviation from the protocol), whereas the opposite is true if a step is prohibited. An optional component or step is one that may or may not be performed, at the discretion of each participating surgeon. Conditions relating to standardization

The type of standardization required for each component and step of an intervention may be classified as mandated, prohibited or optional. A mandatory step, for example, would be essential to perform in all interventions (and if not performed constitutes a deviation from the protocol), whereas the opposite is true if a step is prohibited. An optional component or step is one that may or may not be performed, at the discretion of each participating surgeon. Conditions relating to standardization During trial design, trialists should identify clinical findings or conditions that may influence the type of standardization required, and detail them in the trial protocol so it is clear what action to take when they are encountered. For example, it may be necessary to decide whether to undertake a cholecystectomy at the same time as a bariatric procedure. A trial protocol therefore needs to describe the conditions relating to this clinical situation: for example, a concomitant cholecystectomy may be mandated only among patients with symptomatic gallstone disease (that is, under certain conditions) and prohibited in other patients. Flexibility of standardization A range of flexibility is possible, so that a component or step can be delivered exactly as described within the protocol, within boundaries or totally flexibly. For example, a trial protocol may require surgeons to create an anastomosis using 4·0 polypropylene (exactly as described), any 4·0 or 5·0 monofilament suture (within boundaries) or simply state that this can be performed according to their own preference (totally flexible).

ocol, within boundaries or totally flexibly. For example, a trial protocol may require surgeons to create an anastomosis using 4·0 polypropylene (exactly as described), any 4·0 or 5·0 monofilament suture (within boundaries) or simply state that this can be performed according to their own preference (totally flexible). Section 3: Monitoring of surgical interventions during the trial Monitoring how surgical interventions are actually delivered in a trial (fidelity) is essential to inform the interpretation of results and subsequent implementation of interventions in practice. Three possibilities for recording and reporting fidelity were identified: the intervention, component or step is not delivered at all; an intervention, component or step from another trial group is delivered instead; or an entirely different intervention, component or step is delivered (Table 3). Additionally, the reasons for which the above deviations occur may be crucial and it is therefore recommended that trialists consider recording these throughout the RCT. Table 3 Levels and types of intervention fidelity

onsented to a two‐group trial but, by the time they were randomized, the trial had switched to three groups. It was not feasible to have a further consultation to discuss the trial again, so the randomization for these patients was handled centrally by the trials unit to avoid them being allocated to the new procedure. Another decision that needs to be made is whether the allocation ratio should be chosen to create approximately equally sized groups at the end of the trial, or whether the ratio should be 1 : 1 : 1 going forward, accepting that the sample size for the two original study groups will be larger than that for the new group. In the By‐Band‐Sleeve adaptation, the allocation ratio at each centre was adjusted to achieve approximately equal allocation overall. Although the revised allocation ratio should provide comparable power for all group comparisons, it is not without ‘risks’. A balanced allocation may not be achieved if the trial stops early for any reason, or if the recruitment rates differ significantly from those predicted.

Section 3: Monitoring of surgical interventions during the trial Monitoring how surgical interventions are actually delivered in a trial (fidelity) is essential to inform the interpretation of results and subsequent implementation of interventions in practice. Three possibilities for recording and reporting fidelity were identified: the intervention, component or step is not delivered at all; an intervention, component or step from another trial group is delivered instead; or an entirely different intervention, component or step is delivered (Table 3). Additionally, the reasons for which the above deviations occur may be crucial and it is therefore recommended that trialists consider recording these throughout the RCT. Table 3 Levels and types of intervention fidelity Level of fidelity Types Deviation from intended intervention Did not receive any intervention Received intervention in other trial arm Received an alternative intervention not being evaluated in the trial Deviation from component(s) of the intended intervention Did not receive the component Component delivered according to description in other trial arm Received an alternative component, or component performed in a different way Deviation from step(s) within component(s) of the intended intervention Step not done Step from other trial arm performed Different step performed, or step performed in a different way Example of how the typology can be applied to surgical RCTs

eceived an alternative component, or component performed in a different way Deviation from step(s) within component(s) of the intended intervention Step not done Step from other trial arm performed Different step performed, or step performed in a different way Example of how the typology can be applied to surgical RCTs The typology was used to design the interventions in two surgical RCTs, and subsequently report these details in the trial protocols. The Rescue‐ASDH (Randomized Evaluation of Surgery with Craniectomy for patients Undergoing Evacuation of Acute SubDural Haematoma) trial16 compares the effectiveness and cost‐effectiveness of craniotomy and decompressive craniectomy for acute subdural haematoma. The By‐Band‐Sleeve study17 compares the effectiveness of laparoscopic adjustable gastric band, Roux‐en‐Y gastric bypass and laparoscopic sleeve gastrectomy for patients with severe and complex obesity. In conjunction with two of the present researchers and the trial teams, the typology was used to consider the overall purpose of these interventions and to identify the constituent components and steps. Subsequently, the degree of standardization required for each was established. Both are multicentre pragmatic RCTs and all interventions are undertaken routinely within clinical practice. It was therefore agreed that only the key intervention components needed to be standardized, in order to distinguish the interventions in each trial group from one another.

ion required for each was established. Both are multicentre pragmatic RCTs and all interventions are undertaken routinely within clinical practice. It was therefore agreed that only the key intervention components needed to be standardized, in order to distinguish the interventions in each trial group from one another. As an example, Table 4 lists the components and steps of laparoscopic Roux‐en‐Y gastric bypass (which has a purpose of reconstruction), and the degree of standardization required for each step. The agreed intervention description (as detailed within the trial protocol) is also provided, together with information about fidelity to each aspect of this description. Standardization of the interventions in the Rescue‐ASDH trial (undertaken by surgeons and trialists independent of the typology research team) is described in Tables S1 and S2 (supporting information). Table 4 Standardization of laparoscopic Roux‐en‐Y gastric bypass in the By‐Band‐Sleeve study

As an example, Table 4 lists the components and steps of laparoscopic Roux‐en‐Y gastric bypass (which has a purpose of reconstruction), and the degree of standardization required for each step. The agreed intervention description (as detailed within the trial protocol) is also provided, together with information about fidelity to each aspect of this description. Standardization of the interventions in the Rescue‐ASDH trial (undertaken by surgeons and trialists independent of the typology research team) is described in Tables S1 and S2 (supporting information). Table 4 Standardization of laparoscopic Roux‐en‐Y gastric bypass in the By‐Band‐Sleeve study Components and steps Laparoscopic Roux‐en‐Y gastric bypass Description provided in trial protocol17 Adherence during trial (n = 75) Type Conditions Flexibility Incision and access Establishing pneumoperitoneum Mandatory None Veress/open technique Procedures will be undertaken laparoscopically. Methods used to create a pneumoperitoneum, and the placement of laparoscopic ports and retractors, are at the discretion of the surgeon 75 (100) Insertion of additional ports Optional Poor visibility Flexible Dissection Creation of a horizontal pouch Prohibited n.a. n.a. The pouch can be created according to surgeons' usual practice, although a horizontal gastric pouch that includes fundus is prohibited 75 (100) Reconstruction Measurement of the gastric limb Mandatory None Maximum 150 cm Methods used to create the biliary and gastric limbs are flexible, although upper limits of 75 and 150 cm respectively are recommended 120 (100–150)*

ve approximately equal allocation overall. Although the revised allocation ratio should provide comparable power for all group comparisons, it is not without ‘risks’. A balanced allocation may not be achieved if the trial stops early for any reason, or if the recruitment rates differ significantly from those predicted. The primary reason for adapting a trial to include a new procedure is to keep it up‐to‐date and relevant to clinical practice. It is, however, important to adapt a trial and include a new procedure only when surgeons are comfortable to discuss clinical equipoise with patients. These issues are challenging, because surgeons may naturally believe that the new procedure has advantages and they will find it more interesting/exciting to do a new procedure rather than a standard technique. In these situations, surgeons may state that a randomized evaluation is not possible because it is ‘too late’ and they do not have clinical equipoise. This view is represented in Buxton's law which says: ‘it is always too early for rigorous evaluation until, unfortunately, it's suddenly too late’21. Despite this established mantra, there have been several recent examples of successful surgical trials in which surgeons have been supported to provide balanced information to inform patient decision‐making and ensure successful recruitment22, 23. Because of the success of these other trials, this support was included in By‐Band and By‐Band‐Sleeve. The QRI uses qualitative methods to understand how surgeons communicate information to patients. Consultations are audiorecorded and analysed, and meetings are held with surgeons to provide training in how to optimize informed consent. Although this approach to supporting trial recruitment represents progress for randomized trials, there is a need to use these methods to optimize information provision in other settings when surgeons talk to patients, such as during early‐phase studies and with the introduction of novel techniques.

ual practice, although a horizontal gastric pouch that includes fundus is prohibited 75 (100) Reconstruction Measurement of the gastric limb Mandatory None Maximum 150 cm Methods used to create the biliary and gastric limbs are flexible, although upper limits of 75 and 150 cm respectively are recommended 120 (100–150)* Measurement of the biliary limb Mandatory None Maximum 75 cm 30 (3–60)* Opening of the retrocolic window Optional None Flexible Routing of the Roux limb (antecolic or retrocolic) is flexible Antecolic 21 (28) Retrocolic 54 (72) Anastomoses Gastrojejunostomy Mandatory None Sutured/stapled, 1–2 layers, oral route or intra‐abdominal Anastomoses can be performed as the surgeon chooses (e.g. stapled or sutured, circular or linear, single or double layer) Stapled 75 (100) Circular 10 (13) Linear 65 (87) Jejunojejunostomy Mandatory None Sutured/stapled, 1–2 layers Stapled 75 (100) Triple 25 (33) Single 50 (67) Closure Closure of mesenteric defects Optional None Flexible Closure of mesteric defects is optional Peterson's space 59 (79) Jejunojejunostomy 58 (77) Mesocolon 54 (100)† Other Use of a bougie Optional None Flexible Use of a bougie is optional 66 (88) Values in parentheses are percentages unles indicated otherwise; * values are median (i.q.r.). † Only retrocolic reconstructions were included in the denominator, because a mesocolonic window is not created during antecolic bypasses. n.a., Not applicable.

Other Use of a bougie Optional None Flexible Use of a bougie is optional 66 (88) Values in parentheses are percentages unles indicated otherwise; * values are median (i.q.r.). † Only retrocolic reconstructions were included in the denominator, because a mesocolonic window is not created during antecolic bypasses. n.a., Not applicable. Discussion This study describes a novel framework (a typology) for describing surgical interventions in RCTs. It provides guidance on how to consider the extent of intervention standardization in trial protocols, and subsequent monitoring during the trial itself. The typology requires that the overall purpose of an intervention is described, and that it is deconstructed into constituent components and steps. The deconstructed trial intervention then provides a platform to inform the level of standardization of each component and step to be delivered and monitored within the trial. These factors can be discussed and agreed during trial design (potentially as part of pretrial pilot work), so that details for undertaking the surgical interventions can be provided within the main trial protocol, and subsequently monitored during the trial itself. The typology will help to clarify exactly how interventions were intended to be delivered within RCTs and allow the trialists to monitor adherence to this. Application of the typology to RCTs in surgery has the potential to improve trial conduct, and better to inform the implementation of successful interventions in clinical practice.

l help to clarify exactly how interventions were intended to be delivered within RCTs and allow the trialists to monitor adherence to this. Application of the typology to RCTs in surgery has the potential to improve trial conduct, and better to inform the implementation of successful interventions in clinical practice. It may not always be necessary or appropriate to standardize each component or step of a surgical intervention. This should be driven by the research question, the interventions being compared (including the expertise of those delivering them) and whether the trial is predominantly explanatory or pragmatic15. In explanatory trials, which determine the efficacy of interventions, great detail may be necessary because the interventions are often novel and their safety needs to be assessed within carefully controlled settings. Pragmatic trials, which determine whether interventions are effective in the real world, are often multicentre studies with large numbers of surgeons. Under such circumstances, specifying each operative step (and those of all accompanying co‐interventions) is likely to create difficulties, and ensuring that each step was delivered as planned may be unrealistic. A balance between adequate standardization and practicality is therefore necessary and appropriate. One way of achieving this is to determine the minimum active ingredients of the intervention18 – those that are thought to optimize outcomes or those that are different between the interventions in each trial group – and the degree to which they need to be standardized. In this way, monitoring only the key components may be sufficient, rather than monitoring all components and steps, in order to ensure the intervention is actually delivered as planned.

omes or those that are different between the interventions in each trial group – and the degree to which they need to be standardized. In this way, monitoring only the key components may be sufficient, rather than monitoring all components and steps, in order to ensure the intervention is actually delivered as planned. A potential limitation of this study is that the typology and its categories may not be fully comprehensive. Although further testing could be undertaken with more trial reports, 80 papers were included, providing a total of 160 interventions. The final framework was applied to all papers, and all of the information regarding each intervention could be classified according to the existing typology. Another limitation is that application of the typology was limited to four surgical interventions across two RCTs. A final limitation is that, although specific to surgery, the typology focuses solely on the intervention itself, meaning that, currently, it will need to be used in conjunction with other guidelines such as TIDieR and SPIRIT. Development of a typology for co‐interventions, and identification of the factors that might influence the degree of standardization required (for example explanatory versus pragmatic trials), was beyond the scope of this study, which aimed to derive a classification system from existing literature. Work is ongoing in both of these areas, to develop a comprehensive set of guidelines for surgical RCTs. This will require considerable testing, in both new and ongoing studies, across a variety of different interventions and settings, in order to establish its validity and usefulness.

n system from existing literature. Work is ongoing in both of these areas, to develop a comprehensive set of guidelines for surgical RCTs. This will require considerable testing, in both new and ongoing studies, across a variety of different interventions and settings, in order to establish its validity and usefulness. This typology of surgical interventions provides a framework for deconstructing surgical interventions into their constituent components and steps to ensure that all intervention components are considered a priori. In a pragmatic trial, after identifying all of the components, those deemed to be key or crucial can be agreed, such that parts requiring standardization are described clearly in the protocol and other components can be delivered according to surgeons' individual preferences. This will allow distinction between mandatory, prohibited and optional steps of an intervention, as well as those that can be delivered flexibly. This approach will require surgeons to agree on a few key details about how an intervention should be performed and within what boundaries, rather than all of its individual steps. Thus, other elements can be undertaken according to personal preference, removing the need for surgeons to conform to a detailed, universal, operative script. More importantly, engaging surgeons in designing interventions in this way may increase the likelihood that they will accept the results of RCTs in surgery and, if interventions are deemed to be effective, actually implement them in routine practice.

d for surgeons to conform to a detailed, universal, operative script. More importantly, engaging surgeons in designing interventions in this way may increase the likelihood that they will accept the results of RCTs in surgery and, if interventions are deemed to be effective, actually implement them in routine practice. Supporting information Additional supporting information may be found in the online version of this article: Fig. S1 PRISMA flow diagram of included studies (Word document) Table S1 Standardization of surgical interventions in the Rescue‐ASDH trial (Word document) Table S2 Final descriptions of Rescue‐ASDH interventions for the trial protocol (Word document) Supporting information Fig. S1 PRISMA flow diagram of included studies Table S1 Standardization of surgical interventions in the Rescue‐ASDH trial Table S2 Final descriptions of Rescue‐ASDH interventions for the trial protocol Click here for additional data file.

Table S2 Final descriptions of Rescue‐ASDH interventions for the trial protocol (Word document) Supporting information Fig. S1 PRISMA flow diagram of included studies Table S1 Standardization of surgical interventions in the Rescue‐ASDH trial Table S2 Final descriptions of Rescue‐ASDH interventions for the trial protocol Click here for additional data file. Acknowledgements This work was undertaken with the support of the Medical Research Council (MRC) ConDuCT‐II (Collaboration and innovation for Difficult and Complex randomized controlled Trials In Invasive procedures) Hub (MR/K025643/1). N.S.B. is an National Institute for Health Research (NIHR) Clinical Lecturer and J.A.C. holds an MRC UK Methodology fellowship (G1002292). The By‐Band‐Sleeve and Rescue‐ASDH studies are funded by the NIHR Health Technology Assessment (HTA) programme (project numbers 09/127/53 and 12/35/57 respectively). The views and opinions expressed are those of the authors and do not necessarily reflect those of the HTA programme, NIHR, UK National Health Service or Department of Health. Disclosure: The authors declare no conflict of interest.

e informed consent. Although this approach to supporting trial recruitment represents progress for randomized trials, there is a need to use these methods to optimize information provision in other settings when surgeons talk to patients, such as during early‐phase studies and with the introduction of novel techniques. This article describes an example of adaption from a two‐ to a three‐group surgical trial to include an increasingly prevalent intervention and allow its evaluation in a pragmatic study. The inclusion of an internal pilot and a formal review of success in recruiting to the two‐group design provided a convenient interval in which to consider whether a third group should be added. The justification was based primarily on external sources of information, in contrast to trial adaptation based on accruing evidence within the trial. This approach is recommended for RCTs in surgery to optimize the efficiency of the trial team infrastructure, allow surgeons to undertake an additional intervention, and maintain the relevance of the research question.

sources of information, in contrast to trial adaptation based on accruing evidence within the trial. This approach is recommended for RCTs in surgery to optimize the efficiency of the trial team infrastructure, allow surgeons to undertake an additional intervention, and maintain the relevance of the research question. Collaborators Other By‐Band‐Sleeve study investigators (principal investigator denoted by an asterisk): S. Agrawal*, S. Ajaz, Y. Koak (Homerton University Hospital, London); A. Ahmed*, N. Fakih, S. Hakky, K. Moorthy, S. Purkayastha (Imperial College Healthcare NHS Trust, London); S. Awad, K. Fareed, P. Leeder* (Royal Derby Hospital, Derby); S. Balupuri, W. Carr, N. Jennings*, P. Small (Sunderland Royal Hospital, Sunderland); R. Byrom, N. Davies* (Royal Bournemouth and Christchurch Hospitals); N. Carter*, B. Knight, S. Somers (Queen Alexandra Hospital, Portsmouth); V. Charalampakis, M. Daskalakis, R. Nijar, M. Richardson, R. Singhal*, P. Super (Heart of England NHS Foundation Trust, Birmingham); M. Clarke, A. Cota, I. Finlay* (Royal Cornwall Hospital, Truro); S. Dexter, J. Hayden*, S. Mehta, A. Sarela (St James's University Hospital, Leeds); J. Kelly (University Hospital Southampton, Southampton); D. Mahon, H. Noble (Musgrove Park Hospital, Taunton). Editor's comments

val from the funder to proceed with the adaptation. The transition went smoothly and the study is recruiting steadily to all three procedures (Fig. 2). In February 2017, the 50 per cent recruitment milestone was reached (671 patients randomized into the trial). All study centres continue to be supported with the QRI13. Figure 2 Recruitment into the By‐Band‐Sleeve study at January 2017 BJS-10562-FIG-0002-cDiscussion This paper describes adaptation of a two‐group RCT in bariatric surgery to a three‐group trial. The adaptation was proposed because of changing clinical practice, namely the increasing uptake of a new procedure in the absence of high‐quality evidence of its clinical effectiveness compared with the procedures already being evaluated. The trial team was aware of the interest in the new procedure when designing the By‐Band study, and specified the possibility of an adaptation in the original grant proposal. This stated that if recruitment in the internal pilot phase proved possible the trial team would consider whether adapting to a three‐group trial was relevant (and present a new proposal to the funder). This approach to designing and conducting RCTs in surgery is recommended, although not often used. It ensures the continued relevance of the trial to surgeons, allows efficient use of trial resources, and enables a procedure that has stabilized and is being widely used to undergo comparative rigorous evaluation.

nder). This approach to designing and conducting RCTs in surgery is recommended, although not often used. It ensures the continued relevance of the trial to surgeons, allows efficient use of trial resources, and enables a procedure that has stabilized and is being widely used to undergo comparative rigorous evaluation. Adapting a surgical RCT has proved challenging, with practical and scientific implications and complexities. The majority of experience has been gained in pharmaceutical studies where the pressure to adapt a trial arises from the emergence of competing ‘like‐me’ drugs9, and with pressure to evaluate multiple new drugs. In pharmaceutical trials short‐term or surrogate endpoints are typically used to inform adaptation decisions. In trials with longer‐term endpoints (the primary endpoint of By‐Band‐Sleeve is at 3 years), it is not sensible to adapt the trial based on short‐term outcomes, especially if no appropriate short‐term surrogate measure exists. However, as shown here, there is an opportunity to introduce additional group(s) to evaluate new intervention(s). The By‐Band‐Sleeve study also includes a ‘non‐randomized’ cohort of eligible patients who consent to data collection and follow‐up but not to randomization. Analysis of this cohort will provide important insights into how the trial has been integrated into each clinical practice at the study centres.

Introduction Large‐scale RCTs in surgery can be difficult to design and conduct. There are challenges with recruitment, intervention complexity and outcome assessment. Recruitment may be slow because surgeons and associated clinical teams (such as anaesthetists, nurses, dieticians and psychologists) are unfamiliar with presenting uncertainty and recruiting patients, and there is often a lack of clinical trials research infrastructure in surgical departments1, 2, 3. As a result, extensions to the recruitment period may be required or the trial may be stopped early before the target sample size is reached4, 5. The additional time taken to deliver surgical trials has important implications over and above increased research costs, meaning that the initial research question can become outdated owing to changing practice. Often, new interventions are introduced while the trial is recruiting, despite a lack of evidence from RCTs for effectiveness. This may be due to ‘fashion and popular opinion’, ‘common sense/plausibility’, ‘marketing by industry’ or by ‘influential opinion leaders’. Surgical practice can therefore change, and new interventions can be widely implemented, without evaluation and before outcomes from the trial are available. This contrasts with the introduction of new medicines, which requires formal evaluation and strict adherence to governance procedures before licensing for general use.

cal practice can therefore change, and new interventions can be widely implemented, without evaluation and before outcomes from the trial are available. This contrasts with the introduction of new medicines, which requires formal evaluation and strict adherence to governance procedures before licensing for general use. The adoption of new but unevaluated surgical interventions outside a trial can threaten the viability of an ongoing trial in a number of ways. Participating surgeons may want to undertake the new intervention, reducing the proportion of patients offered the trial. In the case of a new medication, this can be controlled to some extent by a commissioning decision that the medication should be available ‘only in research’6. Although this option may be available for surgical interventions in some countries, surgeons are rarely willing to apply it voluntarily. Thus, patients can remain outside a trial and have the new intervention, as is evidenced by the way in which new surgical interventions are often introduced and adopted7, 8. In the absence of an ‘only in research’ ruling, patients themselves may prefer the new intervention (albeit based on limited evidence and understanding). It is also possible that centres and surgeons may choose not to join the trial because the research question may appear obsolete, although there is a lack of high‐quality evidence to inform such judgements.

ruling, patients themselves may prefer the new intervention (albeit based on limited evidence and understanding). It is also possible that centres and surgeons may choose not to join the trial because the research question may appear obsolete, although there is a lack of high‐quality evidence to inform such judgements. One approach to keeping a trial relevant is to use an adaptive design. The term is used most often in the context of changing interventions, or respecifying the comparator, as evidence emerges about the interventions being evaluated9. For example, by evaluating short‐term outcomes (or side‐effects), a decision can made to drop an intervention when the accruing data confidently show that it could never be effective – on grounds of futility. Importantly, adaptive trial protocols specify criteria for adding or dropping interventions before the trial begins. Although there is increasing knowledge and acceptance of the use of adaptive trial designs for novel pharmaceutical interventions, the principle has been applied less often in trials of surgery. Adapting a surgical trial is challenging. It may be difficult to agree when a new surgical intervention is sufficiently well developed and standardized for evaluation, and to establish when surgeons have sufficient experience of it to contribute to its evaluation in a trial.

One approach to keeping a trial relevant is to use an adaptive design. The term is used most often in the context of changing interventions, or respecifying the comparator, as evidence emerges about the interventions being evaluated9. For example, by evaluating short‐term outcomes (or side‐effects), a decision can made to drop an intervention when the accruing data confidently show that it could never be effective – on grounds of futility. Importantly, adaptive trial protocols specify criteria for adding or dropping interventions before the trial begins. Although there is increasing knowledge and acceptance of the use of adaptive trial designs for novel pharmaceutical interventions, the principle has been applied less often in trials of surgery. Adapting a surgical trial is challenging. It may be difficult to agree when a new surgical intervention is sufficiently well developed and standardized for evaluation, and to establish when surgeons have sufficient experience of it to contribute to its evaluation in a trial. The aim of this article is to describe methods that were used to adapt a surgical RCT to include an additional, emerging surgical technique. The methods were applied in what started as the By‐Band study of two different bariatric surgical techniques10, now adapted to the By‐Band‐Sleeve study to include an additional evaluation of sleeve gastrectomy.

escribe methods that were used to adapt a surgical RCT to include an additional, emerging surgical technique. The methods were applied in what started as the By‐Band study of two different bariatric surgical techniques10, now adapted to the By‐Band‐Sleeve study to include an additional evaluation of sleeve gastrectomy. Methods The By‐Band study was a multicentre pragmatic RCT undertaken in the UK with the aim to evaluate the effectiveness and cost‐effectiveness of two interventions, laparoscopic adjustable gastric band and laparoscopic Roux‐en‐Y gastric bypass. It was designed in two phases. The first phase was set up in two centres to examine recruitment. This internal pilot phase included formal progression criteria/goals (Table 1), which were prespecified and approved by the funder. Formal review of phase 1 was scheduled to take place after the study had been open for 2 years. The funder agreed that if the criteria/goals were met the trial could continue into the second phase. Full details of the protocol are available elsewhere10. The protocol allows all eligible trial participants to join the study. Those who do not consent to randomization are invited to contribute data on the operation chosen and participate in follow‐up. These data will be used to understand the generalizabilty of the study results. Table 1 Details of pre‐agreed recruitment and retention goals (‘progression criteria’) in the internal pilot phase of the study, and actual progress achieved

Methods The By‐Band study was a multicentre pragmatic RCT undertaken in the UK with the aim to evaluate the effectiveness and cost‐effectiveness of two interventions, laparoscopic adjustable gastric band and laparoscopic Roux‐en‐Y gastric bypass. It was designed in two phases. The first phase was set up in two centres to examine recruitment. This internal pilot phase included formal progression criteria/goals (Table 1), which were prespecified and approved by the funder. Formal review of phase 1 was scheduled to take place after the study had been open for 2 years. The funder agreed that if the criteria/goals were met the trial could continue into the second phase. Full details of the protocol are available elsewhere10. The protocol allows all eligible trial participants to join the study. Those who do not consent to randomization are invited to contribute data on the operation chosen and participate in follow‐up. These data will be used to understand the generalizabilty of the study results. Table 1 Details of pre‐agreed recruitment and retention goals (‘progression criteria’) in the internal pilot phase of the study, and actual progress achieved Pre‐agreed goals to be achieved Actual progress achieved To screen 400 patients 333 patients were screened (83 per cent of that expected in original grant application) Sixty per cent of screened patients were eligible 231 patients (69·4 per cent) were found to be eligible To increase recruitment rates from 30 per cent over the first 18 months of recruitment, rising to 50 per cent thereafter* Recruitment rates were 27 per cent over the first 6 months of recruitment, rising to 39 per cent thereafter (target 79 randomized, achieved 80) Less than 5 per cent did not receive allocated treatment 2 of 57 (4 per cent) failed to receive the allocated treatment Less than 5 per cent lost to follow‐up 1 (2 per cent) withdrawn To reconsider the role of sleeve gastrectomy and whether a three‐group study should be proposed Sleeve data presented. The proposal to adapt the trial was approved by the funder * Recruitment rate is the percentage of eligible patients consenting to join the randomized study.

lost to follow‐up 1 (2 per cent) withdrawn To reconsider the role of sleeve gastrectomy and whether a three‐group study should be proposed Sleeve data presented. The proposal to adapt the trial was approved by the funder * Recruitment rate is the percentage of eligible patients consenting to join the randomized study. When the study was designed in 2009–2010, gastric band and bypass accounted for 80 per cent of all bariatric operations in the National Health Service (NHS)11. A new operation, sleeve gastrectomy, was starting to be performed, but accounted for only 8 per cent of operations. Surgeons' experience of sleeve gastrectomy was limited, few long‐term outcome data were available, and there was no consensus on how the operation should be performed. The By‐Band research team therefore decided that it was not appropriate to include sleeve gastrectomy as a third group at that time. This information was presented in the grant application, with a brief proposal to review with the trial steering committee the case for adding a sleeve group at the end of the internal pilot phase. The proposal recognized that, if the trial was adapted, an increase in sample size would likely be needed.

p at that time. This information was presented in the grant application, with a brief proposal to review with the trial steering committee the case for adding a sleeve group at the end of the internal pilot phase. The proposal recognized that, if the trial was adapted, an increase in sample size would likely be needed. Criteria for progression from the pilot to the main trial The goals used for progression from the phase 1 pilot to the main trial (phase 2) comprised five recruitment and protocol adherence‐based measures. The study team also planned to develop a core outcome set during the pilot (Table 1)12. An intervention to optimize recruitment was included in the pilot because of well recognized strong preferences for interventions amongst patients and surgeons. The Quintet Recruitment Intervention (QRI) was used13. This has two major components: understanding recruitment as it happens and then developing a plan of action to address identified difficulties; and optimizing informed consent in collaboration with the RCT chief investigator and the clinical trials unit. The QRI has been used successfully in several surgical trials.

3. This has two major components: understanding recruitment as it happens and then developing a plan of action to address identified difficulties; and optimizing informed consent in collaboration with the RCT chief investigator and the clinical trials unit. The QRI has been used successfully in several surgical trials. Methods for proposing an adaptation to include sleeve gastrectomy Four sources of information were reviewed to determine whether adaptation to include sleeve gastrectomy should be recommended: data on current surgical practice including both NHS and private healthcare provision; published comparative evidence and ongoing trials; expert surgical opinion (trial steering and data monitoring and safety committees, and specialist society); and published data on the stability of sleeve gastrectomy. The evidence collated was submitted to the funder with the proposal to adapt the trial to include a third group, changing the trial identity to By‐Band‐Sleeve. Implementation of the adaptation A list of practical and logistical steps needed to implement an adaptation from two to three groups was drawn up. These steps included considering the implications for trial participants at different stages of the trial (for example, invited to take part but not consented, consented but not randomized, randomized but awaiting surgery, in follow‐up). A strategy for how to proceed was prepared.

from two to three groups was drawn up. These steps included considering the implications for trial participants at different stages of the trial (for example, invited to take part but not consented, consented but not randomized, randomized but awaiting surgery, in follow‐up). A strategy for how to proceed was prepared. Results Progression from phase 1 to phase 2 of the trial The grant opened in January 2012 and the formal review of phase 1 took place in January 2014. The study opened to recruitment in the first centre in November 2012, and in the second in February 2013. Between November 2012 and December 2013, 80 patients were recruited and randomized (79 anticipated). The other goals for progression to the main trial were also met (Table 1). The funder therefore agreed that the trial should progress to phase 2 and expand. Eleven centres are currently participating in the trial; four phase 2 centres opened before the adaptation to By‐Band‐Sleeve, and five opened after the trial had been adapted. Adaptation to include sleeve gastrectomy Changes in current surgical practice

Results Progression from phase 1 to phase 2 of the trial The grant opened in January 2012 and the formal review of phase 1 took place in January 2014. The study opened to recruitment in the first centre in November 2012, and in the second in February 2013. Between November 2012 and December 2013, 80 patients were recruited and randomized (79 anticipated). The other goals for progression to the main trial were also met (Table 1). The funder therefore agreed that the trial should progress to phase 2 and expand. Eleven centres are currently participating in the trial; four phase 2 centres opened before the adaptation to By‐Band‐Sleeve, and five opened after the trial had been adapted. Adaptation to include sleeve gastrectomy Changes in current surgical practice Data from the National Bariatric Surgery Registry were reviewed to inform the rates of each type of operation11. Rates of sleeve gastrectomy increased from 9·0 to 28·1 per cent during phase 1 (Table 2). Over the same period, the rates for gastric bypass remained stable and the proportion of operations using gastric band surgery declined from 32·6 to 15·8 per cent overall. Analysis of data by NHS and private provision showed that, in the NHS, gastric band surgery accounted for just 11·0 per cent of bariatric procedures in 2013, but in the private health sectors band surgery remained the most common operation (Table 2). This pattern was mirrored elsewhere in the world14.

per cent overall. Analysis of data by NHS and private provision showed that, in the NHS, gastric band surgery accounted for just 11·0 per cent of bariatric procedures in 2013, but in the private health sectors band surgery remained the most common operation (Table 2). This pattern was mirrored elsewhere in the world14. Table 2 Rates of laparoscopic Roux‐en‐Y gastric bypass, laparoscopic adjustable gastric band and sleeve gastrectomy in the National Health Service and private sector, 2008–2013 2008–2009 2011 2012 2013 NHS and private practice Band 2132 (32·6) 1316 (25·5) 1358 (24·8) 891 (15·8) Bypass 3817 (58·4) 3030 (58·8) 2894 (52·9) 3176 (56·2) Sleeve 588 (9·0) 809 (15·7) 1218 (22·3) 1587 (28·1) Total 6537 5155 5470 5654 NHS only (HES data) Band n.a. 637 (16·9) 736 (18·2) 506 (11·0) Bypass n.a. 2540 (67·6) 2394 (59·1) 2816 (61·1) Sleeve n.a. 583 (15·5) 922 (22·8) 1290 (28·0) Total n.a. 3760 4052 4612 Private practice only* Band n.a. 679 (48·7) 622 (43·9) 385 (36·9) Bypass n.a. 490 (35·1) 500 (35·3) 360 (34·5) Sleeve n.a. 226 (16·2) 296 (20·9) 297 (28·5) Total n.a. 1395 1418 1042 Values in parentheses are percentages. * Data likely to be under reported. NHS, National Health Service; HES, Hospital Episodes Statistics; n.a., not available. Published comparative evidence and ongoing trials

Band n.a. 679 (48·7) 622 (43·9) 385 (36·9) Bypass n.a. 490 (35·1) 500 (35·3) 360 (34·5) Sleeve n.a. 226 (16·2) 296 (20·9) 297 (28·5) Total n.a. 1395 1418 1042 Values in parentheses are percentages. * Data likely to be under reported. NHS, National Health Service; HES, Hospital Episodes Statistics; n.a., not available. Published comparative evidence and ongoing trials The Cochrane systematic review of bariatric surgery that had informed the original By‐Band grant application was updated and published in 201415. The authors concluded that there remained a lack of high‐quality evidence for the different types of bariatric surgery. The trials reviewed were primarily small single‐centre studies performed outside the UK. They were at risk of bias and focused predominantly on short‐term outcomes up to 1 year. Interrogation of clinical trials registries identified three ongoing trials evaluating sleeve gastrectomy16, 17, 18. These were also small, non‐UK, single‐centre studies. The evidence from completed trials, and the evidence that would be generated from ongoing trials, was judged to be inadequate to draw any definitive conclusion about the effectiveness of the three types of operation. Expert opinion The proposal to adapt the trial was discussed at a special session of the British Obesity and Metabolic Surgery Society annual meeting in January 2014. Those present supported it, and additional centres were recruited. The independent trial steering and data monitoring committees supported the adaptation. Intervention stability

The proposal to adapt the trial was discussed at a special session of the British Obesity and Metabolic Surgery Society annual meeting in January 2014. Those present supported it, and additional centres were recruited. The independent trial steering and data monitoring committees supported the adaptation. Intervention stability At the time when the trial was first designed there was little consensus regarding sleeve gastrectomy itself and how it should be performed and followed up. By the time of the review, best practice guidelines had been published and the fourth international consensus summit for sleeve gastrectomy had taken place19, 20. Meetings confirmed that the procedure was safe, and agreed standards for its conduct were documented. Research question The proposal for adaptation to a three‐group trial was reviewed and accepted by the funder. The original research question, (1) below, was expanded in the three‐group trial to answer three questions based on co‐primary outcomes: Does gastric bypass lead to better quality of life and at least as good weight loss as gastric band (original question)? Does sleeve gastrectomy lead to better quality of life and at least as good weight loss as gastric band (new question)? Does sleeve gastrectomy lead to better quality of life and at least as good weight loss as gastric bypass (new question)?

The proposal for adaptation to a three‐group trial was reviewed and accepted by the funder. The original research question, (1) below, was expanded in the three‐group trial to answer three questions based on co‐primary outcomes: Does gastric bypass lead to better quality of life and at least as good weight loss as gastric band (original question)? Does sleeve gastrectomy lead to better quality of life and at least as good weight loss as gastric band (new question)? Does sleeve gastrectomy lead to better quality of life and at least as good weight loss as gastric bypass (new question)? When calculating the sample size for the adapted design, the same assumptions as for the original calculations were used10. Because the number of hypotheses was increased threefold, the significance levels were adjusted from 5 to 2 per cent for the two‐sided statistical tests of superiority and from 2·5 to 1 per cent for the one‐sided statistical tests of non‐inferiority. Under these assumptions, the revised total sample size for a trial with equal allocation across the three treatment groups was 1341 (447 per group) (Fig. 1). Figure 1 By‐Band‐Sleeve trial design BJS-10562-FIG-0001-cThe formal proposal for the adaptation was submitted at the same time as the request to progress from the pilot to the main trial. Approval and additional funding to support the adaptation was granted in October 2014.

When calculating the sample size for the adapted design, the same assumptions as for the original calculations were used10. Because the number of hypotheses was increased threefold, the significance levels were adjusted from 5 to 2 per cent for the two‐sided statistical tests of superiority and from 2·5 to 1 per cent for the one‐sided statistical tests of non‐inferiority. Under these assumptions, the revised total sample size for a trial with equal allocation across the three treatment groups was 1341 (447 per group) (Fig. 1). Figure 1 By‐Band‐Sleeve trial design BJS-10562-FIG-0001-cThe formal proposal for the adaptation was submitted at the same time as the request to progress from the pilot to the main trial. Approval and additional funding to support the adaptation was granted in October 2014. Implementation of the adaptation Practical and logistical changes necessary to adapt a surgical trial are outlined in Table 3. All patient‐facing study documents had to be revised and submitted as a major amendment to the research ethics committee for approval. Data collection forms needed to be extended to capture the additional surgical procedure. The randomization scheme had to be revised to allow future participants to be allocated to one of the three surgical procedures. The allocation ratio was modified to ensure there would be approximately equal numbers of participants per group at the end of the trial, to provide the same power for all comparisons. Adjustment to the allocation ratio was determined by simulation to ensure balance was achieved, while at the same time ensuring that allocation would be unpredictable to prevent allocation bias. The allocation ratio chosen was specific to each study centre, depending on the numbers recruited before the adaptation and the projected recruitment rate going forward. Each centre needed guidance on: when to collect data using the By‐Band data collection forms and when to collect data using the revised By‐Band‐Sleeve forms; when to start using the revised patient information leaflets; and when the randomization system and database would be switched over and how to manage participants who had given consent but had not been randomized at the time of the switch (Table 3).

forms and when to collect data using the revised By‐Band‐Sleeve forms; when to start using the revised patient information leaflets; and when the randomization system and database would be switched over and how to manage participants who had given consent but had not been randomized at the time of the switch (Table 3). Table 3 Practical and logistical considerations for optimizing adaptation of a two‐group to three‐group surgical trial

forms and when to collect data using the revised By‐Band‐Sleeve forms; when to start using the revised patient information leaflets; and when the randomization system and database would be switched over and how to manage participants who had given consent but had not been randomized at the time of the switch (Table 3). Table 3 Practical and logistical considerations for optimizing adaptation of a two‐group to three‐group surgical trial Consideration Potential problems Suggestions to optimize adaptation Study logo and acronym If the acronym and study logo are specific to the two groups (e.g. By‐Band), they will need to be changed (e.g. to By‐Band‐Sleeve). This may: (1) create changes to study and website materials that would not otherwise be necessary, and (2) jeopardize the branding of the study At the outset, select an acronym and study logo that could encompass a future adaptation Data collection forms New and updated case report forms needed. This may: (1) take a lot of time and involve significant changes to the forms, and (2) lead to major changes to the database Design case report forms in logical sections. For example, separate information and adverse event data that are common for all surgical procedures from that which is specific to one procedure. Organizing the data collection in this way can help minimize the database changes needed Allocation of procedures If equal allocation of participants to groups is applied after the adaptation, this will result in unbalanced numbers of participants in each group at the end of the trial Close working with senior statisticians is recommended. If the allocation ratio is adjusted, simulation of future recruitment at each centre is needed to ensure that the allocation remains concealed and cannot be predicted. Rigorous testing of all changes needs to be performed Alternatively, the allocation ratio can be adjusted so that the numbers per group are approximately equal at the end of the trial when the target sample size is reached. In modifying the allocation ratio, it is necessary: (1) to consider the projected recruitment rate and numbers recruited already in each centre before the adaptation, and (2) to assess the impact the change of ratio will have on future recruitment This work would need to be included in the overall cost of the adaptation Transition pathway for participants at different stages of trial Information provision for participants during trial adaption needs consideration, especially for those part way through recruitment.

change of ratio will have on future recruitment This work would need to be included in the overall cost of the adaptation Transition pathway for participants at different stages of trial Information provision for participants during trial adaption needs consideration, especially for those part way through recruitment. Each (potential) participant will be at one of five stages; they may have: (1) been sent information about the two‐group study but not yet had a consultation, (2) discussed the two‐group study but not yet consented to randomization, (3) been consented to the two‐group study but not yet randomized, (4) been randomized and awaiting surgery, or (5) undergone surgery and be in follow‐up It is recommended that the trial team discusses the transition with individual centres. It is particularly important to agree the process for patients at stages (1) and (3). In By‐Band‐Sleeve, these patients had had a consultation in which two procedures were discussed and it was not feasible for them to have a further consultation. Randomization for these patients was therefore handled centrally by the trials unit to prevent them being allocated to the new procedure The adapted By‐Band‐Sleeve study opened to recruitment in August 2015, 13 months after approval from the funder to proceed with the adaptation. The transition went smoothly and the study is recruiting steadily to all three procedures (Fig. 2). In February 2017, the 50 per cent recruitment milestone was reached (671 patients randomized into the trial). All study centres continue to be supported with the QRI13.

Collaborators Other By‐Band‐Sleeve study investigators (principal investigator denoted by an asterisk): S. Agrawal*, S. Ajaz, Y. Koak (Homerton University Hospital, London); A. Ahmed*, N. Fakih, S. Hakky, K. Moorthy, S. Purkayastha (Imperial College Healthcare NHS Trust, London); S. Awad, K. Fareed, P. Leeder* (Royal Derby Hospital, Derby); S. Balupuri, W. Carr, N. Jennings*, P. Small (Sunderland Royal Hospital, Sunderland); R. Byrom, N. Davies* (Royal Bournemouth and Christchurch Hospitals); N. Carter*, B. Knight, S. Somers (Queen Alexandra Hospital, Portsmouth); V. Charalampakis, M. Daskalakis, R. Nijar, M. Richardson, R. Singhal*, P. Super (Heart of England NHS Foundation Trust, Birmingham); M. Clarke, A. Cota, I. Finlay* (Royal Cornwall Hospital, Truro); S. Dexter, J. Hayden*, S. Mehta, A. Sarela (St James's University Hospital, Leeds); J. Kelly (University Hospital Southampton, Southampton); D. Mahon, H. Noble (Musgrove Park Hospital, Taunton). Editor's comments Acknowledgements This study was funded by the National Institute of Health Research Health Technology Assessment Programme (HTA 09/127/53) and support from the Medical Research Council ConDuCT‐II (Collaboration and innovation for Difficult and Complex randomized controlled Trials In Invasive procedures) Hub for Trials Methodology Research (MR/K025643/1). Disclosure: The authors declare no conflict of interest.

Introduction Breast cancer is the most common cancer among women, with about 1·67 million new patients diagnosed in 20121. With 522 000 annual breast cancer‐related deaths estimated worldwide, it is the leading cause of cancer‐related death in women in developing countries, and second only to lung cancer in more developed regions1. Surgery, either breast‐conserving surgery (BCS) or mastectomy, is the primary treatment for breast cancer. Despite its therapeutic intent, surgery causes physiological stress, which, along with anaesthesia2, can lead to transient immunosuppression during the perioperative period3. Such transient immunosuppression may lead to poorer immune detection of cancer cells3.

or mastectomy, is the primary treatment for breast cancer. Despite its therapeutic intent, surgery causes physiological stress, which, along with anaesthesia2, can lead to transient immunosuppression during the perioperative period3. Such transient immunosuppression may lead to poorer immune detection of cancer cells3. Postoperative bleeding requiring reoperation occurs in up to 4 per cent of women undergoing surgery for breast cancer4. Depending on the age of the patient and extent of primary surgery (mastectomy versus BCS)5, the use of certain prescription drugs (such as selective serotonin reuptake inhibitors (SSRIs) or glucocorticoids) increases the risk of postoperative bleeding requiring reoperation5, 6. However, there is no evidence of an effect of SSRIs and glucocorticoid use on breast cancer recurrence5, 6, 7. Bleeding activates platelets, which can bind tumour cells, promoting immune evasion, angiogenesis, tumour cell survival and metastatic growth8. Cancer is associated with a hypercoagulable state9, 10, with heightened platelet activation and a correlation with poor prognosis11. Thus, patients with breast cancer who develop postoperative bleeding requiring reoperation may be at increased risk of breast cancer recurrence. This cohort study was conducted to investigate the association between bleeding occurring within 14 days of primary breast cancer surgery and the rate of recurrence among patients with breast cancer in Denmark.

Postoperative bleeding requiring reoperation occurs in up to 4 per cent of women undergoing surgery for breast cancer4. Depending on the age of the patient and extent of primary surgery (mastectomy versus BCS)5, the use of certain prescription drugs (such as selective serotonin reuptake inhibitors (SSRIs) or glucocorticoids) increases the risk of postoperative bleeding requiring reoperation5, 6. However, there is no evidence of an effect of SSRIs and glucocorticoid use on breast cancer recurrence5, 6, 7. Bleeding activates platelets, which can bind tumour cells, promoting immune evasion, angiogenesis, tumour cell survival and metastatic growth8. Cancer is associated with a hypercoagulable state9, 10, with heightened platelet activation and a correlation with poor prognosis11. Thus, patients with breast cancer who develop postoperative bleeding requiring reoperation may be at increased risk of breast cancer recurrence. This cohort study was conducted to investigate the association between bleeding occurring within 14 days of primary breast cancer surgery and the rate of recurrence among patients with breast cancer in Denmark. Methods This study was approved by the Danish Data Protection Agency (Record 2007‐58‐0010), the Danish Medicines Agency and the Danish Breast Cancer Group (DBCG). The study is based on routinely collected registry data and according to Danish regulations does therefore not require separate ethical approval.

This cohort study was conducted to investigate the association between bleeding occurring within 14 days of primary breast cancer surgery and the rate of recurrence among patients with breast cancer in Denmark. Methods This study was approved by the Danish Data Protection Agency (Record 2007‐58‐0010), the Danish Medicines Agency and the Danish Breast Cancer Group (DBCG). The study is based on routinely collected registry data and according to Danish regulations does therefore not require separate ethical approval. Setting This was a nationwide cohort study using Danish population‐based registries. Denmark's national health service provides tax‐supported healthcare to Danish citizens and permanent residents, including unrestricted access to hospital care and partial reimbursement for prescribed medications12, 13. At birth or immigration, each person is assigned a unique civil personal registration number (CPR number) that allows unambiguous individual‐level linkage among all Danish administrative and population‐based registries, including medical registries13.

tal care and partial reimbursement for prescribed medications12, 13. At birth or immigration, each person is assigned a unique civil personal registration number (CPR number) that allows unambiguous individual‐level linkage among all Danish administrative and population‐based registries, including medical registries13. Source population and data collection The registry of the DBCG14, 15 and the Danish National Patient Register (DNPR) was used to identify all women with an incident diagnosis of operable stage I–III breast cancer who underwent BCS or mastectomy between 1996 and 2008. To ensure correct retrieval of the exposure, defined as reoperation for postoperative bleeding within 14 days following primary breast cancer‐directed surgery, patients were considered eligible for inclusion in the study if there was a difference of 1 day or less between the recorded date of primary surgery in the DNPR and DBCG database.

etrieval of the exposure, defined as reoperation for postoperative bleeding within 14 days following primary breast cancer‐directed surgery, patients were considered eligible for inclusion in the study if there was a difference of 1 day or less between the recorded date of primary surgery in the DNPR and DBCG database. The DBCG has registered almost all women with invasive breast cancer in Denmark since 197716. Data on tumour and patient characteristics are collected prospectively by treating physicians. The completeness of registration is approximately 95 per cent16. Patients registered in the DBCG database undergo regular follow‐up examinations aimed at detecting recurrent disease17. The following information was obtained from the DBCG database: age and menopausal status at diagnosis, type of surgery, WHO histological tumour type and grade, lymph node status, tumour size, oestrogen receptor (ER) status, receipt of adjuvant chemotherapy, endocrine therapy (ET) and/or radiation therapy, and date and site of recurrence. The DNPR has collected data on all non‐psychiatric hospital admissions since 1977, and on all outpatient and emergency contacts since 1995. Data in the DNPR include the CPR number, one primary diagnosis, and one or more secondary diagnoses classified according to the ICD, as well as data on diagnostic and surgical procedures18.

has collected data on all non‐psychiatric hospital admissions since 1977, and on all outpatient and emergency contacts since 1995. Data in the DNPR include the CPR number, one primary diagnosis, and one or more secondary diagnoses classified according to the ICD, as well as data on diagnostic and surgical procedures18. The DNPR was used to retrieve information on reoperation for bleeding after surgery (Table S1, supporting information) within 14 days following primary surgery for breast cancer. Information was retrieved from the DNPR on potentially confounding other diseases (co‐morbidity) registered up to 10 years before the breast cancer diagnosis. These were summarized using the Charlson Co‐morbidity Index (CCI)19, modified to exclude breast cancer diagnoses. Co‐morbidity prevalent on the date of breast cancer surgery was studied in order to detect diseases that could potentially confound or modify the association between bleeding after surgery and a later breast cancer recurrence20, 21, 22, 23. These included: diabetes, liver disease, chronic pulmonary disease, peripheral and cerebral vascular disease, any other cancer, myocardial infarction and congestive heart failure (Table S2, supporting information). Information on death and emigration was retrieved from the Civil Registration System (CRS). The CRS, established in 1968, contains information on the vital status of all Danish citizens; it is updated daily12.

ease, any other cancer, myocardial infarction and congestive heart failure (Table S2, supporting information). Information on death and emigration was retrieved from the Civil Registration System (CRS). The CRS, established in 1968, contains information on the vital status of all Danish citizens; it is updated daily12. The National Prescription Registry has automatically recorded detailed information on all prescriptions redeemed at Danish community pharmacies since 199524. Information is transferred electronically into the registry at the time of prescription redemption, so the validity of the registry is extremely high25. The registry contains detailed information on dispensed prescriptions, including full Anatomical Therapeutic Chemical codes, and date and quantity dispensed24. Data on drugs that potentially confound the association between bleeding and recurrence were retrieved, including simvastatin and aspirin, which may modify breast cancer prognosis26, 27, and hormone replacement therapy (HRT) (Table S3, supporting information).

rapeutic Chemical codes, and date and quantity dispensed24. Data on drugs that potentially confound the association between bleeding and recurrence were retrieved, including simvastatin and aspirin, which may modify breast cancer prognosis26, 27, and hormone replacement therapy (HRT) (Table S3, supporting information). Variables analysed Age at diagnosis was categorized into decades. Histological grade was defined as low, moderate or high, based on WHO histological tumour type28. Stage was classified as I, II or III according to the UICC classification (6th edition)29. Lymph node status was defined according to number of involved nodes (0, 1–3, 4 or more). Tumour size was categorized as 20 mm or less, or over 20 mm. ER and adjuvant ET were summarized as: ER+/ET+, ER–/ET–, ER+/ET– or ER−/ET+. Surgery type was either mastectomy or BCS. Treatment with adjuvant chemotherapy was categorized dichotomously. Menopausal status at diagnosis was either premenopausal or postmenopausal, classified according to the DBCG.

or less, or over 20 mm. ER and adjuvant ET were summarized as: ER+/ET+, ER–/ET–, ER+/ET– or ER−/ET+. Surgery type was either mastectomy or BCS. Treatment with adjuvant chemotherapy was categorized dichotomously. Menopausal status at diagnosis was either premenopausal or postmenopausal, classified according to the DBCG. Simvastatin and aspirin use were modelled as time‐varying co‐variables. Longitudinal prescription data were used to define time‐updated exposure to these drugs. For each prescription, prescription duration was calculated as pack size (number of pills per pack) multiplied by the number of packages redeemed, assuming that a single pill was taken each day. In defining continuous use, a gap of 30 days was allowed from the end of one prescription (prescription start date + prescription duration) until the start of a new prescription. If a new prescription was redeemed within this window, then exposure was assumed to continue; if not, the patient was considered to have stopped the drug at the end of the 30‐day grace period. The patient could later restart if there were further prescriptions. Finally, the resulting time‐updated current medical exposure variable lagged by 1 year to allow the effect of the drug to accrue, as any effects on cancer are likely to be delayed, and to minimize confounding by indication. HRT was recorded as a baseline co‐variable among women with at least 1 year of prescription history.

lly, the resulting time‐updated current medical exposure variable lagged by 1 year to allow the effect of the drug to accrue, as any effects on cancer are likely to be delayed, and to minimize confounding by indication. HRT was recorded as a baseline co‐variable among women with at least 1 year of prescription history. Breast cancer recurrence was defined according to the DBCG as any local, regional or distant recurrence, or cancer of the contralateral breast up to 10 years after the primary diagnosis14. Follow‐up began 14 days after primary breast cancer surgery (registered in the DNPR) and continued until breast cancer recurrence, death, emigration, 10 years of follow‐up or 1 January 2013 (end of the study period), whichever came first. Statistical analysis The proportion of patients with breast cancer who did or did not undergo reoperation for bleeding after surgery was calculated, by patient, tumour and treatment characteristics. Incidence rates (IRs) of recurrence per 1000 person‐years were calculated, and the 5‐ and 10‐year cumulative incidence of recurrence was estimated according to whether reoperation for bleeding after primary surgery had been undertaken. IRs were also categorized by time after surgery; recurrences developing within 2 years represented very early recurrence, those diagnosed at 2–5 years comprised early recurrence, and recurrences detected after 5 years represented late recurrence30. The proportion of patients with breast cancer receiving mastectomy and BCS over time was calculated, as was the proportion needing a further operation over time.

Statistical analysis The proportion of patients with breast cancer who did or did not undergo reoperation for bleeding after surgery was calculated, by patient, tumour and treatment characteristics. Incidence rates (IRs) of recurrence per 1000 person‐years were calculated, and the 5‐ and 10‐year cumulative incidence of recurrence was estimated according to whether reoperation for bleeding after primary surgery had been undertaken. IRs were also categorized by time after surgery; recurrences developing within 2 years represented very early recurrence, those diagnosed at 2–5 years comprised early recurrence, and recurrences detected after 5 years represented late recurrence30. The proportion of patients with breast cancer receiving mastectomy and BCS over time was calculated, as was the proportion needing a further operation over time. Cox regression models with time from start of follow‐up as the underlying time scale were used to compute crude and adjusted hazard ratios (HRs) for recurrence and associated 95 per cent confidence intervals for reoperation for postoperative bleeding. To model the cause‐specific hazard, patients who died without a breast cancer recurrence were censored at the date of death. The adjusted model included the following potential confounders: age group at diagnosis, menopausal status, receipt of chemotherapy, lymph node status, tumour size, tumour grade, type of primary surgery, ER/ET status, co‐morbidity, baseline HRT, and simvastatin and aspirin use after diagnosis (coded as time‐varying co‐variables lagging by 1 year). The analyses were stratified by age, receipt of chemotherapy, UICC stage and type of primary surgery. Crude and adjusted HRs according to site of recurrence were calculated.

ER/ET status, co‐morbidity, baseline HRT, and simvastatin and aspirin use after diagnosis (coded as time‐varying co‐variables lagging by 1 year). The analyses were stratified by age, receipt of chemotherapy, UICC stage and type of primary surgery. Crude and adjusted HRs according to site of recurrence were calculated. The following sensitivity analyses were conducted: changing the 14‐day window for reoperation and start of follow‐up to 7 days after primary surgery; changing the inclusion criteria from no more than 1 day difference between the recorded date of primary surgery in the DNPR and the DBCG database to no more than 14 days and no more than 31 days; changing the study population to include only patients with stage I and II disease at diagnosis; and excluding patients with previous cancers. Analyses were performed using Stata® version 13 (StataCorp, College Station, Texas, USA). Results A total of 33 162 patients with breast cancer who underwent BCS or mastectomy between 1996 and 2008 were identified. The cohort consisted of 30 711 women after exclusion of 2425 women with more than 1 day difference in the date of surgery, or inconsistency in type of surgery, between the DNPR and the DBCG database, and 26 women who died or had an event registered before the start of follow‐up (within 14 days after primary breast cancer surgery). The proportion of patients treated with BCS versus mastectomy increased in recent years accompanied by a decline in the rate of reoperation. Median follow‐up was 7·0 years.

DBCG database, and 26 women who died or had an event registered before the start of follow‐up (within 14 days after primary breast cancer surgery). The proportion of patients treated with BCS versus mastectomy increased in recent years accompanied by a decline in the rate of reoperation. Median follow‐up was 7·0 years. Reoperation after surgery Overall, 767 patients (2·5 per cent) had at least one reoperation within 14 days of the primary surgery. Compared with women who were not reoperated, a higher proportion of patients who underwent reoperation were postmenopausal (75·1 versus 72·5 per cent), and had co‐morbid disease (CCI score of at least 1: 23·2 versus 20·1 per cent), a history of cerebrovascular disease (5·2 versus 3·4 per cent) and moderate‐grade tumours (13·0 versus 11·0 per cent) (Tables 1 and 2). Reoperated patients were more likely to have undergone mastectomy than BCS as primary surgery (69·3 versus 57·9 per cent) and less likely to receive chemotherapy (28·7 versus 33·6 per cent). A higher proportion of patients without reoperation had stage III cancer (18·1 versus 14·0 per cent). Overall, 21·0 per cent of women in the breast cancer cohort had been prescribed aspirin or simvastatin during follow‐up, and 41·6 per cent had been prescribed HRT before the breast cancer diagnosis. Reoperated patients were more likely to be concurrent aspirin users. Table 1 Baseline characteristics of 30 711 patients diagnosed with stage I–III breast cancer in Denmark, 1996–2008, according to reoperation for postoperative bleeding

Reoperation after surgery Overall, 767 patients (2·5 per cent) had at least one reoperation within 14 days of the primary surgery. Compared with women who were not reoperated, a higher proportion of patients who underwent reoperation were postmenopausal (75·1 versus 72·5 per cent), and had co‐morbid disease (CCI score of at least 1: 23·2 versus 20·1 per cent), a history of cerebrovascular disease (5·2 versus 3·4 per cent) and moderate‐grade tumours (13·0 versus 11·0 per cent) (Tables 1 and 2). Reoperated patients were more likely to have undergone mastectomy than BCS as primary surgery (69·3 versus 57·9 per cent) and less likely to receive chemotherapy (28·7 versus 33·6 per cent). A higher proportion of patients without reoperation had stage III cancer (18·1 versus 14·0 per cent). Overall, 21·0 per cent of women in the breast cancer cohort had been prescribed aspirin or simvastatin during follow‐up, and 41·6 per cent had been prescribed HRT before the breast cancer diagnosis. Reoperated patients were more likely to be concurrent aspirin users. Table 1 Baseline characteristics of 30 711 patients diagnosed with stage I–III breast cancer in Denmark, 1996–2008, according to reoperation for postoperative bleeding All patients Recurrence Total person‐years Reoperation (n = 767) No reoperation (n = 29 944) Reoperation (n = 126) No reoperation (n = 4643) Reoperation No reoperation Overall 5241 200 685 Age at diagnosis (years) ≤ 29 0 (0) 98 (0·3) 0 (0) 32 (0·7) 0 578 30–39 30 (3·9) 1357 (4·5) 8 (6·3) 311 (6·7) 217 9073 40–49 112 (14·6) 5070 (16·9) 20 (15·9) 838 (18·0) 850 36 701 50–59 237 (30·9) 8962 (29·9) 43 (34·1) 1455 (31·3) 1683 63 381 60–69 230 (30·0) 9258 (30·9) 31 (24·6) 1357 (29·2) 1550 61 232 70–79 131 (17·1) 4254 (14·2) 23 (18·3) 576 (12·4) 816 25 602 ≥ 80 27 (3·5) 945 (3·2) 1 (0·8) 74 (1·6) 124 4118 Menopausal status at diagnosis Premenopausal 191 (24·9) 8226 (27·5) 36 (28·6) 1380 (29·7) 1411 59 317 Postmenopausal 576 (75·1) 21 704 (72·5) 90 (71·4) 3262 (70·3) 3830 141 296 Missing 0 (0) 14 (0·0) 0 (0) 1 (0·0) 0 72 Charlson Co‐morbidity Index score 0 589 (76·8) 23 913 (79·9) 110 (87·3) 3879 (83·5) 4152 165 150 1 107 (14·0) 3357 (11·2) 12 (9·5) 446 (9·6) 701 20 666 2 47 (6·1) 1683 (5·6) 2 (1·6) 209 (4·5) 265 9947 ≥ 3 24 (3·1) 991 (3·3) 2 (1·6) 109 (2·3) 123 4922 Specific co‐morbidities Myocardial infarction 15 (2·0) 356 (1·2) 1 (0·8) 42 (0·9) 79 1987 Congestive heart failure 18 (2·3) 385 (1·3) 1 (0·8) 35 (0·8) 74 1937 Vascular disease 21 (2·7) 518 (1·7) 1 (0·8) 68 (1·5) 127 2818 Cerebrovascular disease 40 (5·2) 1013 (3·4) 1 (0·8) 114 (2·5) 274 5597 Chronic pulmonary disease 39 (5·1) 1459 (4·9) 7 (5·6) 174 (3·7) 211 8467 Diabetes types 1 and 2 20 (2·6) 811 (2·7) 1 (0·8) 114 (2·5) 112 4491 Diabetes with organ damage 8 (1·0) 346 (1·2) 1 (0·8) 41 (0·9) 41 1824 Liver disease 10 (1·3) 250 (0·8) 1 (0·8) 33 (0·7) 29 1296 Any other cancer 24 (3·1) 1286 (4·3) 1 (0·8) 154 (3·3) 152 7360 Values in parentheses are percentages.

) 174 (3·7) 211 8467 Diabetes types 1 and 2 20 (2·6) 811 (2·7) 1 (0·8) 114 (2·5) 112 4491 Diabetes with organ damage 8 (1·0) 346 (1·2) 1 (0·8) 41 (0·9) 41 1824 Liver disease 10 (1·3) 250 (0·8) 1 (0·8) 33 (0·7) 29 1296 Any other cancer 24 (3·1) 1286 (4·3) 1 (0·8) 154 (3·3) 152 7360 Values in parentheses are percentages. Table 2 Baseline tumour characteristics and treatments of 30 711 patients diagnosed with stage I–III breast cancer in Denmark, 1996–2008, according to reoperation for postoperative bleeding

) 174 (3·7) 211 8467 Diabetes types 1 and 2 20 (2·6) 811 (2·7) 1 (0·8) 114 (2·5) 112 4491 Diabetes with organ damage 8 (1·0) 346 (1·2) 1 (0·8) 41 (0·9) 41 1824 Liver disease 10 (1·3) 250 (0·8) 1 (0·8) 33 (0·7) 29 1296 Any other cancer 24 (3·1) 1286 (4·3) 1 (0·8) 154 (3·3) 152 7360 Values in parentheses are percentages. Table 2 Baseline tumour characteristics and treatments of 30 711 patients diagnosed with stage I–III breast cancer in Denmark, 1996–2008, according to reoperation for postoperative bleeding All patients Recurrence Total person‐years Reoperation (n = 767) No reoperation (n = 29 944) Reoperation (n = 126) No reoperation (n = 4643) Reoperation No reoperation Overall 5241 200 685 UICC stage I 284 (37·0) 10 852 (36·2) 36 (28·6) 1157 (24·9) 2095 78 669 II 367 (47·8) 13 465 (45·0) 52 (41·3) 1844 (39·7) 2539 92 554 III 107 (14·0) 5406 (18·1) 38 (30·2) 1620 (34·9) 550 28 262 Missing 9 (1·2) 221 (0·7) 0 (0) 22 (0·5) 57 1200 Tumour size (mm) ≤ 20 438 (57·1) 17 190 (57·4) 57 (45·2) 2026 (43·6) 3160 121 891 > 20 321 (41·9) 12 544 (41·9) 67 (53·2) 2574 (55·4) 2021 77 267 Missing 8 (1·0) 210 (0·7) 2 (1·6) 43 (0·9) 60 1528 Lymph node status Negative 405 (52·8) 15 522 (51·8) 51 (40·5) 1807 (38·9) 2963 111 142 1–3 postive nodes 255 (33·2) 9147 (30·5) 38 (30·2) 1266 (27·3) 1731 62 306 ≥ 4 positive nodes 104 (13·6) 5151 (17·2) 37 (29·4) 1563 (33·7) 533 26 735 Missing 3 (0·4) 124 (0·4) 0 (0) 7 (0·2) 14 502 Histological grade Low 621 (81·0) 24 522 (81·9) 105 (83·3) 3846 (82·8) 4218 163 024 Moderate 100 (13·0) 3301 (11·0) 11 (8·7) 548 (11·8) 714 22 769 High 44 (5·7) 1992 (6·7) 9 (7·1) 222 (4·8) 297 13 972 Missing 2 (0·3) 129 (0·4) 1 (0·8) 27 (0·6) 11 920 ER/adjuvant ET status ER–/ET– 134 (17·5) 5818 (19·4) 21 (16·7) 1174 (25·3) 892 35 750 ER+/ET– 184 (24·0) 7143 (23·9) 25 (19·8) 1087 (23·4) 1399 52 922 ER+/ET+ 420 (54·8) 15 985 (53·4) 76 (60·3) 2177 (46·9) 2736 104 739 ER–/ET+ 5 (0·7) 181 (0·6) 1 (0·8) 27 (0·6) 39 1330 Unknown 24 (3·1) 817 (2·7) 3 (2·4) 178 (3·8) 174 5944 Type of primary surgery Mastectomy 373 (48·6) 10 838 (36·2) 65 (51·6) 1867 (40·2) 2527 74 573 Mastectomy + RT 159 (20·7) 6486 (21·7) 34 (27·0) 1445 (31·1) 1074 41 563 BCS + RT 235 (30·6) 12 620 (42·1) 27 (21·4) 1331 (28·7) 1639 84 550 Adjuvant chemotherapy Yes 220 (28·7) 10 075 (33·6) 33 (26·2) 1628 (35·1) 1509 65 009 No 547 (71·3) 19 869 (66·4) 93 (73·8) 3015 (64·9) 3732 135 676 HRT before diagnosis Yes 316 (41·2) 12 452 (41·6) 37 (29·4) 1634 (35·2) 2220 83 790 No 451 (58·8) 17 492 (58·4) 89 (70·6) 3009 (64·8) 3021 116 896 Drugs taken during study period Simvastatin 148 (19·3) 6286 (21·0) 7 (5·6) 349 (7·5) 538 22 527 Aspirin (high and low doses) 190 (24·8) 6233 (20·8) 15 (11·9) 556 (12·0) 532 17 ,613 V

e diagnosis Yes 316 (41·2) 12 452 (41·6) 37 (29·4) 1634 (35·2) 2220 83 790 No 451 (58·8) 17 492 (58·4) 89 (70·6) 3009 (64·8) 3021 116 896 Drugs taken during study period Simvastatin 148 (19·3) 6286 (21·0) 7 (5·6) 349 (7·5) 538 22 527 Aspirin (high and low doses) 190 (24·8) 6233 (20·8) 15 (11·9) 556 (12·0) 532 17 ,613 V alues in parentheses are percentages. ER, oestrogen receptor; ET, endocrine therapy; RT, radiotherapy; BCS, breast‐conserving surgery; HRT, hormone replacement therapy.

e diagnosis Yes 316 (41·2) 12 452 (41·6) 37 (29·4) 1634 (35·2) 2220 83 790 No 451 (58·8) 17 492 (58·4) 89 (70·6) 3009 (64·8) 3021 116 896 Drugs taken during study period Simvastatin 148 (19·3) 6286 (21·0) 7 (5·6) 349 (7·5) 538 22 527 Aspirin (high and low doses) 190 (24·8) 6233 (20·8) 15 (11·9) 556 (12·0) 532 17 ,613 V alues in parentheses are percentages. ER, oestrogen receptor; ET, endocrine therapy; RT, radiotherapy; BCS, breast‐conserving surgery; HRT, hormone replacement therapy. Recurrence after reoperation for bleeding Overall, 4769 patients developed breast cancer recurrence during follow‐up. The IR of recurrence was 24·0 (95 per cent c.i. 20·2 to 28·6) and 23·1 (22·5 to 23·8) per 1000 person‐years for reoperated and non‐reoperated patients respectively (Table 3). Regardless of reoperation status, the incidence rate was higher in the first 2 years after surgery, followed by a decrease (Table S4, supporting information). The 1‐year IR of recurrence was 29·1 (13·1 to 44·1) and 21·3 (19·7 to 23·1) per 1000 person‐years for reoperated and non‐reoperated patients respectively. The IR of recurrence in the second year after primary surgery was 40·7 (28·3 to 58·6) per 1000 person‐years for reoperated patients and 34·7 (32·6 to 36·9) per 1000 person‐years for non‐reoperated patients. After 5 years of follow‐up, the IRs for patients who did and did not undergo reoperation were similar: 27·7 (22·6 to 33·9) and 26·9 (26·1 to 27·8) per 1000 person‐years respectively. The 5‐year cumulative incidence of recurrence was 12·8 and 12·5 per cent for patients with and without reoperation respectively; the 10‐year cumulative incidence of recurrence was 19·9 per cent for reoperated patients and 18·9 per cent for non‐reoperated patients (Table S5, supporting information).

‐years respectively. The 5‐year cumulative incidence of recurrence was 12·8 and 12·5 per cent for patients with and without reoperation respectively; the 10‐year cumulative incidence of recurrence was 19·9 per cent for reoperated patients and 18·9 per cent for non‐reoperated patients (Table S5, supporting information). Table 3 Incidence rates and hazard ratios for breast cancer recurrence, according to reoperation for postoperative bleeding, among 30 711 women diagnosed with stage I–III breast cancer in Denmark, 1996–2008 with follow‐up to 31 December 2012 No. of recurrences Person‐years Crude incidence rate (per 100 000 person‐years) Unadjusted hazard ratio Adjusted hazard ratio* Overall (reoperation within 14 days)† No reoperation 4643 200 685 23·1 (22·5, 23·8) 1·00 (reference) 1·00 (reference) Reoperation 126 5241 24·0 (20·2, 28·6) 1·05 (0·88, 1·25) 1·06 (0·89, 1·26) Reoperation within 7 days† No reoperation 4650 201 520 23·1 (22·4, 23·7) 1·00 (reference) 1·00 (reference) Reoperation 121 4995 24·2 (20·3, 28·9) 1·06 (0·88, 1·27) 1·08 (0·91, 1·30) Values in parentheses are 95 per cent confidence intervals. Hazard ratios with 95 per cent confidence intervals are shown.

No reoperation 4643 200 685 23·1 (22·5, 23·8) 1·00 (reference) 1·00 (reference) Reoperation 126 5241 24·0 (20·2, 28·6) 1·05 (0·88, 1·25) 1·06 (0·89, 1·26) Reoperation within 7 days† No reoperation 4650 201 520 23·1 (22·4, 23·7) 1·00 (reference) 1·00 (reference) Reoperation 121 4995 24·2 (20·3, 28·9) 1·06 (0·88, 1·27) 1·08 (0·91, 1·30) Values in parentheses are 95 per cent confidence intervals. Hazard ratios with 95 per cent confidence intervals are shown. * Hazard ratios were adjusted for age (as a categorical variable), menopausal status at diagnosis (premenopausal, postmenopausal), lymph node status (negative, 1–3 positive nodes, at least 4 positive nodes), tumour size (20 mm or smaller, larger than 20 mm), histological grade (low, moderate, high), type of surgery, oestrogen receptor (ER) status and receipt of endocrine therapy (ET) (ER+/ET–, ER+/ET+, ER–/ET–, ER–/ET+), receipt of chemotherapy (yes, no), simvastatin use and aspirin use (both as time‐varying co‐variables lagging by 1 year), co‐morbidity, and receipt of hormone replacement therapy before diagnosis (yes, no). † The total number of patients with recurrence is not identical here because two patients died or developed a recurrence before the start of follow‐up on day 14.

* Hazard ratios were adjusted for age (as a categorical variable), menopausal status at diagnosis (premenopausal, postmenopausal), lymph node status (negative, 1–3 positive nodes, at least 4 positive nodes), tumour size (20 mm or smaller, larger than 20 mm), histological grade (low, moderate, high), type of surgery, oestrogen receptor (ER) status and receipt of endocrine therapy (ET) (ER+/ET–, ER+/ET+, ER–/ET–, ER–/ET+), receipt of chemotherapy (yes, no), simvastatin use and aspirin use (both as time‐varying co‐variables lagging by 1 year), co‐morbidity, and receipt of hormone replacement therapy before diagnosis (yes, no). † The total number of patients with recurrence is not identical here because two patients died or developed a recurrence before the start of follow‐up on day 14. Among 767 patients who underwent reoperation, there were 126 recurrences in 5241 person‐years of follow‐up. Among 29 944 women who did not undergo reoperation, there were 4643 recurrences in 200 685 person‐years of follow‐up. After adjusting for potential confounders, no association between bleeding after surgery and breast cancer recurrence was observed (adjusted HR 1·06, 95 per cent c.i. 0·89 to 1·26), regardless of time interval of exposure (7 or 14 days after primary operation) (Table 3). This lack of association did not change in sensitivity analyses in which the study population included only patients with stage I and II disease at diagnosis, patients with previous cancers were excluded, or patients with a difference in surgery date between the DNPR and DBCG database of no more than 14 days and no more than 31 days were included (Table S6, supporting information). The estimates did not vary by site of breast cancer recurrence (Fig. 1), and there was no evidence of effect modification in models stratified by age, tumour stage, type of primary surgery or receipt of chemotherapy (Fig. 2).

o more than 14 days and no more than 31 days were included (Table S6, supporting information). The estimates did not vary by site of breast cancer recurrence (Fig. 1), and there was no evidence of effect modification in models stratified by age, tumour stage, type of primary surgery or receipt of chemotherapy (Fig. 2). Figure 1 Forest plot showing associations between reoperation for postoperative bleeding and anatomical site of recurrence. Hazard ratios with 95 per cent confidence intervals are shown. Hazard ratios were adjusted for age (as a categorical variable), menopausal status at diagnosis (premenopausal, postmenopausal), lymph node status (negative, 1–3 positive nodes, at least 4 positive nodes), tumour size (20 mm or smaller, larger than 20 mm), histological grade (low, moderate, high), type of surgery, oestrogen receptor (ER) status and receipt of endocrine therapy (ET) (ER+/ET–, ER+/ET+, ER–/ET–, ER–/ET+), receipt of chemotherapy (yes, no), simvastatin use and aspirin use (both as time‐varying co‐variables lagging by 1 year), co‐morbidity, and receipt of hormone replacement therapy before diagnosis (yes, no). CNS, central nervous system

R) status and receipt of endocrine therapy (ET) (ER+/ET–, ER+/ET+, ER–/ET–, ER–/ET+), receipt of chemotherapy (yes, no), simvastatin use and aspirin use (both as time‐varying co‐variables lagging by 1 year), co‐morbidity, and receipt of hormone replacement therapy before diagnosis (yes, no). CNS, central nervous system BJS-10592-FIG-0001-cFigure 2 Forest plot showing associations between reoperation for postoperative bleeding and rate of breast cancer recurrence, stratified by age, UICC stage and type of primary therapy. Hazard ratios with 95 per cent confidence intervals are shown. Hazard ratios were adjusted for age (as a categorical variable), menopausal status at diagnosis (premenopausal, postmenopausal), lymph node status (negative, 1–3 positive nodes, at least 4 positive nodes), tumour size (20 mm or smaller, larger than 20 mm), histological grade (low, moderate, high), type of surgery, oestrogen receptor (ER) status and receipt of endocrine therapy (ET) (ER+/ET–, ER+/ET+, ER–/ET–, ER–/ET+), receipt of chemotherapy (yes, no), simvastatin use and aspirin use (both as time‐varying co‐variables lagging by 1 year), co‐morbidity, and receipt of hormone replacement therapy before diagnosis (yes, no). RT, radiotherapy; BCS, breast‐conserving surgery

of endocrine therapy (ET) (ER+/ET–, ER+/ET+, ER–/ET–, ER–/ET+), receipt of chemotherapy (yes, no), simvastatin use and aspirin use (both as time‐varying co‐variables lagging by 1 year), co‐morbidity, and receipt of hormone replacement therapy before diagnosis (yes, no). RT, radiotherapy; BCS, breast‐conserving surgery BJS-10592-FIG-0002-cDiscussion Previous research in Danish patients reported an association between re‐excision (owing to insufficient surgical margins within 2 months of BCS) and increased risk of ipsilateral breast tumour recurrence31. This finding was, however, largely explained by residual disease31. The hypothesis for the present study was that patients who undergo reoperation for postoperative bleeding would be at increased risk of ipsilateral breast tumour recurrence. No evidence was found of an association between reoperation for bleeding after surgery and later breast cancer recurrence, regardless of time interval of exposure (7 or 14 days after the primary operation). Furthermore, the estimates did not vary in analyses stratified by clinical factors, the extent of primary surgery, or by site of breast cancer recurrence. A slight increase in early recurrence among reoperated patients was observed, but the estimates are imprecise.

f exposure (7 or 14 days after the primary operation). Furthermore, the estimates did not vary in analyses stratified by clinical factors, the extent of primary surgery, or by site of breast cancer recurrence. A slight increase in early recurrence among reoperated patients was observed, but the estimates are imprecise. Research suggests that mastectomy is associated with a higher risk of intraoperative bleeding and postoperative complications than BCS32, 33, 34. However, mastectomy alone and BCS combined with radiotherapy have equal efficacy in terms of preventing breast cancer recurrence35. Results from the present study show that the association of postoperative bleeding with breast cancer recurrence is not modified by the extent of primary surgery. The associations observed for reoperation and breast cancer recurrence are not in line with those seen in patients undergoing surgery for gastrointestinal cancers. For example, intraoperative blood loss associated with surgery for upper gastrointestinal tract tumours decreases the activity of natural killer cells, which are the body's primary defence mechanism against cancer36. Research suggests that blood loss during surgery, regardless of whether blood transfusion is given, is a risk factor for peritoneal recurrence after curative resection of gastric cancer37. The mechanisms for the lack of concordance between these findings and those of the present study on breast cancer are unclear. Blood loss that can be controlled by further operation could be less extensive than blood loss that is sufficient to warrant a blood transfusion.

e after curative resection of gastric cancer37. The mechanisms for the lack of concordance between these findings and those of the present study on breast cancer are unclear. Blood loss that can be controlled by further operation could be less extensive than blood loss that is sufficient to warrant a blood transfusion. The main strengths of this study include its large size and population‐based nationwide design within a setting of universal tax‐supported healthcare. The prospective data collection reduced the potential for selection bias and ensured virtually complete follow‐up. Furthermore, comprehensive data on potential confounders, including prescription drug data, were available. The crude estimates were quite similar to the adjusted estimates, and thus there was little evidence of confounding. It is also a strength that reoperation for bleeding after surgery has a surgical procedure code and is therefore well recorded in the database. Although the positive predictive value of this specific procedure code has not been assessed in the DNPR, it is expected to be high, as hospitals in Denmark are reimbursed only after registration of surgical procedures. It is nevertheless possible that other operative procedures could be misclassified as reoperation owing to postoperative bleeding. These include the codes for reoperation for postoperative infection or reoperation owing to other causes, which may include insufficient surgical margins (Table S1, supporting information). However, the latter misclassification is likely to bias the present findings away from the lack of effect of reoperation as residual disease is well known to be associated with recurrence31. The impact of postoperative infection on later breast cancer recurrence remains unclear.

gins (Table S1, supporting information). However, the latter misclassification is likely to bias the present findings away from the lack of effect of reoperation as residual disease is well known to be associated with recurrence31. The impact of postoperative infection on later breast cancer recurrence remains unclear. Earlier studies38, 39 used blood transfusion as a proxy for perioperative bleeding. However, in the case of breast cancer surgery, perioperative bleeding does not always result in blood transfusion. Furthermore, patients who receive blood transfusions are often sicker, with disseminated cancer, and more extensive co‐morbidity. The present study has some limitations. Information was missing on the extent of postoperative bleeding, in terms of actual blood loss. There was no information available on surgical complications that may have precipitated such bleeding. Another concern is the risk of selection bias due to exclusion of patients; however, the excluded patients were younger, had less advanced disease stages at diagnosis, and were less likely to receive mastectomy and ET (Table S7, supporting information). The sensitivity analyses also showed that the inclusion of these patients did not change the present findings (Table S6, supporting information). No information was available on the type of axillary surgery. However, from 2001 to 2006 the sentinel node technique was gradually introduced in Denmark40. During the study interval, all women with metastasis of any size in the axilla were offered axillary clearance level I + II as standard care.

pporting information). No information was available on the type of axillary surgery. However, from 2001 to 2006 the sentinel node technique was gradually introduced in Denmark40. During the study interval, all women with metastasis of any size in the axilla were offered axillary clearance level I + II as standard care. Aspirin has been shown to decrease the risk of breast cancer mortality in some26, but not all41, 42 studies, whereas simvastatin has been consistently associated with a decreased risk of breast cancer recurrence/mortality43. Information on prescribed aspirin was available, but it was not possible to account for aspirin bought over the counter. Aspirin formulations are available over the counter in Denmark but, if prescribed, almost exclusively done so in low doses for cardiovascular prevention. Over‐the‐counter aspirin is available only in small packs, and supplies for regular use are usually prescribed by physicians and reimbursable via the Danish National Health Insurance System. The proportions of total sales of low‐dose aspirin dispensed by prescription, and thus captured in prescription registries, is high (92 per cent in 2012)44, so residual confounding regarding aspirin is expected to be a minor issue. No information on prescription compliance was available. In Denmark, patients pay part of the cost of redeemed prescriptions, so the estimates are likely to reflect actual use. Adjustment for prescribed aspirin and simvastatin did not change the findings. Finally, despite the large study size, reoperation for postoperative bleeding was relatively rare in this population and thus the precision of some of the estimates is low.

ed prescriptions, so the estimates are likely to reflect actual use. Adjustment for prescribed aspirin and simvastatin did not change the findings. Finally, despite the large study size, reoperation for postoperative bleeding was relatively rare in this population and thus the precision of some of the estimates is low. The findings of the present study have important clinical implications, and provide reassurance to patients and physicians that reoperation for postoperative bleeding does not increase the risk of breast cancer recurrence. Patients who undergo reoperation for bleeding are unlikely to need more aggressive adjuvant therapy. Breast cancer surgery involves a soft tissue surface and is often characterized by extensive dissection, which increases the risk of postoperative bleeding; the results may therefore be relevant to other soft tissue surgical procedures. Supporting information Additional supporting information may be found online in the supporting information tab for this article. Table S1 ICD‐10 codes for surgical procedures among women with stage I, II or III breast cancer in Denmark, 1996–2008 (Word document) Table S2 ICD codes for co‐morbidities (Word document) Table S3 Confounder drugs (Word document) Table S4 Incidence of breast cancer recurrence for patients with stage I, II or III breast cancer in Denmark, 1996–2008, according to need for reoperation for postoperative bleeding, stratified by time after surgery (Word document)

Table S2 ICD codes for co‐morbidities (Word document) Table S3 Confounder drugs (Word document) Table S4 Incidence of breast cancer recurrence for patients with stage I, II or III breast cancer in Denmark, 1996–2008, according to need for reoperation for postoperative bleeding, stratified by time after surgery (Word document) Table S5 Five‐ and 10‐year cumulative incidence of breast cancer recurrence for patients with stage I, II or III breast cancer in Denmark, 1996–2008, according to need for reoperation for postoperative bleeding (Word document) Table S6 Breast cancer recurrences and hazard ratios for patients with stage I and II breast cancer, for patients without any previous cancers, and for patients with more than 1 day between the primary surgery date registered in the Danish National Patient Register and the Danish Breast Cancer Group database (in Denmark, 1996–2008), according to need for reoperation for postoperative bleeding (Word document) Table S7 Comparison of baseline characteristics of patients retained in the cohort versus those excluded (Word document) Supporting information Table S1 ICD‐10 codes for surgical procedures among women with stage I, II or III breast cancer in Denmark, 1996–2008 Table S2 ICD codes for co‐morbidities Table S3 Confounder drugs Table S4 Incidence of breast cancer recurrence for patients with stage I, II or III breast cancer in Denmark, 1996–2008, according to need for reoperation for postoperative bleeding, stratified by time after surgery

Table S1 ICD‐10 codes for surgical procedures among women with stage I, II or III breast cancer in Denmark, 1996–2008 Table S2 ICD codes for co‐morbidities Table S3 Confounder drugs Table S4 Incidence of breast cancer recurrence for patients with stage I, II or III breast cancer in Denmark, 1996–2008, according to need for reoperation for postoperative bleeding, stratified by time after surgery Table S5 Five‐ and 10‐year cumulative incidence of breast cancer recurrence for patients with stage I, II or III breast cancer in Denmark, 1996–2008, according to need for reoperation for postoperative bleeding Table S6 Breast cancer recurrences and hazard ratios for patients with stage I and II breast cancer, for patients without any previous cancers, and for patients with more than 1 day between the primary surgery date registered in the Danish National Patient Register and the Danish Breast Cancer Group database (in Denmark, 1996–2008), according to need for reoperation for postoperative bleeding Table S7 Comparison of baseline characteristics of patients retained in the cohort versus those excluded Click here for additional data file.

Table S6 Breast cancer recurrences and hazard ratios for patients with stage I and II breast cancer, for patients without any previous cancers, and for patients with more than 1 day between the primary surgery date registered in the Danish National Patient Register and the Danish Breast Cancer Group database (in Denmark, 1996–2008), according to need for reoperation for postoperative bleeding Table S7 Comparison of baseline characteristics of patients retained in the cohort versus those excluded Click here for additional data file. Acknowledgements This work was supported by grants from the Danish Cancer Society (R117‐A7305‐14‐S7, R91‐A7311‐14‐S9; to R.N.P.), the Novo Nordisk Foundation (NNF14OC0012281; to D.P.C.‐F.), the Lundbeck Foundation (R167‐2013‐15861; to D.P.C.‐F.), the Elvira and Rasmus Riisforts Foundation (to D.P.C.‐F.), the Helga and Peter Korning Foundation (to D.P.C.‐F.), the Programme for Clinical Research Infrastructure (PROCRIN) established by the Lundbeck Foundation and the Novo Nordisk Foundation (to H.T.S.), Torben og Alice Frimodts Foundation (to R.N.P.), Ferd og Ellen Hindgauls Foundation (to R.N.P.), the Oticon Foundation (to R.N.P.), and the Foundation of 1870 (to R.N.P.). K.B. is funded by a Wellcome Trust/Royal Society Sir Henry Dale fellowship (107731/Z/15/Z). The funding agencies had no role in design of the study; the collection, analysis and interpretation of the data; the writing of the article; or the decision to submit the article for publication.

Foundation of 1870 (to R.N.P.). K.B. is funded by a Wellcome Trust/Royal Society Sir Henry Dale fellowship (107731/Z/15/Z). The funding agencies had no role in design of the study; the collection, analysis and interpretation of the data; the writing of the article; or the decision to submit the article for publication. The Department of Clinical Epidemiology is involved in studies that receive funding from various companies as research grants to (and administered by) Aarhus University. None of these studies have any relation to the present work. Disclosure: The authors declare no other conflict of interest.

Introduction Low anterior resection with total mesorectal excision is regarded as one of the optimal surgical treatments for potentially curable carcinoma of the rectum1, 2. Because of the low anastomosis close to the pelvic floor, patients often receive a temporary ileostomy at the time of the resection to reduce the risk of symptomatic anastomotic dehiscence3 and its clinical consequences3, 4, 5, 6. However, studies4, 7, 8, 9 have reported considerable morbidity related to the temporary ileostomy, with complication rates of up to 50 per cent. Most patients with a temporary ileostomy have the stoma for at least 3 months, and it is not unusual for it to be left in place much longer. For some patients the stoma becomes permanent10. Data regarding quality of life (QOL) in patients receiving a diverting stoma as part of their rectal cancer treatment are limited11, 12, 13, 14. In prospective studies it has been suggested that patients with a stoma may suffer from impaired health‐related quality of life (HRQOL)15, which may improve at stoma closure2. Complications such as stoma leakage, parastomal skin irritation, dietary restrictions, retraction and prolapse of the stoma have been reported to have significant impact on the patient's daily life12. The aim of the present study was to compare HRQOL at 3, 6 and 12 months after rectal resection for cancer in a multicentre RCT comparing early versus late closure of a temporary ileostomy (EASY trial)16, 17.

Data regarding quality of life (QOL) in patients receiving a diverting stoma as part of their rectal cancer treatment are limited11, 12, 13, 14. In prospective studies it has been suggested that patients with a stoma may suffer from impaired health‐related quality of life (HRQOL)15, which may improve at stoma closure2. Complications such as stoma leakage, parastomal skin irritation, dietary restrictions, retraction and prolapse of the stoma have been reported to have significant impact on the patient's daily life12. The aim of the present study was to compare HRQOL at 3, 6 and 12 months after rectal resection for cancer in a multicentre RCT comparing early versus late closure of a temporary ileostomy (EASY trial)16, 17. Methods The EASY trial was designed as a randomized multicentre trial16 comparing early with late closure of a temporary ileostomy regarding risk of complications. Screening for and inclusion of participants was made after index surgery (total mesorectal excision for rectal cancer including creation of a temporary ileostomy). Exclusion criteria were diabetes mellitus, ongoing steroid treatment, signs of postoperative complications at clinical evaluation 1–4 days after rectal resection and inability to understand the Danish or Swedish language. Patients with no adverse signs were invited to participate and, after informed consent, underwent further investigation with contrast‐enhanced CT or a flexible endoscopy of the rectum, or both, performed 6–8 days after stoma creation to ensure that no patient with signs of anastomotic leakage was included. Patients were randomized to either the intervention group with early closure (day 8–13 after stoma creation) or the control group with late closure (more than 12 weeks after stoma creation) of the ileostomy.

erformed 6–8 days after stoma creation to ensure that no patient with signs of anastomotic leakage was included. Patients were randomized to either the intervention group with early closure (day 8–13 after stoma creation) or the control group with late closure (more than 12 weeks after stoma creation) of the ileostomy. The primary endpoint of the study was the mean number of complications after rectal resection and up to 12 months; these results have been published previously17. The present paper reports the secondary endpoints, HRQOL and QOL, at 3, 6 and 12 months after the index operation. The study was approved in Denmark by the Science Ethics Committee for the Capital Region (H‐1‐2010‐113) and in Sweden by the Regional Ethics Approval Committee in Göteborg (Dnr 064‐2011). Before inclusion, patients were informed about the study and all participating patients returned a signed consent form. The project was approved by the Data Protection Agency in Denmark, and by the Personal Data Representative at Sahlgrenska University Hospital, Göteborg, Sweden. The protocol was registered at https://www.clinicaltrials.gov (NCT01287637) before patient inclusion.

The study was approved in Denmark by the Science Ethics Committee for the Capital Region (H‐1‐2010‐113) and in Sweden by the Regional Ethics Approval Committee in Göteborg (Dnr 064‐2011). Before inclusion, patients were informed about the study and all participating patients returned a signed consent form. The project was approved by the Data Protection Agency in Denmark, and by the Personal Data Representative at Sahlgrenska University Hospital, Göteborg, Sweden. The protocol was registered at https://www.clinicaltrials.gov (NCT01287637) before patient inclusion. Patients Eight hospitals in Denmark and Sweden participated in the study, but three centres (with a total of 8 patients) were excluded as they failed to maintain a screening log. Consenting patients were asked to complete questionnaires at 3, 6 and 12 months after stoma creation (rectal resection). The questionnaires included the 36‐item Short Form 36 (SF‐36®; Rand Corporation, Santa Monica, California, USA) and the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ) CR29 and C30. Data regarding demographic details, tumour stage and height, chemoradiotherapy and all complications within 12 months of surgery were registered in case report forms.

Santa Monica, California, USA) and the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ) CR29 and C30. Data regarding demographic details, tumour stage and height, chemoradiotherapy and all complications within 12 months of surgery were registered in case report forms. Randomization Consenting patients who fulfilled the inclusion criteria were randomized either to the intervention group with early closure of the ileostomy or to the control group with late closure. Randomization was executed in computer‐generated blocks of six. The randomization was performed on the surgical ward using sequentially numbered thick, opaque and sealed envelopes. Blinding of the intervention was not possible. Health‐related quality‐of‐life instruments Short Form 36 SF‐36® is a generic tool that evaluates patients' self‐reported quality of life18, 19. It consists of 36 items that measure eight dimensions of health on a multi‐item scale, including social and physical function. The scoring scale ranges from 0 to 100, with lower scores indicating worse health. The instrument has been validated, and for comparison in this study a Swedish reference population was used20. EORTC QLQ‐C30 and QLQ‐CR29

SF‐36® is a generic tool that evaluates patients' self‐reported quality of life18, 19. It consists of 36 items that measure eight dimensions of health on a multi‐item scale, including social and physical function. The scoring scale ranges from 0 to 100, with lower scores indicating worse health. The instrument has been validated, and for comparison in this study a Swedish reference population was used20. EORTC QLQ‐C30 and QLQ‐CR29 The EORTC QLQ‐C30 is a questionnaire developed to assess QOL in patients with cancer, and consists of five functional scales (physical, role, cognitive, emotional and social), three symptom scales (fatigue, pain and nausea, and vomiting), one global health status and QOL scale, and six single‐item measures (dyspnoea, insomnia, appetite loss, constipation, diarrhoea and financial difficulties)21. A high score on the functional scale represents a high level of functioning, whereas a high score on the symptom scale represents a high level of symptoms. The EORTC QLQ‐CR29 was designed for use in patients undergoing treatment for colorectal cancer. It was derived from the EORTC QLQ‐C38 questionnaire, as there was a need for an update in the colorectal module22. The questions assess disease symptoms, side‐effects of treatment, body image, future perspective, and sexual function and interest. Both questionnaires have been validated internationally and were available in Danish and Swedish versions22, 23. All questionnaires were administered at 3, 6 and 12 months after stoma creation.

The EORTC QLQ‐CR29 was designed for use in patients undergoing treatment for colorectal cancer. It was derived from the EORTC QLQ‐C38 questionnaire, as there was a need for an update in the colorectal module22. The questions assess disease symptoms, side‐effects of treatment, body image, future perspective, and sexual function and interest. Both questionnaires have been validated internationally and were available in Danish and Swedish versions22, 23. All questionnaires were administered at 3, 6 and 12 months after stoma creation. Statistical analysis The study was part of a RCT with power calculated for the primary endpoint. Group sizes in the EASY trial were set to 72 patients per group to evaluate complication rates16, 17. In Krouse et al.24 a minimally important difference of 8 units was used for the different scales of SF‐36®. Physical and mental component scores had a standard deviation (s.d.) of up to 15 units, and the eight specific scales had s.d. values in the range 17–33. With group sizes of 72 or 55 and a s.d. of 17 or 15 respectively, there will be 80 per cent power to detect a true difference of 8 units.

different scales of SF‐36®. Physical and mental component scores had a standard deviation (s.d.) of up to 15 units, and the eight specific scales had s.d. values in the range 17–33. With group sizes of 72 or 55 and a s.d. of 17 or 15 respectively, there will be 80 per cent power to detect a true difference of 8 units. The questionnaires SF‐36®25, EORTC QLQ‐C3026 and QLQ‐CR2923 were scored according to the methods recommended by the developers; missing data were handled as instructed in the scoring manuals. Before analysis, based on the literature and previous data, the authors chose to present the functional scales and global health status/QOL of the EORTC QLQ‐C30 questionnaire, and the functional scales (urinary frequency, stool frequency and body image) of the EORTC QLQ‐CR29 questionnaire. At 12‐month follow‐up, several patients (20 in the intervention group and 16 in the control group) were, by mistake, given an incomplete questionnaire in which questions 16–30 of the EORTC QLQ‐C30 were missing. Before analysis, the decision was made to include physical and role functioning, as these functional scales are scored using questions 1–15 in accordance with the EORTC manual. No other scales or items were analysed for these patients in the EORTC QLQ‐C30 at 12‐month follow‐up. However, EORTC QLQ‐CR29 and SF‐36® were analysed at 12‐month follow‐up. As the distribution of the incomplete questionnaires was independent of the observable characteristics of the patients who received them, interpreting the missing data as being completely random is reasonable and imputation was considered unnecessary.

ever, EORTC QLQ‐CR29 and SF‐36® were analysed at 12‐month follow‐up. As the distribution of the incomplete questionnaires was independent of the observable characteristics of the patients who received them, interpreting the missing data as being completely random is reasonable and imputation was considered unnecessary. Owing to the characteristics of the data, the different scales of SF‐36® and EORTC were summarized by median (i.q.r.) values, and group comparisons were made using the Wilcoxon rank sum test and the two‐sample Hodges–Lehmann estimator. SF‐36® scores were compared with those in a general Swedish reference population20. For each of the eight scales, the individual levels were compared with age‐matched (17–34, 35–49, 50–64, and 65 or more years) mean levels from the reference population. The proportion of patients with values below the reference levels was compared between the intervention and control group using a χ2 test. The software packages SPSS® version 23 (IBM, Armonk, New York, USA), SAS® version 9.4 (SAS Institute, Cary, North Carolina, USA) and R version 3.2.327 were used for statistical analysis. Scores for SF‐36®, EORTC QLQ‐C30 and QLQ‐CR29 were derived and summarized by median (i.q.r.) values.

d control group using a χ2 test. The software packages SPSS® version 23 (IBM, Armonk, New York, USA), SAS® version 9.4 (SAS Institute, Cary, North Carolina, USA) and R version 3.2.327 were used for statistical analysis. Scores for SF‐36®, EORTC QLQ‐C30 and QLQ‐CR29 were derived and summarized by median (i.q.r.) values. Results The EASY trial assessed 418 patients for eligibility. After exclusion of 291 patients, 127 patients were randomized (Fig. 1). A further 15 patients were excluded, eight from three centres that were excluded from the study as they failed to maintain a screening log. In summary, 112 patients were included from February 2011, with the last follow‐up in November 2015. Some 55 patients in the intervention (early closure) group and 57 in the control (late closure) group were available for analysis. There were no violations of the randomization. Except for a larger proportion of women in the intervention group, baseline demographic characteristics and clinical data were similar in the two groups (Table 1).

he intervention (early closure) group and 57 in the control (late closure) group were available for analysis. There were no violations of the randomization. Except for a larger proportion of women in the intervention group, baseline demographic characteristics and clinical data were similar in the two groups (Table 1). Figure 1 Participant flow diagram, as in the EASY trial17. *Paralytic ileus (24), Hartmann procedure with intersphincteric dissection (16), delayed postoperative recovery (15), perioperative complications (7), other infection (5), reoperation (7), high stoma output (5), pulmonary embolism (1), ulcerative colitis (1), extensive cancer disease (3), cardiovascular disease (2), language difficulty (5), diabetes (28), permanent or no stoma (29), steroid treatment (3), other (8). †Centre 6 (2), centre 7 (3), centre 8 (3). ‡Allocated to early closure, but not possible to perform surgery within 8–13 days (1); early closure outside the study (2); randomized, but no further information available (1). §3 months: Short Form 36 (SF‐36®) (52), European Organisation for Research and Treatment of Cancer (EORTC) QLQ‐C30 (52), EORTC QLQ‐CR29 (52); 6 months: 3 months: SF‐36® (52), EORTC QLQ‐C30 (50), EORTC QLQ‐CR29 (50); 12 months: 3 months: SF‐36® (50), EORTC QLQ‐C30 (30), EORTC QLQ‐CR29 (50). ¶3 months: SF‐36® (53), EORTC QLQ‐C30 (51), EORTC QLQ‐CR29 (52); 6 months: 3 months: SF‐36® (53), EORTC QLQ‐C30 (53), EORTC QLQ‐CR29 (53); 12 months: SF‐36® (47), EORTC QLQ‐C30 (31), EORTC QLQ‐CR29 (47). #Missing from follow‐up at 12 months. **Patient did not have closure; missing from follow‐up at 12 months

CR29 (50). ¶3 months: SF‐36® (53), EORTC QLQ‐C30 (51), EORTC QLQ‐CR29 (52); 6 months: 3 months: SF‐36® (53), EORTC QLQ‐C30 (53), EORTC QLQ‐CR29 (53); 12 months: SF‐36® (47), EORTC QLQ‐C30 (31), EORTC QLQ‐CR29 (47). #Missing from follow‐up at 12 months. **Patient did not have closure; missing from follow‐up at 12 months BJS-10680-FIG-0001-cTable 1 Baseline and preoperative characteristics of patients randomized to early or late closure Early closure (n = 55) Late closure (n = 57) Age (years)* 67 (36–82) 67 (39–81) Sex ratio (F : M) 31 : 24 21 : 36 BMI (kg/m2)* 24 (17–32) 23 (19–35) Co‐morbidity 23 (42) 24 (42)† Ischaemic heart disease 5 8 Hypertension 17 13 COPD 2 2 Renal disease 0 0 Other 9‡ 4§ Radiotherapy 16 (29) 16 (28) Long‐term 5 (9) 5 (9) Adjuvant chemotherapy 22 (40) 23 (40) Marital status Single 5 (9) 9 (16) Married 42 (76) 43 (75) Widowed 8 (15) 5 (9) Higher education 34 (62) 37 (65) Employed Yes 25 (45) 25 (44) No 29 (53)¶ 32 (56) Smoker 6 (11) 4 (7) No. of pack years* 30 (16–30)# 26 (20–50)¶ Alcohol intake > 60 g/day 0 (0) 0 (0)¶ Lower border of tumour (cm from anal verge) 5–9 27 (49) 24 (42) 10–15 27 (49) 33 (58) > 15 1 (2) 0 (0) UICC clinical stage** I 12 (22) 19 (33) II 21 (38) 13 (23) III 18 (33) 20 (35) IV 3 (5) 1 (2) Method of evaluation of anastomosis before ileostomy closure CT 14 (25) 19 (33) Rectoscopy 14 (25) 10 (18) CT + rectoscopy 27 (49) 28 (49) Total length of hospital stay†† (days) * 14 (11–42) 14 (7–44) Values in parentheses are percentages unless indicated otherwise; * values are median (range). † Data missing for two patients.

I 12 (22) 19 (33) II 21 (38) 13 (23) III 18 (33) 20 (35) IV 3 (5) 1 (2) Method of evaluation of anastomosis before ileostomy closure CT 14 (25) 19 (33) Rectoscopy 14 (25) 10 (18) CT + rectoscopy 27 (49) 28 (49) Total length of hospital stay†† (days) * 14 (11–42) 14 (7–44) Values in parentheses are percentages unless indicated otherwise; * values are median (range). † Data missing for two patients. ‡ Asthma (2), depression (1), idiopathic thrombocytopenic purpura (1), lymphoma (1), Waldenström's macroglobulinaemia (1), osteoporosis (1), Sjögren syndrome (1), thyrotoxicosis (1). § Depression (1), hyperlipidaemia (1), hypothyroidism (1), meningioblastoma (1). ¶ Data missing for one patient. # Data missing for three patients. ** Data missing for one patient in each group; in addition, three patients in the late closure group had T0 N0 M0 disease and were therefore not classified. †† For both index surgery and loop ileostomy closure. COPD, chronic obstructive pulmonary disease; UICC, International Union Against Cancer. Table reproduced with permission from Wolters Kluwer Health Inc. Danielsen AK, Park J, Jansen JE, Bock D, Skullman S, Wedin A et al. Early closure of a temporary ileostomy in patients with rectal cancer: a multicenter randomized controlled trial. Ann Surg 2017; 265 (2): 284–290. https://journals.lww.com/annalsofsurgery/. Questionnaire response rates were generally around 90 (range 82–95) per cent, excluding the EORTC QLQ‐C30 at 12‐month follow‐up, owing to missing questions as described in the Methods section (Fig. 1).

†† For both index surgery and loop ileostomy closure. COPD, chronic obstructive pulmonary disease; UICC, International Union Against Cancer. Table reproduced with permission from Wolters Kluwer Health Inc. Danielsen AK, Park J, Jansen JE, Bock D, Skullman S, Wedin A et al. Early closure of a temporary ileostomy in patients with rectal cancer: a multicenter randomized controlled trial. Ann Surg 2017; 265 (2): 284–290. https://journals.lww.com/annalsofsurgery/. Questionnaire response rates were generally around 90 (range 82–95) per cent, excluding the EORTC QLQ‐C30 at 12‐month follow‐up, owing to missing questions as described in the Methods section (Fig. 1). SF‐36® scores were similar between the two groups, with no differences in the physical component score or the mental component score, but significant differences in role physical, bodily pain and mental health at 3, 12 and 12 months respectively (Table 2). All dimensions in SF‐36® improved over time. At 3 months, a majority of patients in both groups scored values below mean levels in the reference population20, especially regarding role physical. At 12 months, 52–85 per cent of the patients scored higher than the reference group, with physical functioning scoring the highest among the dimensions. Table 2 SF‐36® scores at 3, 6 and 12 months after rectal resection 3 months 6 months 12 months Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P †

SF‐36® scores were similar between the two groups, with no differences in the physical component score or the mental component score, but significant differences in role physical, bodily pain and mental health at 3, 12 and 12 months respectively (Table 2). All dimensions in SF‐36® improved over time. At 3 months, a majority of patients in both groups scored values below mean levels in the reference population20, especially regarding role physical. At 12 months, 52–85 per cent of the patients scored higher than the reference group, with physical functioning scoring the highest among the dimensions. Table 2 SF‐36® scores at 3, 6 and 12 months after rectal resection 3 months 6 months 12 months Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Physical functioning Early 90 (75–95) 0 (−5, 5) 0·646 90 (81·7–100) 0 (−5, 5) 0·630 95 (70–100) 0 (0, 5) 0·322 Late 90 (80–95) 90 (80–95) 95 (90–100) Role physical Early 75 (50–96·9) 12·5 (0, 18·8) 0·025 81·3 (50–100) 6·3 (0, 18·8) 0·140 81·3 (56·3–100) 0 (−6·3, 6·3) 0·718 Late 62·5 (43·8–75) 75 (50–93·8) 87·5 (75–100) Bodily pain Early 80 (52–100) 0 (−10, 0) 0·858 74 (62–100) 0 (−16, 0) 0·264 79 (51–100) 0 (0, 20) 0·035 Late 74 (62–100) 84 (63–100) 100 (74–100) General health Early 71·6 (52–88·5) −5 (−15, 2) 0·139 77 (56–87) 0 (−10, 5) 0·820 74·5 (45–92) 5 (−5, 16·8) 0·279 Late 77 (67–87) 77 (65–87) 82 (72–87) Vitality Early 62·5 (43·8–81·3) 4·2 (−6·3, 12·5) 0·441 68·8 (50–81·3) 0 (−12·5, 6·3) 0·796 68·8 (50–81·3) 6·3 (0, 12·5) 0·196 Late 68·8 (56·3–81·3) 68·8 (56·3–81·3) 75 (62·5–87·5) Social functioning Early 75 (62·5–100) 0 (−12·5, 0) 0·468 87·5 (62·5–100) 0 (0, 0) 0·976 87·5 (62·5–100) 0 (0, 12·5) 0·415 Late 87·5 (75–100) 87·5 (62·5–100) 100 (75–100) Role emotional Early 83·3 (58·3–100) 0 (0, 8·3) 0·345 87·5 (66·7–100) 0 (0, 0) 0·923 95·8 (66·7–100) 0 (0, 0) 0·697 Late 83·3 (50–100) 83·3 (75–100) 95·8 (75–100) Mental health Early 80 (55–90) 5 (−5, 10) 0·217 80 (60–90) −5 (−10, 5) 0·291 80 (60–90) 10 (0, 15) 0·020 Late 85 (65–90) 85 (70–95) 85 (75–95) Mental component score Early 52·5 (40·7–58·6) 1 (−2·6, 5) 0·588 54·4 (42·8–58·6) 0·2 (−3, 3·9) 0·939 54·1 (42·6–58·5) 2·5 (−0·7, 6·3) 0·105 Late 53 (44·8–57·8) 54·6 (46·9–57·5) 56·6 (52·9–59·2) Physical component score Early 51·8 (40·9–58·2) −0·5(−3·8, 3·4) 0·823 53·3 (43·3–57·1) −0·2 (−3·6, 3) 0·900 54·1 (44·5–59) 1·6 (−1, 6·1) 0·281 Late 51·2 (46·9–54·8) 52·2 (45·8–57·9) 56·8 (51–59·4) * Two‐sample Hodges–Lehmann (H–L) estimator with 95 per cent asymptotic confidence limits in parentheses. SF‐36®, Short Form 36.

cal component score Early 51·8 (40·9–58·2) −0·5(−3·8, 3·4) 0·823 53·3 (43·3–57·1) −0·2 (−3·6, 3) 0·900 54·1 (44·5–59) 1·6 (−1, 6·1) 0·281 Late 51·2 (46·9–54·8) 52·2 (45·8–57·9) 56·8 (51–59·4) * Two‐sample Hodges–Lehmann (H–L) estimator with 95 per cent asymptotic confidence limits in parentheses. SF‐36®, Short Form 36. † Wilcoxon rank sum test for difference between early and late closure. EORTC QLQ‐C30 and QLQ‐CR29 scores were comparable between intervention and control groups. Emotional functioning was lower in the early closure group at 3 and 6 months, but similar to the late closure group at 12 months (Table 3). No statistically significant differences were seen in the dimensions of the QLQ‐CR29 questionnaire (Table 4). Table 3 EORTC QLQ‐C30 scores at 3, 6 and 12 months after rectal resection 3 months 6 months 12 months Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P †

EORTC QLQ‐C30 and QLQ‐CR29 scores were comparable between intervention and control groups. Emotional functioning was lower in the early closure group at 3 and 6 months, but similar to the late closure group at 12 months (Table 3). No statistically significant differences were seen in the dimensions of the QLQ‐CR29 questionnaire (Table 4). Table 3 EORTC QLQ‐C30 scores at 3, 6 and 12 months after rectal resection 3 months 6 months 12 months Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Global quality of life Early 75 (50–83·3) 0 (−8·3, 8·3) 0·941 66·7 (50–83·3) 0 (−8·3, 8·3) 0·961 83·3 (50–91·7) 0 (−16·7, 8·3) 0·889 Late 66·7 (58·3–83·3) 66·7 (66·7–83·3) 83·3 (66·7–91·7) Physical functioning Early 93·3 (73·3–100) 0 (0, 6·7) 0·634 93·3 (80–100) 0 (−6·7, 0) 0·433 93·3 (73·3–100) 0 (−13·3, 0) 0·137 Late 93·3 (73·3–100) 93·3 (80–100) 100 (80–100) Role functioning Early 83·3 (66·7–100) −16·7 (−16·7, 0) 0·066 100 (66·7–100) 0 (0, 0) 0·503 100 (66·7–100) 0 (0, 0) 0·793 Late 66·7 (50–100) 83·3 (66·7–100) 100 (66·7–100) Emotional functioning Early 83·3 (66·7–100) 8·3 (0, 16·7) 0·023 83·3 (66·7–100) −8·3 (−16·7, 0) 0·031 91·7 (66·7–100) 0 (−8·3, 0) 0·409 Late 91·7 (83·3–100) 91·7 (75–100) 91·7 (83·3–100) Cognitive functioning Early 100 (83·3–100) 0 (0, 0) 0·447 83·3 (83·3–100) 0 (−16·7, 0) 0·131 100 (66·7–100) 0 (0, 0) 0·652 Late 100 (83·3–100) 100 (83·3–100) 100 (83·3–100) Social functioning Early 83·3 (66·7–100) 0 (0, 0) 0·583 83·3 (66·7–100) 0 (0, 0) 0·882 83·3 (66·7–100) 0 (−16·7, 0) 0·142 Late 83·3 (66·7–100) 83·3 (66·7–100) 100 (66·7–100) * Two‐sample Hodges–Lehmann (H–L) estimator with 95 per cent asymptotic confidence limits in parentheses. EORTC, European Organisation for Research and Treatment of Cancer.

3·3 (66·7–100) 0 (0, 0) 0·583 83·3 (66·7–100) 0 (0, 0) 0·882 83·3 (66·7–100) 0 (−16·7, 0) 0·142 Late 83·3 (66·7–100) 83·3 (66·7–100) 100 (66·7–100) * Two‐sample Hodges–Lehmann (H–L) estimator with 95 per cent asymptotic confidence limits in parentheses. EORTC, European Organisation for Research and Treatment of Cancer. † Wilcoxon rank sum test for difference between early and late. Table 4 EORTC QLQ‐CR29 scores for selected functions at 3, 6 and 12 months after rectal resection 3 months 6 months 12 months Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Median (i.q.r.) H–L* P † Urinary frequency Early 16·7 (0–33·3) 0 (0, 16·7) 0·323 16·7 (0–50) 0 (−16·7, 0) 0·353 16·7 (0–33·3) 0 (0, 16·7) 0·268 Late 16·7 (0–50) 16·7 (8·3–41·7) 33·3 (0–50) Stool frequency Early 33·3 (16·7–50) 33·3 (16·7, 33·3) < 0·001   33·3 (16·7–50) 16·7 (0, 33·3) 0·068 33·3 (16·7–50) 0 (0, 16·7) 0·611 Late 0 (0–16·7) 16·7 (0–66·7) 33·3 (16·7–50) Body image Early 88·9 (66·7–100) 0 (−11·1, 0) 0·715 88·9 (77·8–100) 0 (0, 11·1) 0·364 94·4 (77·8–100) 0 (0, 0) 0·502 Late 77·8 (66·7–100) 88·9 (66·7–100) 100 (88·9–100) * Two‐sample Hodges–Lehmann (H–L) estimator with 95 per cent asymptotic confidence limits in parentheses. EORTC, European Organisation for Research and Treatment of Cancer. † Wilcoxon rank sum test for difference between early and late closure.

Urinary frequency Early 16·7 (0–33·3) 0 (0, 16·7) 0·323 16·7 (0–50) 0 (−16·7, 0) 0·353 16·7 (0–33·3) 0 (0, 16·7) 0·268 Late 16·7 (0–50) 16·7 (8·3–41·7) 33·3 (0–50) Stool frequency Early 33·3 (16·7–50) 33·3 (16·7, 33·3) < 0·001   33·3 (16·7–50) 16·7 (0, 33·3) 0·068 33·3 (16·7–50) 0 (0, 16·7) 0·611 Late 0 (0–16·7) 16·7 (0–66·7) 33·3 (16·7–50) Body image Early 88·9 (66·7–100) 0 (−11·1, 0) 0·715 88·9 (77·8–100) 0 (0, 11·1) 0·364 94·4 (77·8–100) 0 (0, 0) 0·502 Late 77·8 (66·7–100) 88·9 (66·7–100) 100 (88·9–100) * Two‐sample Hodges–Lehmann (H–L) estimator with 95 per cent asymptotic confidence limits in parentheses. EORTC, European Organisation for Research and Treatment of Cancer. † Wilcoxon rank sum test for difference between early and late closure. Discussion In this RCT no significant differences were observed in HRQOL within 12 months after rectal resection for cancer when early and late closure of temporary ileostomy were compared. Global QOL generally improved later in the follow‐up period (6–12 months), and at 12 months the results were comparable, not only between the two groups but also with respect to both age‐matched reference populations28, 29 and previous findings30. Although there was a tendency for improvement over time in global QOL and role functioning, no clinically significant changes were seen in QLQ‐C30 scores. The definition of a clinically significant change over time was based on the suggestion31 that a difference of 5–10 points be considered a ‘ little change’ and a difference of 10–20 points a ‘ moderate change’. When comparing the SF‐36® scores with Swedish reference data20, a general improvement was seen during the follow‐up interval.

f a clinically significant change over time was based on the suggestion31 that a difference of 5–10 points be considered a ‘ little change’ and a difference of 10–20 points a ‘ moderate change’. When comparing the SF‐36® scores with Swedish reference data20, a general improvement was seen during the follow‐up interval. No baseline data for preoperative assessment were available, owing to the fact that the patients were enrolled and randomized after the rectal resection. This is considered a weakness in the evaluation of HRQOL. Even though reference data were available, there was no opportunity to investigate the development and changes in HRQOL from preoperative to postoperative values. In addition, considering the length of time needed for recovery after rectal cancer surgery, follow‐up at 18 and 24 months might have been of value to observe any further development and potential improvement in HRQOL.

tunity to investigate the development and changes in HRQOL from preoperative to postoperative values. In addition, considering the length of time needed for recovery after rectal cancer surgery, follow‐up at 18 and 24 months might have been of value to observe any further development and potential improvement in HRQOL. The lack of difference between the two groups differs somewhat from the findings of previous prospective studies2, 13, which have reported that patients with a temporary stoma may suffer from impaired HRQOL in comparison with patients who had undergone a similar operation but without a covering stoma, such as a high anterior resection. However, the results of previous studies are contradictory. A Danish study32 suggested that a temporary stoma left patients with feelings of uncertainty, that closure of the stoma was seen as a crucial event, and that knowing the date for closure was important. A prospective study11 found improved global QOL following stoma reversal after low anterior resection for rectal cancer, and physical function, as measured by both the QLQ‐C30 and SF‐36® questionnaires, was also significantly improved. However, a prospective interview study14 reported no change in QOL following closure of a temporary stoma after rectal cancer surgery, and the patient's personality rather than clinical variables has a strong and lasting influence on QOL33. As suggested in a review article34 from 2011, closure of the stoma and prompt redirection of intestinal contents to the rectum 3–6 months after rectal resection might be considered to be a cause of negative impact on patients' physical, social and psychological health for several months. Gastrointestinal symptoms, such as increased stool frequency, urgency, diarrhoea and persistent problems including low anterior resection syndrome, do not occur in the presence of a temporary stoma. Consequently, stoma reversal might result in the appearance of these symptoms. Although this was not observed in the present study, this could be because of the small size of the cohort. In the COLOR II trial35, patients with rectal cancer appeared to need a longer time to recover compared with those with colonic cancer. This could perhaps explain the lack of difference between the groups in the present study, as both involved extensive surgery for rectal cancer, where timing of the stoma closure is not reflected in the HRQOL follow‐up.

ith rectal cancer appeared to need a longer time to recover compared with those with colonic cancer. This could perhaps explain the lack of difference between the groups in the present study, as both involved extensive surgery for rectal cancer, where timing of the stoma closure is not reflected in the HRQOL follow‐up. The EASY trial17 found that it was safe and advantageous to close the loop ileostomy early in patients with no clinical or radiological signs of anastomotic leakage. However, the present study did not find a link between this clinical advantage and patients' HRQOL. Acknowledgements The authors express their gratitude to the doctors, nurses and staff at the participating centres for their help and assistance during the course of this trial. A.K.D. has received funding from the Research Council at Herlev and Gentofte Hospital, Denmark. S.S. has received funding from the Health and Medical Care Committee of the Regional Executive Board and Region Västra Götaland (numbers 311031 and 390991) and from Lions Cancerfond West, Sweden. E.A. has received funding from the Swedish Research Council (2012‐1786), Swedish Cancer Society (2013/500), Sahlgrenska University Hospital (agreement concerning research and education of doctors ALFGBG‐366481, ALFGBG‐526501 and ALFGBG‐493341), Swedish Society of Medicine (SLS‐247661 and SLS‐412151) and Lions Väst Cancer Foundation. Disclosure: The authors declare no conflict of interest.

Introduction Abdominal aortic aneurysm (AAA) is defined as an abnormal dilatation of the abdominal aorta of 30 mm or more, and constitutes a significant health problem worldwide1. Each year in England and Wales, AAAs cause over 4000 deaths following aortic rupture2, with approximately 8000 patients a year undergoing surgery to prevent this3. In 2013, the National Health Service (NHS) AAA screening programme (NAAASP) was fully rolled out across England, based on evidence from several RCTs suggesting that AAA‐related mortality was reduced through participation in AAA screening4. The NAAASP currently invites all men in their 65th year to receive a one‐off non‐invasive abdominal ultrasound scan. In England in 2015–2016, 227 543 men were screened and 2549 (1·1 per cent) were diagnosed with an AAA; however, only 723 men (0·3 per cent) had an AAA large enough (at least 55 mm) to require referral for consideration of surgery5. This highlights one of the major issues with screening for AAA in that, although it remains cost‐effective, the majority of patients identified do not require immediate surgery and are subsequently entered into ongoing surveillance, either 6‐monthly or annually. Most men with a screen‐detected AAA will spend 3–5 years in surveillance before reaching the threshold for elective AAA repair, rising to over 7 years for men with a 30‐mm AAA6. Currently 13 104 men in England are in AAA surveillance5.

nd are subsequently entered into ongoing surveillance, either 6‐monthly or annually. Most men with a screen‐detected AAA will spend 3–5 years in surveillance before reaching the threshold for elective AAA repair, rising to over 7 years for men with a 30‐mm AAA6. Currently 13 104 men in England are in AAA surveillance5. This has led to questions being raised over the psychological impact of AAA screening. Some have even suggested that AAA screening may do more harm than good7. A small number of observational studies have investigated quality of life (QoL) in those who are identified at screening to have an AAA8, 9, 10, 11, 12, 13, demonstrating varying results and conclusions when comparing screened and unscreened cohorts. The United Kingdom Aneurysm Growth Study (UKAGS) is a prospective observational cohort study currently recruiting men with AAA identified through the English NHS AAA screening programme (NAAASP) and the Welsh AAA screening programme (WAAASP), with the aim of investigating the growth rates of small AAAs. All recruited men (those with an AAA and controls without) are sent an annual self‐completed postal questionnaire to obtain longitudinal clinical and QoL data. This resource was used to assess the contemporary impact of screening for AAA on men who attend the NAAASP and WAAASP.

stigating the growth rates of small AAAs. All recruited men (those with an AAA and controls without) are sent an annual self‐completed postal questionnaire to obtain longitudinal clinical and QoL data. This resource was used to assess the contemporary impact of screening for AAA on men who attend the NAAASP and WAAASP. Methods NAAASP and WAAASP invite all men in England and Wales during their 65th year of age to attend AAA screening14. Eligible men are sent an invitation letter to attend a local clinic for an ultrasound scan. A technician measures the maximal anteroposterior inner wall to inner wall diameter of the infrarenal aorta. Those with a diameter of less than 3 cm are discharged; those with a diameter between 3·0 and 5·4 cm are offered ultrasound surveillance every 6 or 12 months (based on AAA diameter); and men with an aortic diameter above 5·4 cm are directly referred for possible surgical repair.

r wall diameter of the infrarenal aorta. Those with a diameter of less than 3 cm are discharged; those with a diameter between 3·0 and 5·4 cm are offered ultrasound surveillance every 6 or 12 months (based on AAA diameter); and men with an aortic diameter above 5·4 cm are directly referred for possible surgical repair. UKAGS is a prospective observational cohort study that recruits men with an AAA, as well as individuals without AAA (controls) who have attended AAA screening, from NAAASP or WAAASP15. All recruited men are sent annual self‐completion postal questionnaires (Appendix S1, supporting information) to obtain longitudinal clinical information and QoL information. Additionally, those with AAA on initial screening undergo annual ultrasound screening measurements through the standard surveillance procedures14. UKAGS is currently recruiting individuals from 14 units across England and Wales, and aims to recruit 20 000 men over 5 years. Ethical approval has been granted by an NHS research ethics committee, and men have provided their written informed consent upon recruitment for QoL data collection and analyses.

procedures14. UKAGS is currently recruiting individuals from 14 units across England and Wales, and aims to recruit 20 000 men over 5 years. Ethical approval has been granted by an NHS research ethics committee, and men have provided their written informed consent upon recruitment for QoL data collection and analyses. Data collection and quality‐of‐life assessments Data collected at baseline included AAA diameter (inner wall to inner wall measurement), demographics, standard cardiovascular co‐morbidities and QoL‐related fields. This includes eight questions adapted from the Medical Outcomes Study Short Form 36 questionnaire (SF‐8), a survey that has previously been recommended specifically for vascular disease‐related QoL outcome analyses16 and has demonstrated high reliability and validity17. Several other QoL questionnaires are available, and have been used by other groups. The present questions were considered most suitable for the UKAGS.

‐8), a survey that has previously been recommended specifically for vascular disease‐related QoL outcome analyses16 and has demonstrated high reliability and validity17. Several other QoL questionnaires are available, and have been used by other groups. The present questions were considered most suitable for the UKAGS. The SF‐8 questionnaire uses single‐item scales addressing eight domains of general health, physical functioning, role limitations (due to physical health), bodily pain, vitality, social functioning, mental health and role limitations (due to emotional health). These parameters are then used to produce two outcome measures of QoL: the Physical Component Summary (PCS) and the Mental Component Summary (MCS). The scores range from 0 to 100, where zero indicates the lowest and 100 the highest level of health, calibrated so that the average score is 50, with a standard deviation of 1017. Scores were calculated using the QualityMetric Health Outcomes™ Scoring Software 4.5 (Optum®; QualityMetric, Lincoln, Rhode Island, USA). Additionally, using a Likert scale (1, not at all; 5, all the time), men with a known AAA were asked how often they thought about their aneurysm and how often they had thought about the potential for aneurysm growth in the preceding 4 weeks.

mes™ Scoring Software 4.5 (Optum®; QualityMetric, Lincoln, Rhode Island, USA). Additionally, using a Likert scale (1, not at all; 5, all the time), men with a known AAA were asked how often they thought about their aneurysm and how often they had thought about the potential for aneurysm growth in the preceding 4 weeks. As recruitment occurred through the traditional screening pathway, participants were recruited into the study across an extended period (September 2011 to July 2015). Relative to each man's baseline recruitment date, QoL data collection follow‐up was divided into four groups after initial screening: 0–12, 13–24, 25–36 and 37 or more months. The primary outcome measure was comparing PCS and MCS at each time point between men with a diagnosed AAA on initial screening and those without an AAA (control group). Data collection was done at yearly follow‐up intervals with recruitment from NAAASP and WAAASP into the study occurring continuously throughout the 4‐year study. Thus, participants were in the study for varying lengths of time and had completed a varying number of questionnaires (Table 1). Table 1 Total number of questionnaire respondents for each time interval after initial screening

Data collection was done at yearly follow‐up intervals with recruitment from NAAASP and WAAASP into the study occurring continuously throughout the 4‐year study. Thus, participants were in the study for varying lengths of time and had completed a varying number of questionnaires (Table 1). Table 1 Total number of questionnaire respondents for each time interval after initial screening Time from initial screening (months) No aneurysm Aneurysm Total 0–12 4807 174 4981 13–24 4232 238 4470 25–36 914 142 1056 ≥ 37 52 93 145 Statistical analysis The SF‐8 QoL data collected were analysed using ANOVA, comparing the AAA group with the no‐AAA group at each interval. Regression analyses accounted for potential confounding factors, including demographics and co‐morbidities. Mean growth rate (cm/month) during surveillance was calculated and linear regression was used to identify any association of growth rate with the QoL parameters. ANOVA was used to assess changes in the frequencies over time with which respondents thought about their aneurysm and aneurysm growth; data for the intervals of 13–24, 25–36 and 37 or more months were compared with those for the 0–12‐month interval, which acted as a baseline value. Continuous variables are presented as mean(s.d.) or mean(s.e.m.) values, as appropriate. A Pearson χ2 test was used to compare categorical variables and a paired t test to compare continuous data. Data were analysed using IBM SPSS® version 22.0 (IBM, Armonk, New York, USA). P < 0·050 was considered statistically significant.

uous variables are presented as mean(s.d.) or mean(s.e.m.) values, as appropriate. A Pearson χ2 test was used to compare categorical variables and a paired t test to compare continuous data. Data were analysed using IBM SPSS® version 22.0 (IBM, Armonk, New York, USA). P < 0·050 was considered statistically significant. Results A total of 5011 men were recruited into the study, of whom 381 (7·6 per cent) had an AAA identified via screening. Overall, they were followed for a mean of 19·0(9·1) months from their initial screening appointment. Men with an AAA were older (age 72·6 versus 69·8 years; P < 0·001), had a higher BMI (28·1 versus 27·0 kg/m2; P < 0·001) and were more likely to be a current smoker (15·1 versus 5·2 per cent; P < 0·001) than those in the control group (Table 2). When comparing co‐morbidities to the control group, men with an AAA were more likely to have diabetes mellitus (18·8 versus 10·0 per cent), ischaemic heart disease (12·2 versus 4·4 per cent), high cholesterol (53·2 versus 30·8 per cent), previous stroke (6·1 versus 2·9 per cent) and a previous myocardial infarction (21·1 versus 5·8 per cent) (all P < 0·001). Table 2 Demographics of men included No AAA (n = 4630) AAA (n = 381) P † Age (years)* 69·8(3·5) 72·6(5·5) < 0·001‡ BMI (kg/m2)* 27·0(4·4) 28·1(4·3) < 0·001‡ AAA diameter at initial screening (mm)* 17·8(0·2) 36·1(0·7) < 0·001‡

Men with an AAA were older (age 72·6 versus 69·8 years; P < 0·001), had a higher BMI (28·1 versus 27·0 kg/m2; P < 0·001) and were more likely to be a current smoker (15·1 versus 5·2 per cent; P < 0·001) than those in the control group (Table 2). When comparing co‐morbidities to the control group, men with an AAA were more likely to have diabetes mellitus (18·8 versus 10·0 per cent), ischaemic heart disease (12·2 versus 4·4 per cent), high cholesterol (53·2 versus 30·8 per cent), previous stroke (6·1 versus 2·9 per cent) and a previous myocardial infarction (21·1 versus 5·8 per cent) (all P < 0·001). Table 2 Demographics of men included No AAA (n = 4630) AAA (n = 381) P † Age (years)* 69·8(3·5) 72·6(5·5) < 0·001‡ BMI (kg/m2)* 27·0(4·4) 28·1(4·3) < 0·001‡ AAA diameter at initial screening (mm)* 17·8(0·2) 36·1(0·7) < 0·001‡ Current smoker 237 of 4596 (5·2) 57 of 378 (15·1) < 0·001 Diabetes mellitus 454 of 4536 (10·0) 69 of 367 (18·8) < 0·001 IHD 197 of 4507 (4·4) 43 of 352 (12·2) < 0·001 High cholesterol 1401 of 4542 (30·8) 194 of 365 (53·2) < 0·001 Previous stroke 134 of 4612 (2·9) 23 of 376 (6·1) < 0·001 Previous MI 269 of 4614 (5·8) 80 of 379 (21·1) < 0·001 Values in parentheses are percentages unless indicated otherwise; * values are mean(s.d.). AAA, abdominal aortic aneurysm; IHD, ischaemic heart disease; MI, myocardial infarction. † Pearson χ2 test, except ‡ paired t test.

Current smoker 237 of 4596 (5·2) 57 of 378 (15·1) < 0·001 Diabetes mellitus 454 of 4536 (10·0) 69 of 367 (18·8) < 0·001 IHD 197 of 4507 (4·4) 43 of 352 (12·2) < 0·001 High cholesterol 1401 of 4542 (30·8) 194 of 365 (53·2) < 0·001 Previous stroke 134 of 4612 (2·9) 23 of 376 (6·1) < 0·001 Previous MI 269 of 4614 (5·8) 80 of 379 (21·1) < 0·001 Values in parentheses are percentages unless indicated otherwise; * values are mean(s.d.). AAA, abdominal aortic aneurysm; IHD, ischaemic heart disease; MI, myocardial infarction. † Pearson χ2 test, except ‡ paired t test. Quality of life For the PCS, scores in the AAA group were significantly lower at 0–12, 13–24 and 25–36 months than those in the control group (P < 0·001, P = 0·028 and P < 0·001 respectively) (Table 3 and Fig. 1 a). Over 37 months after screening, differences in PCS were non‐significant. For the MCS, scores were significantly lower immediately after screening in men with an AAA versus the control group (P < 0·001) (Table 4 and Fig. 1 b). However, after 12 months, MCS scores from the AAA cohort returned to baseline levels, equivalent to those of men with no AAA, and continued thus for the remainder of the follow‐up. Table 3 SF‐8 Physical Component Summary scores for men with and those without an abdominal aortic aneurysm for each time interval after initial screening Time from initial screening (months) PCS score P *

Quality of life For the PCS, scores in the AAA group were significantly lower at 0–12, 13–24 and 25–36 months than those in the control group (P < 0·001, P = 0·028 and P < 0·001 respectively) (Table 3 and Fig. 1 a). Over 37 months after screening, differences in PCS were non‐significant. For the MCS, scores were significantly lower immediately after screening in men with an AAA versus the control group (P < 0·001) (Table 4 and Fig. 1 b). However, after 12 months, MCS scores from the AAA cohort returned to baseline levels, equivalent to those of men with no AAA, and continued thus for the remainder of the follow‐up. Table 3 SF‐8 Physical Component Summary scores for men with and those without an abdominal aortic aneurysm for each time interval after initial screening Time from initial screening (months) PCS score P * No AAA AAA 0–12 51·4(7·9) 47·6(8·9) < 0·001 13–24 50·7(8·4) 49·5(8·7) 0·028 25–36 50·8(8·3) 47·5(10·0) < 0·001 ≥ 37 51·3(8·8) 48·5(9·4) 0·077 The SF‐8 questionnaire included eight questions adapted from the Medical Outcomes Study Short Form 36. Values are mean(s.d.). PCS, Physical Component Summary; AAA, abdominal aortic aneurysm. * ANOVA. Figure 1 Changes in SF‐8 scores for a Physical Component Summary (PCS) score and b Mental Component Summary (MCS) score after initial screening. Values are mean(s.e.m.). The SF‐8 questionnaire included eight questions adapted from the Medical Outcomes Study Short Form 36 BJS-10721-FIG-0001-cTable 4 SF‐8 Mental Component Summary scores for men with and those without an abdominal aortic aneurysm for each time interval after initial screening

Figure 1 Changes in SF‐8 scores for a Physical Component Summary (PCS) score and b Mental Component Summary (MCS) score after initial screening. Values are mean(s.e.m.). The SF‐8 questionnaire included eight questions adapted from the Medical Outcomes Study Short Form 36 BJS-10721-FIG-0001-cTable 4 SF‐8 Mental Component Summary scores for men with and those without an abdominal aortic aneurysm for each time interval after initial screening Time from initial screening (months) MCS score P * No AAA AAA 0–12 54·0(7·0) 51·9(8·3) < 0·001 13–24 53·5(7·5) 53·2(7·4) 0·610 25–36 54·0(7·1) 52·9(7·1) 0·884 ≥ 37 52·7(7·6) 53·7(6·7) 0·408 The SF‐8 questionnaire included eight questions adapted from the Medical Outcomes Study Short Form 36. Values are mean(s.d.). MCS, Mental Component Summary; AAA, abdominal aortic aneurysm. * ANOVA. Regression analysis was done, adjusting for the additional co‐variables collected, for both PCS and MCS (Tables S1 and S2, supporting information). The lower PCS scores remained significant across all AAA groups after screening (P < 0·001), whereas the MCS scores overall showed no differences between the AAA and control group (P = 0·443). The effect of growth rate on QoL was analysed by comparing QoL to the mean growth rate (cm/month) recorded for all patients with AAA. QoL for both MCS and PCS showed no relationship with growth rate (Figs S1 and S2, supporting information).

Regression analysis was done, adjusting for the additional co‐variables collected, for both PCS and MCS (Tables S1 and S2, supporting information). The lower PCS scores remained significant across all AAA groups after screening (P < 0·001), whereas the MCS scores overall showed no differences between the AAA and control group (P = 0·443). The effect of growth rate on QoL was analysed by comparing QoL to the mean growth rate (cm/month) recorded for all patients with AAA. QoL for both MCS and PCS showed no relationship with growth rate (Figs S1 and S2, supporting information). Impact of AAA and AAA growth ANOVA demonstrated a progressive reduction in the frequency with which men with an AAA had thought about their aneurysm in the preceding 4 weeks, at 13–24 months (P = 0·025), 25–36 months (P = 0·040) and 37 months or more (P = 0·005), all showing a significant reduction relative to baseline values at 0–12 months (Fig. 2 a; Table S3, supporting information). When men with a small AAA were asked how often they had thought about aneurysm growth in the preceding 4 weeks, there was also a significant reduction in frequency at 25–36 months (P = 0·004) and 37 months or more (P = 0·006) (Fig. 2 b; Table S4, supporting information). Figure 2 Likert scores of the frequency with which men had thought about a their abdominal aortic aneurysm (AAA) and b AAA growth in the preceding 4 weeks. Values are mean(s.d.). *P < 0·050, †P ≤ 0·010 (ANOVA)

Impact of AAA and AAA growth ANOVA demonstrated a progressive reduction in the frequency with which men with an AAA had thought about their aneurysm in the preceding 4 weeks, at 13–24 months (P = 0·025), 25–36 months (P = 0·040) and 37 months or more (P = 0·005), all showing a significant reduction relative to baseline values at 0–12 months (Fig. 2 a; Table S3, supporting information). When men with a small AAA were asked how often they had thought about aneurysm growth in the preceding 4 weeks, there was also a significant reduction in frequency at 25–36 months (P = 0·004) and 37 months or more (P = 0·006) (Fig. 2 b; Table S4, supporting information). Figure 2 Likert scores of the frequency with which men had thought about a their abdominal aortic aneurysm (AAA) and b AAA growth in the preceding 4 weeks. Values are mean(s.d.). *P < 0·050, †P ≤ 0·010 (ANOVA) BJS-10721-FIG-0002-cDiscussion This analysis demonstrates that men diagnosed with an AAA through screening have a transient reduction in mental QoL during the first year, but this then returns to normal. Furthermore, with time, a man diagnosed with an AAA is likely to think progressively less about it and its growth. This study also demonstrates that men with an AAA have a consistently lower physical QoL than men without an AAA, even when adjusted for co‐variables.

uring the first year, but this then returns to normal. Furthermore, with time, a man diagnosed with an AAA is likely to think progressively less about it and its growth. This study also demonstrates that men with an AAA have a consistently lower physical QoL than men without an AAA, even when adjusted for co‐variables. This is not the first time that a transient impact of mental QoL has been demonstrated in screening programmes: colorectal, prostate and breast cancer programmes have all been shown to have small effects on QoL that tend to diminish with long‐term follow‐up18, 19, 20. Faecal occult blood testing for colorectal cancer had no impact on mental QoL18, prostate cancer screening showed no influence on physical, psychological or social functioning19, and in women recalled after breast cancer screening, the initial anxiety and depression associated with the appointment had decreased significantly, even after a few days20. However, contemporaneous data on QoL in men screened for AAA were lacking. For AAA screening, the majority of the QoL evidence on this topic has come from the Multi‐Aneurysm Screening Study (MASS)9. Like the present study, the longer‐term mental impact appeared to be negligible in MASS. Using similar methodology, the MASS data set demonstrated a significant reduction in mental component scores for QoL 6 weeks after initial AAA diagnosis, yet, as in the present study, an improvement was seen 12 months after diagnosis, back to near‐baseline values.

r‐term mental impact appeared to be negligible in MASS. Using similar methodology, the MASS data set demonstrated a significant reduction in mental component scores for QoL 6 weeks after initial AAA diagnosis, yet, as in the present study, an improvement was seen 12 months after diagnosis, back to near‐baseline values. Smaller historical studies have shown similar findings. Lucarotti and colleagues11 showed that initial screening investigations caused mild anxiety that did not persist following AAA diagnosis. Wanhainen and co‐workers8 found that only individuals with a low QoL score before screening were susceptible to potential negative effects. If such transient negative effects on mental QoL are seen in men after screening, more work is warranted to evaluate the potential benefit that could occur from introducing counselling, with discussion surrounding the presence and growth of the AAA, and how this may impact on the man with the disease.

negative effects. If such transient negative effects on mental QoL are seen in men after screening, more work is warranted to evaluate the potential benefit that could occur from introducing counselling, with discussion surrounding the presence and growth of the AAA, and how this may impact on the man with the disease. The finding that PCS scores were lower in men with an AAA compared with controls, even after adjustment for confounders, probably reflects the co‐morbid nature of many men with AAA. This observation was also seen in MASS9, where men with an AAA had lower physical QoL scores, from both the Short Form 36 (QualityMetric, Lincoln, Rhode Island, USA) and EuroQol – 5D (EuroQol Group, Rotterdam, The Netherlands) questionnaires. AAA is an independent marker of cardiovascular risk, with a documented 3·0 per cent per year risk of cardiovascular death in patients with a small AAA21. AAA and cardiovascular disease have been shown to share risk factors22, 23.

Island, USA) and EuroQol – 5D (EuroQol Group, Rotterdam, The Netherlands) questionnaires. AAA is an independent marker of cardiovascular risk, with a documented 3·0 per cent per year risk of cardiovascular death in patients with a small AAA21. AAA and cardiovascular disease have been shown to share risk factors22, 23. It might be assumed that AAA growth would have a negative effect on the MCS score. Here, no association between these parameters was seen. Indeed, the longer a man had been involved in the surveillance programme, the less he reported thinking about AAA growth. These findings all support the suggestion that AAA screening does not have a significant or long‐term effect on QoL. These findings were echoed by Dahlberg and colleagues24, who demonstrated that the reassurance a patient received from increased AAA surveillance and the positive reinforcement of screening programmes eventually outweighed the potential negative effects that might be anticipated about worsening health. One of the key limitations of this work is the fact that QoL scores were not available before screening. As the scores of the no‐AAA participants acted as a control group, and remained stable throughout the study, it can be assumed that QoL would be similar before screening. There was a discrepancy in the longer‐term follow‐up rate between men with an AAA and no‐AAA controls, where reduced compliance may reflect decreasing engagement in those without an AAA.

nts acted as a control group, and remained stable throughout the study, it can be assumed that QoL would be similar before screening. There was a discrepancy in the longer‐term follow‐up rate between men with an AAA and no‐AAA controls, where reduced compliance may reflect decreasing engagement in those without an AAA. The recruitment method employed meant that some men were recruited at the time of the first screening scan and some during AAA surveillance. To allow for this, the time of recruitment was recorded and used to adjust analyses of AAA growth accordingly. This is reflected in the difference in the mean age of men with and those without an AAA at baseline. It was not possible to conduct a direct regression analysis of the SF‐8 data set owing to the clustered annual follow‐up of the men, yet ANOVA was able to provide a more suitable and relevant alternative. The conclusions drawn from this work are applicable only to men screened as positive for AAA. The overall effects of screening on the larger number of men with negative scans are not yet determined. Collaborators Other UK Aneurysm Growth Study investigators include: R. Pathak, M. Brooks, P. Hayes, C. Imray, J. Quarmby, S. Choksy, J. J. Earnshaw, C. P. Shearman, E. Grocott, T. Rix, I. Chetter, W. Tennant, G. Libertiny, T. Sykes, M. Dayer, L. Pike, A. Pherwani, C. Nice, N. Browning, C. McCollum, S. Yusuf, M. Gannon, J. Barwell, S. Baker, S. R. Vallabhaneni, A. Davies. Supporting information Appendix S1. Questionnaire sent to all recruited men for self‐completion Table S1 Regression analysis for Physical Component Summary score

Collaborators Other UK Aneurysm Growth Study investigators include: R. Pathak, M. Brooks, P. Hayes, C. Imray, J. Quarmby, S. Choksy, J. J. Earnshaw, C. P. Shearman, E. Grocott, T. Rix, I. Chetter, W. Tennant, G. Libertiny, T. Sykes, M. Dayer, L. Pike, A. Pherwani, C. Nice, N. Browning, C. McCollum, S. Yusuf, M. Gannon, J. Barwell, S. Baker, S. R. Vallabhaneni, A. Davies. Supporting information Appendix S1. Questionnaire sent to all recruited men for self‐completion Table S1 Regression analysis for Physical Component Summary score Table S2 Regression analysis for Mental Component Summary score Table S3 Likert scores of the frequency with which men had thought about their AAA in the preceding 4 weeks Table S4 Likert scores of the frequency with which men had thought about their AAA growth in the preceding 4 weeks Fig. S1 Linear regression for aneurysm growth rate and Physical Component Summary score Fig. S2 Linear regression for aneurysm growth rate and Mental Component Summary score Click here for additional data file. Acknowledgements A.S. and D.S. are funded by the National Institute for Health Research, and A.S. is also funded by the Academy of Medical Sciences (grant number SGCL13). The United Kingdom Aneurysm Growth Study is funded by the British Heart Foundation (grant number CS/14/2/30841). Disclosure: The authors declare no conflict of interest.

group are shown in parentheses. pCR, pathological complete response; cCR, clinical complete response; CRM, circumferential resection margin; LRC, locoregional control; DFS, disease‐free survival; GI, gastrointestinal; 5‐FU, 5‐fluorouracil; RFS, relapse‐free survival; OS, overall survival; CSS, cancer‐specific survival. A phase II trial66 was designed to assess the pCR (primary endpoint) following neoadjuvant therapy with panitumumab and RT. Of 19 enrolled patients, 17 were evaluable for pathology assessment. Although no pCR was observed, seven patients (41 per cent) had grade 3 Dworak pathological tumour regression. As the primary endpoint was not achieved, the authors were unable to make any recommendation for the use of panitumumab in treatment of LARC. Similar findings were reported in other phase II trials where the primary endpoint of pCR was not achieved67, 68, and toxicity was high69. EXPERT‐C70 was a randomized phase II trial of neoadjuvant CAPOX with or without cetuximab, followed by capecitabine‐based CRT with or without cetuximab, followed by surgery and then adjuvant CAPOX with or without cetuximab in 165 high‐risk patients with rectal cancer. Cetuximab did not improve the primary outcome (pCR), so it was not felt to have contributed significantly to increased radiation‐induced cytotoxicity. The EXPERT‐C trial did, however, find that TP53 tumour suppressor protein wild‐type status was a predictive biomarker in favour of cetuximab‐based therapy.

ectal cancer. Cetuximab did not improve the primary outcome (pCR), so it was not felt to have contributed significantly to increased radiation‐induced cytotoxicity. The EXPERT‐C trial did, however, find that TP53 tumour suppressor protein wild‐type status was a predictive biomarker in favour of cetuximab‐based therapy. The prospective phase II EXCITE trial71, published in 2017, focused on the addition of cetuximab to an irinotecan–capecitabine‐based neoadjuvant CRT regimen in 82 patients. Fourteen patients (17 per cent) had a pCR. As a side point of interest, contrary to the planned protocol, four patients achieved an endoscopically and MRI‐confirmed cCR, and were managed using the emerging watch‐and‐wait approach. Overall 24 patients (29 per cent) had an excellent clinical or pathological response. Using next‐generation sequencing, 46 per cent of matched biopsy–resection specimens were discrepant for EGFR pathway mutations. Intratumoral heterogeneity was suggested as a possible explanation, manifesting as a geographical biopsy miss or chemoradiation‐driven emergence of new mutations.

Introduction There is the potential for large routinely collected clinical data sets to improve healthcare delivery1, 2, 3. More specifically, systematic and statistical examination of operating records could provide novel insights into surgical practice4, 5, 6, 7, 8. Given that operating theatres are one of the costliest elements of a hospital9, these advances present important new opportunities for understanding how to improve surgical quality and efficiency10, 11. Examining routinely collected data from the operating theatre allows the creation of natural experiments (where exposure to the event of interest has not been manipulated experimentally12, 13). This allows the description, characterization and prediction of organizational, administrative and human factor‐related behaviours. Previous studies have implicated a number of factors that influence the amount of time it takes for a surgeon to perform a procedure. Beyond the surgeon's level of skill14, 15, research has demonstrated the influence of external drivers on procedure duration, including: patient characteristics15 (for example age14, 16, co‐morbidities such as BMI17, 18, 19), the surgical team20, 21, hospital size22 and case mix23. Recent reviews24, 25 have suggested that the way in which surgeons prepare for an operation can also affect performance, with some preparation techniques resulting in shorter operating times. The majority of these studies involved simulated contexts26, 27 and conclusions about best practice for preparation in the real world remain unclear.

have suggested that the way in which surgeons prepare for an operation can also affect performance, with some preparation techniques resulting in shorter operating times. The majority of these studies involved simulated contexts26, 27 and conclusions about best practice for preparation in the real world remain unclear. Operating lists present a natural experiment to test the hypothesis that surgeons will warm up progressively through practice, and that such benefits will be ameliorated when surgeons switch procedures in a theatre list. Examination of the impact of list order may also yield important information about service organization efficiency28. Anecdotal evidence suggests that some surgeons schedule what is perceived to be their most difficult operation first. This may be sensible but, to date, there is no evidence base to support such practice. Similarly, there has been no investigation of whether a list should be unimodal (1 procedure and/or method, for example open or minimally invasive) or multimodal in composition. Analysis of routinely collected information on surgical procedures presents an opportunity to move away from using clinical intuition and experience‐driven decisions to data‐driven decision‐making29, 30, 31, 32. To this end, the aim of this study was to examine the effect of operating list order on duration of operation in procedures performed across all 38 Spire Healthcare hospitals, one of the largest providers of private healthcare in the UK.

tion and experience‐driven decisions to data‐driven decision‐making29, 30, 31, 32. To this end, the aim of this study was to examine the effect of operating list order on duration of operation in procedures performed across all 38 Spire Healthcare hospitals, one of the largest providers of private healthcare in the UK. Methods This study received ethical approval from the Spire Healthcare Research Ethics Committee. Data were collated from Spire Healthcare's electronic patient record system (SAP SE, Walldorf, Germany) across all 38 UK hospitals. To practice in a private hospital in the UK, a surgeon must be on the General Medical Council's specialist register and hold, or have held in the past 5 years, a substantive consultant post within the National Health Service (NHS) or a Defence Medical Services hospital. Consequently, all procedures in these hospitals were performed by experienced consultant surgeons, assisted if appropriate by trainees. Patient demographics, procedural/operative information, prognosticators of operative outcome (ASA physical status grade33) and duration of hospital stay were included in the data set. Age was divided into groups, to allow adequate anonymization of data (18 or less, 19–24, 25–34, 35–44, 45–54, 55–64, 65–75, over 75 years). No additional information (such as sex or co‐morbidity) was available to the research team.

SA physical status grade33) and duration of hospital stay were included in the data set. Age was divided into groups, to allow adequate anonymization of data (18 or less, 19–24, 25–34, 35–44, 45–54, 55–64, 65–75, over 75 years). No additional information (such as sex or co‐morbidity) was available to the research team. The 35 most frequently observed operations in the data set were the primary focus of investigation. No restriction was placed on the surgical subspecialty, type of procedure performed, or techniques used by the operating surgeon to perform the procedure. The collated data were parsed to allow further analysis; individual surgeon's operating lists were identified, and any list that contained one of the most frequent 35 operations was included in the data set (98 291 theatre lists). Any other procedure performed during one of those lists was also included in the data set (255 757 procedures in total). Component operations were allocated absolute and procedure‐specific order numbers. The absolute list number refers to the number of procedures performed by the operating surgeon on the list, whereas the procedure‐specific list number is the number of times a certain procedure has been performed by the surgeon on a list. All cases that involved a change from the previous procedure were coded as a switch, because they involved some form of task switching (Table 1). Table 1 Illustration of absolute and procedure‐specific list number and switch classification

Component operations were allocated absolute and procedure‐specific order numbers. The absolute list number refers to the number of procedures performed by the operating surgeon on the list, whereas the procedure‐specific list number is the number of times a certain procedure has been performed by the surgeon on a list. All cases that involved a change from the previous procedure were coded as a switch, because they involved some form of task switching (Table 1). Table 1 Illustration of absolute and procedure‐specific list number and switch classification Procedure Absolute list no. Procedure‐specific list no. Switch Laparoscopic cholecystectomy 1 1 n.a. Open inguinal hernia repair 2 1 Yes Laparoscopic cholecystectomy 3 2 Yes Laparoscopic cholecystectomy 4 3 No Open inguinal hernia repair 5 2 Yes n.a., Not applicable. Procedures were also classified by method (open or minimally invasive surgery) and complexity, in accordance with the AXA Specalist Procedure Codes, which are used to grade the magnitude of surgical procedures in UK independent hospitals (Table S1, supporting information)34.

Procedure Absolute list no. Procedure‐specific list no. Switch Laparoscopic cholecystectomy 1 1 n.a. Open inguinal hernia repair 2 1 Yes Laparoscopic cholecystectomy 3 2 Yes Laparoscopic cholecystectomy 4 3 No Open inguinal hernia repair 5 2 Yes n.a., Not applicable. Procedures were also classified by method (open or minimally invasive surgery) and complexity, in accordance with the AXA Specalist Procedure Codes, which are used to grade the magnitude of surgical procedures in UK independent hospitals (Table S1, supporting information)34. From the original data set comprising 478 519 individual procedures, 8807 were excluded because they had no surgeon identifier associated with them and 1422 because no start time was recorded (Fig.  1). Thirty‐two duplicate records were also removed. Although such instances were relatively trivial to identify, a more difficult challenge in analysing routinely collected data lay in identifying cases where erroneous data might have been entered, for example the wrong start time or procedure type, or instances where missing data might have been due to the procedure ultimately being cancelled. All of these factors are likely to influence procedure order classification. This introduced noise, which it was reasoned would work against the hypothesis being tested (because the hypothesis suggests that the preceding operation (n – 1) affects the subsequent one; where data are missing, using n – 2 would make it more likely that the hypothesis would be rejected). Importantly, because of the statistical power afforded by a data set of this size, this noise in the data was tolerated rather than adjusting list order numbering, which would have required subjective inferences.

one; where data are missing, using n – 2 would make it more likely that the hypothesis would be rejected). Importantly, because of the statistical power afforded by a data set of this size, this noise in the data was tolerated rather than adjusting list order numbering, which would have required subjective inferences. Figure 1 Flow chart illustrating how sample sizes were determined for the linear‐mixed effects and matched analyses from the original data set BJS-10804-FIG-0001-cDuration of operation was employed as the primary outcome measure. Generally, this measure was defined as the time from skin incision to skin closure. In procedures where a skin incision is not made (such as endoscopic examination), the time taken for the procedure to be performed (defined as time from insertion to withdrawal of the endoscope for endoscopic examinations) was used. This measure was chosen because it is strongly correlated with surgeon performance, the focus of the study35, and previous studies36, 37, 38, 39, 40 have shown a relationship between this variable and clinical outcomes across a range of operations. In addition, duration of operation is recorded routinely in Spire Healthcare hospitals, and is not affected by loss of patients to follow‐up, unlike other measures of clinical outcome such as hospital death. Duration of hospital stay (in minutes) was investigated as a secondary outcome measure.

s a range of operations. In addition, duration of operation is recorded routinely in Spire Healthcare hospitals, and is not affected by loss of patients to follow‐up, unlike other measures of clinical outcome such as hospital death. Duration of hospital stay (in minutes) was investigated as a secondary outcome measure. Statistical analysis Operating times are zero‐bound and present a skewed distribution14. Therefore, all analyses focused on changes in natural logarithmic operating time, which can be seen as equivalent to measuring proportional time changes for relatively small magnitudes. A model was created to capture the effects of absolute list order, procedure‐specific list order and switching acrosss the full range of list positions to understand the relationship between list composition and duration of operation. It was also reasoned that different operations might yield distinctly different patterns of results, and so the analysis was conducted at a procedure level to allow individual cases to be compared against the same types of procedure.

ons to understand the relationship between list composition and duration of operation. It was also reasoned that different operations might yield distinctly different patterns of results, and so the analysis was conducted at a procedure level to allow individual cases to be compared against the same types of procedure. As the data set included information on factors known to correlate with postoperative outcomes (ASA grade5, 6, 41, 42 and age43, 44, 45), these potential confounders were controlled for. Different ages and ASA grades, along with different surgical procedures would imply different normal operating times; therefore, these baseline operating times were treated as random effects, shared by all operations of the same type, on the same age group and with the same ASA grade. In reality, operations on patients in similar age groups or with similar ASA grades will have similar baselines; for example, a 34‐year‐old patient with an ASA grade of II is more similar to a 40‐year‐old with an ASA grade of II than an 80‐year‐old with a grade of IV. However, assumptions were not made about the relationship between duration of operation and these factors. Instead, a statistically more conservative approach was adopted by assuming that these random effects were independent between pairs of operations (unless all 3 of these variables were identical). Restricted‐likelihood maximization via the Lme4 package46 was used to fit the linear mixed‐effects model in R (R Project for Statistical Computing, Vienna, Austria), and the effect size and probability values (α threshold of P < 0·050) estimated for the fixed effects of list order (absolute and procedure‐specific) and switching.

‐likelihood maximization via the Lme4 package46 was used to fit the linear mixed‐effects model in R (R Project for Statistical Computing, Vienna, Austria), and the effect size and probability values (α threshold of P < 0·050) estimated for the fixed effects of list order (absolute and procedure‐specific) and switching. For a closer examination of the primary effects observed in the data, a form of matched analysis was subsequently performed on a subset of the data. This analysis was inspired by (but not identical to) a novel method for identifying causal relationships in natural experiments47. Here, the data were stratified into multiple sets of pairs by explicitly matching individuals who had the same age, ASA grade and operation type, but differed in list order by one position. Specifically, the data were first filtered by procedure type, then all cases that were ordered as procedures 1 and 2 were separated into different data frames (list order 1 and list order 2). All cases in list order 1 (presented in a randomly determined order) were examined to determine which elements of list order 2 had the same age group and ASA grade. If a case could be matched, this pair was included in the subsequent analysis and removed from the pool. In the event of multiple matches from list order 2 with list order 1, the computer program randomly selected one case for the pair and the non‐selected case(s) were returned to the pool for a possible future match. Each patient was paired to only one other individual, and only patients for whom a pair could be found were included. The matching process terminated when no more unique pairs could be found. This approach represents a method for statistically controlling for all the potential confounding variables available in the data set.

atient was paired to only one other individual, and only patients for whom a pair could be found were included. The matching process terminated when no more unique pairs could be found. This approach represents a method for statistically controlling for all the potential confounding variables available in the data set. In addition to these primary analyses, it was determined whether these effects translated across surgical method (open and minimally invasive); and whether the impact of list order varied according to procedure complexity. The procedures were separated by classifying them as those performed using open or minimally invasive techniques to address the first question, and by complexity for the second question, and the matched analysis repeated. To provide a measure of the magnitude (or effect size) of the analysed variables on list order, the change in the log scale for the linear mixed‐effects model was reported, along with mean difference in the log duration of the procedures in the matched analysis. Change in log duration is, to a high degree of approximation, the geometric average of the proportional percentage change in procedure duration; the percentage change is therefore referred to for all outcomes to provide an intuitive means of understanding these data.

duration of the procedures in the matched analysis. Change in log duration is, to a high degree of approximation, the geometric average of the proportional percentage change in procedure duration; the percentage change is therefore referred to for all outcomes to provide an intuitive means of understanding these data. Results Surgical lists containing the 35 operations performed most frequently between 1 April 2013 and 31 May 2015 were analysed (255 757 procedures). The linear mixed‐effects model revealed statistically reliable differences in changes in duration of operation for the fixed effects of absolute list order, procedure‐specific list order and switching when pooled across operations (all P < 0·001). The effect sizes (which can be treated as percentage changes in operating time as a function of list position change) were largest for procedure‐specific list order and switching. The percentage change in duration of operation for each procedure is shown in Fig.  2.

tching when pooled across operations (all P < 0·001). The effect sizes (which can be treated as percentage changes in operating time as a function of list position change) were largest for procedure‐specific list order and switching. The percentage change in duration of operation for each procedure is shown in Fig.  2. Figure 2 Forest plots showing percentage change in duration of operation for the top 35 procedures in the database based on the influence of fixed model parameters: a absolute list order, b procedure‐specific list order and c procedure switch. Negative values indicate the percentage reduction in duration of operation given an increase in each parameter, and positive values the percentage increase. The top row in each panel shows the overall effect of each fixed parameter. Error bars represent 95 per cent confidence intervals. Procedures are identified by AXA Specialist Procedure Codes (Table S1, supporting information)

of operation given an increase in each parameter, and positive values the percentage increase. The top row in each panel shows the overall effect of each fixed parameter. Error bars represent 95 per cent confidence intervals. Procedures are identified by AXA Specialist Procedure Codes (Table S1, supporting information) BJS-10804-FIG-0002-cFor absolute list order, there was a statistically significant list order effect, suggesting that each position in the list decreases duration of operation by 0·39 (95 per cent c.i. 0·35 to 0·44) per cent across all operations. These effects were substantially greater when considering the benefits acrued when the same procedure was repeated in a list, with the effect of procedure‐specific list order leading to a 0·98 (0·88 to 1·09) per cent reduction in duration. There was a cost associated with switching between different procedures in a list, leading to an increase in duration of operation by 6·48 (6·05 to 6·90) per cent for each increase in position in list order. To illustrate this effect on individual procedures, Fig.  3 shows the influence of task repetition on operating time for three routine procedures. There was a marked similarity in the pattern of results across these procedures, indicating that fatigue, inattention and monotony‐related performance impairment following multiple repetitions of a procedure were not present in these data.

the influence of task repetition on operating time for three routine procedures. There was a marked similarity in the pattern of results across these procedures, indicating that fatigue, inattention and monotony‐related performance impairment following multiple repetitions of a procedure were not present in these data. Figure 3 Duration of operation as a function of procedure‐specific list order for three routine procedures: a primary open inguinal hernia repair (with mesh); b oesophagogastroduodenoscopy (with biopsy of lesion) and c lens implant for cataract. Error bars represent 95 per cent confidence intervals BJS-10804-FIG-0003-cUsing the same linear model to analyse duration of hospital stay, overall a statistically reliable effect of absolute and procedure‐specific list order was found (both P < 0·001), but not for switching (P = 0·136). Specifically, the data indicated that, for every increase in absolute list position, duration of hospital stay increased by 0·55 (95 per cent c.i. 0·50 to 0·61) per cent. However, procedure‐specific list order resulted in a decrease in length of stay by 0·72 (0·58 to 0·85) per cent.

, but not for switching (P = 0·136). Specifically, the data indicated that, for every increase in absolute list position, duration of hospital stay increased by 0·55 (95 per cent c.i. 0·50 to 0·61) per cent. However, procedure‐specific list order resulted in a decrease in length of stay by 0·72 (0·58 to 0·85) per cent. The matched analysis allowed a focus on the impact of repeating a procedure in more detail on the primary outcome measure of duration of operation. A total of 48 632 pairs were matched from of a maximum 48 699 cases (99·9 per cent of all cases; the sample size was constrained by the number of cases with a procedure‐specific list order of 2 in the data set). Here, a statistically reliable improvement was found (P < 0·050) in 29 of the 35 procedures; the change in operating time ranged from a reduction of 3·84 (95 per cent c.i. 1·47 to 6·21) per cent to 17·25 (10·69 to 23·81) per cent (Fig.  4). Pooling across all 35 procedures showed a 6·18 (5·64 to 6·72) per cent reduction in operating time on average when performing the second procedure relative to the first (P < 0·001). Figure 4 Forest plot from matched analysis illustrating the percentage change in duration of operation for procedure‐specific list order 2 procedures compared with list order 1. Error bars represent 95 per cent confidence intervals. Procedures are identified by AXA Specialist Procedure Codes (Table S1, supporting information)

4 Forest plot from matched analysis illustrating the percentage change in duration of operation for procedure‐specific list order 2 procedures compared with list order 1. Error bars represent 95 per cent confidence intervals. Procedures are identified by AXA Specialist Procedure Codes (Table S1, supporting information) BJS-10804-FIG-0004-cSupplementary analyses allowed these results to be assessed in more detail. Conducting the matched analysis separately for open and minimally invasive procedures showed comparable effects of list order on duration of operation, indicating that this phenomenon transcends surgical method (Fig.  5 a). Figure 5 Percentage change in duration of operation time according to a surgical method and b complexity for matched wake analysis. Error bars represent 95 per cent confidence intervals BJS-10804-FIG-0005-cFinally, by separating procedures based on their complexity, a weak positive trend was found, but overall there were no differences in effect size as a function of complexity, with the reduction in operating time ranging from 4·80 to 7·50 per cent (Fig.  5 b).

Figure 5 Percentage change in duration of operation time according to a surgical method and b complexity for matched wake analysis. Error bars represent 95 per cent confidence intervals BJS-10804-FIG-0005-cFinally, by separating procedures based on their complexity, a weak positive trend was found, but overall there were no differences in effect size as a function of complexity, with the reduction in operating time ranging from 4·80 to 7·50 per cent (Fig.  5 b). Discussion These results support the suggestion that areas where surgical performance can be improved can be identified through analysis of routinely collected large data sets. The order in which surgical procedures are done has a relationship with their duration. The effects were similar for open and minimally invasive procedures, and procedures of differing complexity. In contrast, switching between different procedures resulted in increased duration of operation. These changes in operating time are particularly significant given that they were observed in highly trained individuals with several years of practice. The results are all the more remarkable when the wide range of factors that can potentially influence procedure duration is considered. The consistency of this pattern of results across procedure type, method and complexity provides compelling evidence that operating list order plays an important role in surgical performance.

ults are all the more remarkable when the wide range of factors that can potentially influence procedure duration is considered. The consistency of this pattern of results across procedure type, method and complexity provides compelling evidence that operating list order plays an important role in surgical performance. The effects of list order on duration of hospital stay are more difficult to interpret. A reduction in length of stay demonstrated in some investigations, with an increase in duration demonstrated in others, brings into question the practical significance of these findings. They are likely to reflect the complex, multifactorial nature of duration of hospital stay, which is affected to a much greater degree by social and institutional factors than operating time. Consequently, efforts were focused on understanding the effects of list order on duration of operation in the follow‐up matched analysis. The data have practical implications. There was an overall 6·18 per cent saving in operating time (as large as 17·25 per cent in some procedures) for repeating the same procedure on the list, even after controlling for age and ASA grade. This control is particularly important as anecdotal evidence indicates that surgeons typically take these factors into account when compiling their lists, but the data indicate that the process of list ordering itself influences the duration of operation above and beyond the variance captured by age and ASA grade.

ade. This control is particularly important as anecdotal evidence indicates that surgeons typically take these factors into account when compiling their lists, but the data indicate that the process of list ordering itself influences the duration of operation above and beyond the variance captured by age and ASA grade. From the perspective of service delivery, the results indicate that lists involving a combination of procedures take longer to complete than those that include only one procedure type (the overall cost of switching was estimated as a 6·48 per cent increase in duration). Although increased time with task switching has long been established in experimental psychology48, this is the first demonstration of its influence in surgical performance. Where possible, theatre lists should be confined to a single procedure type and method. One limitation of the present analyses is that the data cannot address the issue of the mechanisms of performance facilitation. However, the results do triangulate with existing empirical work showing that warm up reduces operating times and task switching increases completion times25. This indicates a need to explore these areas through further research. An understanding of the optimal preparatory routines to drive performance improvement will be necessary to harness the potential of the phenomenon reported here.

g that warm up reduces operating times and task switching increases completion times25. This indicates a need to explore these areas through further research. An understanding of the optimal preparatory routines to drive performance improvement will be necessary to harness the potential of the phenomenon reported here. The present data were derived from a private healthcare provider. It is worth noting that the majority of UK healthcare delivery is provided by the NHS, but the private sector is used by 10–22 per cent of the population (depending on region)49, 50. This data set was chosen for two reasons. First, the data could be pooled across multiple hospitals (a considerable logistical challenge in the NHS)51. Second, all operations performed in private UK hospitals must be conducted by trained consultant surgeons, eliminating training effects and ensuring that all procedures on a theatre list were performed by a single practitioner. Further work is required to establish whether these effects also exist in the NHS, where surgeons performing the procedures vary in experience. Environmental factors such as distractions, ward rounds and general resource constraints and prioritization also differ between private and NHS hospitals.

single practitioner. Further work is required to establish whether these effects also exist in the NHS, where surgeons performing the procedures vary in experience. Environmental factors such as distractions, ward rounds and general resource constraints and prioritization also differ between private and NHS hospitals. The effects reported here are modest. Yet, it is evident that the aggregation of even small gains has the potential to produce substantial benefits when scaled across a health service; this is particularly important given the growing economic pressures to optimize elective surgery10. Although the full extent of the impact of these effects remains to be seen, to put the present results in context, 52 per cent of the procedures analysed involved switching from the preceding operation, and the present results demonstrate that switching is responsible for a 6·48 per cent increase in duration of operation. Over the course of a year in a typical large hospital, avoiding switching could lead to a meaningful reduction in operating time. Editor's comments Supporting information Table S1 Procedure code modality and difficulty classification Click here for additional data file. Acknowledgements T.W.P. and F.M. contributed equally to this work. The authors acknowledge the assistance provided by C. Sinfield, A. Tchaikovsky and J. J. DeGorter at Spire Healthcare. This work was supported by funding from Rays of Hope charity to T.W.P. and a Medical Research Council grant (R17427) to M.M.‐W. Disclosure: The authors declare no conflict of interest.

Introduction Rectal cancer treatment has continued to improve in recent years as a result of optimized surgical technique, advances in staging, pathological quality control and multidisciplinary management. Neoadjuvant chemoradiotherapy (CRT) is considered the standard of care for locally advanced rectal cancer (LARC). It is well recognized that the response to neoadjuvant CRT is both variable and unpredictable for the individual patient, and techniques to risk‐stratify patients and predict response are an expanding area of research. Favourable responses to CRT are independently associated with conferring a long‐term survival advantage to patients who undergo resection, and in more recent years the possibility of deferral of surgery and organ preservation has also been raised1. A complete response to CRT may be classified as either a clinical complete response (cCR) or a pathological complete response (pCR). Although the two terms are often used interchangeably, these responses are assessed differently, and one does not necessarily imply the other. A pCR is based on pathological findings after resection, commonly using the Dworak or Mandard tumour regression grading systems. A cCR is defined according to a combination of clinical examination (including digital rectal examination), radiological (in particular diffusion‐weighted MRI) and endoscopic appearances.

er. A pCR is based on pathological findings after resection, commonly using the Dworak or Mandard tumour regression grading systems. A cCR is defined according to a combination of clinical examination (including digital rectal examination), radiological (in particular diffusion‐weighted MRI) and endoscopic appearances. Following the initial description by Habr‐Gama and colleagues2, there are now a growing number of series reporting the use of neoadjuvant CRT as the sole treatment for rectal cancer that undergoes a cCR, resulting in further interest in the role of organ preservation in rectal cancer3. It is, however, important to be able to differentiate which tumours are more susceptible to undergoing a cCR. At present, the most reliable predictor of an increased response is tumour stage, with early tumours more likely to display a cCR. The use of CRT in combination with local excision is perhaps becoming better defined in early T1 rectal cancers, but its value in more advanced cancer is less clear4. The STAR‐TReC trial (ISRCTN14240288)5 will compare three different strategies for more advanced tumours up to T3b N0, and assess the feasibility of randomizing to a trial with organ preservation arms. However, the role of neoadjuvant CRT as sole treatment for even more locally advanced tumours that perhaps threaten the circumferential resection margin (CRM) is unknown, and it is likely that studies examining such tumours will need to incorporate the development of intensified CRT regimens.

h organ preservation arms. However, the role of neoadjuvant CRT as sole treatment for even more locally advanced tumours that perhaps threaten the circumferential resection margin (CRM) is unknown, and it is likely that studies examining such tumours will need to incorporate the development of intensified CRT regimens. Patients who have an apparent cCR may be offered entry into a watch‐and‐wait surveillance policy after a full and complete discussion. If patients are fit for intervention, salvage surgery is recommended for those who display tumour regrowth, which is most often luminal rather than nodal1. There is clearly an interest in both predicting patients who may undergo a cCR or pCR and/or improving cCR and/or pCR rates as there are currently no reliable clinical (apart from earlier stage), biochemical or molecular predictive biomarkers in clinical practice. Radiotherapy (RT) is typically delivered via either a short‐ or long‐course strategy, the latter being employed to downstage tumours. A recent short study by the UK National Bowel Cancer Audit6 revealed that the median time from completion of CRT to surgical resection is currently 11 weeks in the UK, suggesting that the concept of delayed resection is gaining traction in clinical practice. A recent study7 suggested that increasing the interval between the end of CRT and surgical resection improves the response rate. Similarly, short‐course RT may be combined with a delayed interval to surgery; the recent Stockholm III trial8 has demonstrated improved tumour regression over traditional short‐course treatment.

cent study7 suggested that increasing the interval between the end of CRT and surgical resection improves the response rate. Similarly, short‐course RT may be combined with a delayed interval to surgery; the recent Stockholm III trial8 has demonstrated improved tumour regression over traditional short‐course treatment. Radiosensitizers are employed routinely to improve the radiosensitivity of rectal cancer to RT; the standard of care is a concurrent single‐agent fluoropyrimidine. A number of studies have analysed novel agents or combination therapies that aim to improve radiosensitivity and cCR and/or pCR rates. The critical target for RT is DNA and the accumulation of DNA damage, particularly DNA double‐strand breaks, and the ability of tumour cells to repair this damage, contributes significantly to the therapeutic effect. Some agents and combination therapies (such as oxaliplatin, irinotecan and poly(ADP‐ribose) polymerase (PARP) inhibitors) might typically take advantage of this by creating additional DNA damage or inhibiting DNA damage repair, exacerbating the effects of RT. The aim of this review was to summarize the current and novel agents that have been employed in the treatment of LARC, and to consider their role in the context of cCR and organ preservation. A summary of radiosensitizing agents is provided in Fig. 1.

age or inhibiting DNA damage repair, exacerbating the effects of RT. The aim of this review was to summarize the current and novel agents that have been employed in the treatment of LARC, and to consider their role in the context of cCR and organ preservation. A summary of radiosensitizing agents is provided in Fig. 1. Figure 1 Summary of current and potential radiosensitizing agents. 5‐dFCR, 5′‐deoxy‐5‐fluorocytidine; 5‐dFUR, 5′‐deoxy‐5‐fluorouridine; 5‐FU, 5‐fluorouracil; dUMP, deoxyuridine monophosphate; dTMP, deoxythymidine monophosphate; SSB, single‐strand break; DSB, double‐strand break; TOPO, topoisomerase; EGFR, epidermal growth factor receptor; VEGF, vascular endothelial growth factor; PARP, poly(ADP‐ribose)polymerase; COX, cyclo‐oxygenase; HDAC, histone deacetylase

fluorouracil; dUMP, deoxyuridine monophosphate; dTMP, deoxythymidine monophosphate; SSB, single‐strand break; DSB, double‐strand break; TOPO, topoisomerase; EGFR, epidermal growth factor receptor; VEGF, vascular endothelial growth factor; PARP, poly(ADP‐ribose)polymerase; COX, cyclo‐oxygenase; HDAC, histone deacetylase BJS-10993-FIG-0001-cMethods A literature search was performed for published full‐text articles using PubMed, Cochrane and Scopus databases using the search criteria string (‘radiosensitiser’ OR ‘radiosensitizer’) AND (‘rectal’ OR ‘rectum’) AND ‘cancer’ in November 2017. Additional papers were detected by scanning the references of relevant papers. Search results were initially included based on a relevant title, and these papers were then read in full. Inclusion criteria were: papers published in the English language, those with a focus on rectal cancer, all study types, and articles published between 1990 and 2017. Studies focusing on a primary malignancy other than rectal cancer were excluded. Two reviewers were involved at each stage, with search results being loaded into the Covidence system to enable joint reviews to take place methodically. Once eligible articles had been identified, a search was undertaken to exclude duplicated results or duplicated data sets to produce the final list of papers included (Fig. 2). Figure 2 PRISMA diagram showing selection of articles for review

BJS-10993-FIG-0001-cMethods A literature search was performed for published full‐text articles using PubMed, Cochrane and Scopus databases using the search criteria string (‘radiosensitiser’ OR ‘radiosensitizer’) AND (‘rectal’ OR ‘rectum’) AND ‘cancer’ in November 2017. Additional papers were detected by scanning the references of relevant papers. Search results were initially included based on a relevant title, and these papers were then read in full. Inclusion criteria were: papers published in the English language, those with a focus on rectal cancer, all study types, and articles published between 1990 and 2017. Studies focusing on a primary malignancy other than rectal cancer were excluded. Two reviewers were involved at each stage, with search results being loaded into the Covidence system to enable joint reviews to take place methodically. Once eligible articles had been identified, a search was undertaken to exclude duplicated results or duplicated data sets to produce the final list of papers included (Fig. 2). Figure 2 PRISMA diagram showing selection of articles for review BJS-10993-FIG-0002-cStandard chemotherapy regimens 5‐Fluorouracil 5‐Fluorouracil (5‐FU) is an antimetabolite fluoropyrimidine. It is one of the most common chemotherapeutic agents used in cancer treatment, in particular breast and colorectal cancer. It was the first agent to be used as a radiosensitizer in conjunction with RT, predominantly in colorectal cancer. It works by inhibiting essential biosynthetic processes, and also by affecting cellular DNA and RNA functions. The mechanism of cytotoxicity of 5‐FU has been ascribed to the misincorporation of fluoronucleotides into RNA and DNA, and to inhibition of the nucleotide synthetic enzyme thymidylate synthase9.

rectal cancer. It works by inhibiting essential biosynthetic processes, and also by affecting cellular DNA and RNA functions. The mechanism of cytotoxicity of 5‐FU has been ascribed to the misincorporation of fluoronucleotides into RNA and DNA, and to inhibition of the nucleotide synthetic enzyme thymidylate synthase9. There are a number of mechanisms by which 5‐FU could increase radiation sensitivity at the cellular level. One is thought to involve the killing of S‐phase cells, which are relatively radioresistant10, 11. This does not account for all of the increased radiation sensitivity produced by the drug because non‐cytotoxic concentrations can also increase sensitivity. Radiosensitization under non‐cytotoxic conditions occurs only when cells are incubated with the drug before and during radiation treatment. Several studies have suggested that 5‐FU should be given continuously during a course of fractionated radiation to achieve radiosensitization of most fractions12, 13. UK National Institute for Health and Care Excellence guidance14 focusing on stage III tumours, examining randomized comparisons of bolus versus infusional regimens, suggests that infusional therapy is equivalent to bolus treatment in terms of effectiveness, but has relatively reduced toxicity. Owing to concerns regarding the increased cost of infusional treatment and patient inconvenience, there remains geographical variation across the UK in the technique employed14, although 5‐FU has largely been superseded by oral capecitabine. Trials of 5‐FU are summarized in Table 1.

but has relatively reduced toxicity. Owing to concerns regarding the increased cost of infusional treatment and patient inconvenience, there remains geographical variation across the UK in the technique employed14, although 5‐FU has largely been superseded by oral capecitabine. Trials of 5‐FU are summarized in Table 1. Table 1 Summary of fluoropyrimidine agents Results (%)* Reference Phase Disease stage Test drug Single/combination regimen Cohort size pCR rate cCR rate Other endpoints Toxicity 15 III T3–4 N–/+ 5‐FU Adjuvant RT versus RT + 5‐FU 204 n.a. – 5‐year recurrence 41·5 (62·7) Increased risk of GI or haematological problems with 5‐FU use; only 1 severe 16 III T3–4 N+/– 5‐FU Neoadjuvant RT versus RT + 5‐FU 773 11·4 (3·6) – 5‐year LR 8·1 (16·5) 5‐year OS 67 (67) Grade 3–4 toxicity 14·6 (3·6) 17 III T3–4 N–/+ 5‐FU Neoadjuvant versus adjuvant RT + 5‐FU 267 15·0 0·8 5‐year DFS 64·7 (53·4) 5‐year OS 74·5 (65·6) Grade 4 GI disturbance 24 (13) 18 III T3–4 N–/+ 5‐FU Neoadjuvant RT versus RT + 5‐FU 1011 13·7 (5·3) – – – 19 III T3–4 N–/+ 5‐FU Neoadjuvant versus adjuvant RT + 5‐FU 823 8 – 5‐year LRR 6 (13) 5‐year DFS 68 (65) 5‐year OS 76 (74) Grade 3–4 toxicity 27 versus 40 20 II T3–4 N–/+ Capecitabine Single 95 12 – Downstaging 76 Grade 3 toxicity 3 21 II T3–4 N–/+ Capecitabine Single 54 18 – Downstaging 51 Sphincter salvage 67 Grade 3–4 GI toxicity 2 22 II T3–4 N–/+ Capecitabine Single 31 7 – Downstaging 54 3‐year DFS 60 3‐year OS 77 Grade 3–4 GI toxicity 36 Proctitis 32 23 III T3–4 N–/+ Capecitabine Capecitabine + oxaliplatin versus 5‐FU 42 24 (14) – Downstaging 81 (67) Grade 3 GI toxicity 19 (14)

5‐year OS 76 (74) Grade 3–4 toxicity 27 versus 40 20 II T3–4 N–/+ Capecitabine Single 95 12 – Downstaging 76 Grade 3 toxicity 3 21 II T3–4 N–/+ Capecitabine Single 54 18 – Downstaging 51 Sphincter salvage 67 Grade 3–4 GI toxicity 2 22 II T3–4 N–/+ Capecitabine Single 31 7 – Downstaging 54 3‐year DFS 60 3‐year OS 77 Grade 3–4 GI toxicity 36 Proctitis 32 23 III T3–4 N–/+ Capecitabine Capecitabine + oxaliplatin versus 5‐FU 42 24 (14) – Downstaging 81 (67) Grade 3 GI toxicity 19 (14) Haematological 19 (14) 24 III T3–4 N–/+ Capecitabine Capecitabine versus 5‐FU 392 14 (5) – LR 6 (7) 5‐year OS 76 (67) 3‐year DFS 75 (67) Grade 3–4 GI toxicity 9 (2) 25 III T3–4 N–/+ Capecitabine Capecitabine +/– oxaliplatin versus 5‐FU +/– oxaliplatin 1608 20·7 (17·8) – Downstaging 21·1 (21·3) Sphincter salvage 59·3 (59·4) Grade 3–5 GI toxicity 11·7 (11·7) Addition of oxaliplatin increased GI disturbance (P < 0·001) * Results for control group are shown in parentheses. pCR, pathological complete response; cCR, clinical complete response; 5‐FU, 5‐fluorouracil; RT, radiotherapy; n.a., not applicable; GI, gastrointestinal; LR, local recurrence; OS, overall survival; DFS, disease‐free survival; LRR, locoregional recurrence.

P < 0·001) * Results for control group are shown in parentheses. pCR, pathological complete response; cCR, clinical complete response; 5‐FU, 5‐fluorouracil; RT, radiotherapy; n.a., not applicable; GI, gastrointestinal; LR, local recurrence; OS, overall survival; DFS, disease‐free survival; LRR, locoregional recurrence. Krook and colleagues15 first conducted an RCT to assess adjuvant RT with or without systemic bolus 5‐FU chemotherapy, and confirmed an improvement in local relapse rates with a survival benefit in favour of 5‐FU‐based RT in comparison with RT alone. The phase III French Federation Francophone de Cancerologie Digestive (FFCD) 9203 study16 randomized patients with stage II–III rectal cancer to receive RT alone or with infusional 5‐FU/leucovorin. Patients in both arms subsequently underwent surgery and four cycles of 5‐FU/leucovorin. The preoperative chemoradiation arm showed a significant improvement in pCR rate (11·4 versus 3·6 per cent; P < 0·050) and local relapse rate (8·1 versus 16·5 per cent; P < 0·050). The 5‐year survival in both arms was 67 per cent.

Patients in both arms subsequently underwent surgery and four cycles of 5‐FU/leucovorin. The preoperative chemoradiation arm showed a significant improvement in pCR rate (11·4 versus 3·6 per cent; P < 0·050) and local relapse rate (8·1 versus 16·5 per cent; P < 0·050). The 5‐year survival in both arms was 67 per cent. The National Surgical Adjuvant Breast and Bowel Project (NSABP) R‐03 phase III study17 compared the use of neoadjuvant and adjuvant CRT in patients with T3–T4 or node‐positive rectal cancer using 5‐FU and leucovorin. Those receiving neoadjuvant therapy had a pCR rate of 15·0 per cent and 5‐year disease‐free survival (DFS) rate of 64·7 per cent. Among those undergoing adjuvant CRT, 39·2 per cent had sphincter‐saving surgery (versus 47·8 per cent in the neoadjuvant cohort) and a DFS rate of 53·4 per cent. Five‐year overall survival rates were 74·5 and 65·6 per cent in the neoadjuvant and adjuvant treatment groups respectively, supporting the use of CRT before rather than after operation. In a 2012 Cochrane review, Petersen and colleagues26 considered the use of 5‐FU for additional adjuvant chemotherapy. The pooled data from 21 RCTs, including almost 10 000 patients, found improved DFS and overall survival with use of adjuvant chemotherapy. Owing to lack of tumour stage‐specific data, however, it was not possible to draw a link of benefit to specific locally advanced tumours, potentially indicating an area for further work.

ooled data from 21 RCTs, including almost 10 000 patients, found improved DFS and overall survival with use of adjuvant chemotherapy. Owing to lack of tumour stage‐specific data, however, it was not possible to draw a link of benefit to specific locally advanced tumours, potentially indicating an area for further work. The phase III European Organisation for Research and Treatment of Cancer (EORTC) 22921 study18 randomized patients with stage II–III rectal cancer to neoadjuvant RT alone versus RT with concurrent bolus 5‐FU/leucovorin, with subsequent randomization to postoperative chemotherapy or not. The authors concluded that adding 5‐FU‐based chemotherapy, either before (as part of CRT) or after operation, conferred a significant advantage in terms of local control. The German Rectal Cancer Study Group19 randomly assigned 823 patients with clinical stage T3 or T4 or node‐positive disease to receive either preoperative or postoperative CRT. The results showed a significantly lower 5‐year cumulative incidence of local relapse in favour of the preoperative treatment group (6 versus 13 per cent; P = 0·006). Five‐year DFS (68 versus 65 per cent) and overall survival (76 versus 74 per cent) rates were similar. Significant tumour downstaging was seen after preoperative combined treatment, with a pCR rate of 8 per cent.

incidence of local relapse in favour of the preoperative treatment group (6 versus 13 per cent; P = 0·006). Five‐year DFS (68 versus 65 per cent) and overall survival (76 versus 74 per cent) rates were similar. Significant tumour downstaging was seen after preoperative combined treatment, with a pCR rate of 8 per cent. In a pooled analysis of 5‐FU phase II–III trials27 including 3157 patients, the pCR rate was 13·5 per cent. On multivariable analysis, statistically significant factors for a higher pCR rate were the addition of a second chemotherapy agent and the method of continuous infusion.

incidence of local relapse in favour of the preoperative treatment group (6 versus 13 per cent; P = 0·006). Five‐year DFS (68 versus 65 per cent) and overall survival (76 versus 74 per cent) rates were similar. Significant tumour downstaging was seen after preoperative combined treatment, with a pCR rate of 8 per cent. In a pooled analysis of 5‐FU phase II–III trials27 including 3157 patients, the pCR rate was 13·5 per cent. On multivariable analysis, statistically significant factors for a higher pCR rate were the addition of a second chemotherapy agent and the method of continuous infusion. Capecitabine The development of an oral 5‐FU drug was driven by the desire to overcome the perceived limitations associated with intravenous infusion of 5‐FU, such as extended hospital stay, the need for intravenous lines and associated healthcare costs. Capecitabine (Xeloda®; Roche, Basle, Switzerland) is an oral prodrug of 5‐FU; it is a fluoropyrimidine carbamate that undergoes a three‐step in vivo enzymatic conversion to 5‐FU. The final step is mediated by the enzyme thymidine phosphorylase, which is upregulated in tumour tissue compared with adjacent healthy tissue. This theoretically allows selective activation of the drug and low systemic toxicity28. After oral administration, capecitabine passes rapidly and extensively through the intestinal membrane as an intact molecule. Capecitabine is not cytotoxic itself; the only cytotoxic moiety is 5‐FU, which is generated preferentially in human cancer cells. Preferential activation of capecitabine to 5‐FU in malignant tumour was demonstrated in animal models bearing human xenografts29. Table 1 provides an overview of clinical trials of capecitabine.

citabine is not cytotoxic itself; the only cytotoxic moiety is 5‐FU, which is generated preferentially in human cancer cells. Preferential activation of capecitabine to 5‐FU in malignant tumour was demonstrated in animal models bearing human xenografts29. Table 1 provides an overview of clinical trials of capecitabine. The first phase II trials20, 21 showed that RT plus capecitabine is well tolerated and easier to administer than protracted intravenous infusion of 5‐FU, with a pCR rate comparable to intravenous infusion of 5‐FU for LARC. In 2008, another phase II trial22 of 31 patients with LARC showed that capecitabine was well tolerated orally and had radiosensitizing effects comparable to those of neoadjuvant 5‐FU therapy. In 2011, Swellengrebel et al.30 again showed that oral capecitabine had an acceptable acute toxicity profile in a cohort of 147 patients with LARC. In 2015, Saha and co‐workers23 conducted a randomized control pilot study comparing capecitabine–oxaliplatin (CAPOX) as a radiosensitizer with 5‐FU–leucovorin; the two arms were comparable in terms of objective response rate, pCR rate, R0 resection and toxicity profile23. Noh and colleagues31, 32 investigated different timings for administration of capecitabine among 171 patients undergoing RT followed by total mesorectal excision (TME) 4–6 weeks after neoadjuvant therapy and assessed the radiosensitization response by pCR. The optimal radiosensitizing effects of capecitabine were achieved if it was administered 1 h before RT.

ferent timings for administration of capecitabine among 171 patients undergoing RT followed by total mesorectal excision (TME) 4–6 weeks after neoadjuvant therapy and assessed the radiosensitization response by pCR. The optimal radiosensitizing effects of capecitabine were achieved if it was administered 1 h before RT. A phase III RCT24 between 2002 and 2007 recruited nearly 400 patients with stage II and III LARC, comparing CRT with capecitabine versus 5‐FU. The primary endpoint was 5‐year overall survival, which in the capecitabine group was non‐inferior to that in the 5‐FU group (76 versus 67 per cent; P < 0·001). The local recurrence rate was similar in the two groups (6 versus 7 per cent); however, the rate of distant metastasis was 9 per cent lower in the capecitabine group, with increased 3‐year DFS. The phase III NSABP R‐04 trial25 aimed to ascertain the optimal neoadjuvant chemotherapy regimen alongside RT for stage II–III rectal cancer. Infusion of 5‐FU and oral capecitabine with or without intravenous oxaliplatin were compared in 1608 patients. The pCR rate was 17·8 per cent for those receiving 5‐FU and 20·7 per cent among those receiving capecitabine. Sphincter salvage rates were largely comparable between the groups at 59·4 and 59·3 per cent respectively, as were rates of tumour downstaging (21·3 versus 21·1 per cent). The addition of oxaliplatin led to a small increase in pCR rate, but a reduction in sphincter salvage and downstaging, and a significant increase in toxicity.

salvage rates were largely comparable between the groups at 59·4 and 59·3 per cent respectively, as were rates of tumour downstaging (21·3 versus 21·1 per cent). The addition of oxaliplatin led to a small increase in pCR rate, but a reduction in sphincter salvage and downstaging, and a significant increase in toxicity. Additional chemotherapy agents to enhance radiosensitivity Oxaliplatin Oxaliplatin is a third‐generation platinum‐based drug that enhances radiation‐induced cytotoxicity via irreparable DNA damage through formation of interstrand and intrastrand crosslinks, induction of G2/M cell‐cycle arrest, blockage of DNA replication and inhibition of transcription33, 34. Preclinical data indicated potent radiosensitizing properties of the drug, with synergism between oxaliplatin and RT35, 36; these findings have been applied to several clinical trials for patients with LARC. Phase I–II studies focusing on the addition of oxaliplatin to 5‐FU‐based neoadjuvant CRT reported promising results33. pCR rates varied between 7 and 28 per cent, compared with 8–15 per cent in the 5‐FU‐alone group34, 37. Single weekly dosing was the most effective regimen, with diarrhoea and neuropathy the most commonly reported adverse effects.

on the addition of oxaliplatin to 5‐FU‐based neoadjuvant CRT reported promising results33. pCR rates varied between 7 and 28 per cent, compared with 8–15 per cent in the 5‐FU‐alone group34, 37. Single weekly dosing was the most effective regimen, with diarrhoea and neuropathy the most commonly reported adverse effects. Six large phase III trials to date have compared fluoropyrimidine CRT with or without oxaliplatin. The results from these trials are summarized in Table 2. The ACCORD 12/prodige 2 trial38 randomized 598 patients to standard capecitabine‐based neoadjuvant CRT or additional weekly dosing with oxaliplatin together with an increased radiation dose. The difference in pCR rate of 13·9 versus 19·2 per cent was not significant (P = 0·09), although ACCORD had been powered to detect an increase from 11 to 20 per cent with CAPOX. There was an increase in grade 3–4 toxicity with oxaliplatin. Table 2 Summary of other chemotherapy agents Results (%)* Reference Phase Disease stage Test drug Single/combination Cohort size pCR rate cCR rate Other endpoints Toxicity 38 III T3–4 N–/+ Oxaliplatin Capecitabine + oxaliplatin versus capecitabine alone 598 19·2 (13·9) 0·7 (0) Positive CRM 9·9 (19·3) Grade 3–4 toxicity 25 (1) 39 III T3–4 N–/+ Oxaliplatin 5‐FU + oxaliplatin versus 5‐FU alone 1236 17 (13) – 3‐year DFS 75·9 (71·2) Grade 3–4 toxicity 23 (20) 40 III T3–4 N–/+ Oxaliplatin 5‐FU +/– oxaliplatin versus capecitabine +/– oxaliplatin 1608 – – LR 11·2 (11·8) 5‐year DFS 66·4 (67·7)

Reference Phase Disease stage Test drug Single/combination Cohort size pCR rate cCR rate Other endpoints Toxicity 38 III T3–4 N–/+ Oxaliplatin Capecitabine + oxaliplatin versus capecitabine alone 598 19·2 (13·9) 0·7 (0) Positive CRM 9·9 (19·3) Grade 3–4 toxicity 25 (1) 39 III T3–4 N–/+ Oxaliplatin 5‐FU + oxaliplatin versus 5‐FU alone 1236 17 (13) – 3‐year DFS 75·9 (71·2) Grade 3–4 toxicity 23 (20) 40 III T3–4 N–/+ Oxaliplatin 5‐FU +/– oxaliplatin versus capecitabine +/– oxaliplatin 1608 – – LR 11·2 (11·8) 5‐year DFS 66·4 (67·7) 5‐year OS 81·3 (79) Addition of oxaliplatin significantly increased toxicity (P < 0·001) 41 III T3–4 N–/+ Oxaliplatin Oxaliplatin +5‐FU versus 5‐FU alone 1094 – – 3‐year DFS 74·5 (73·9) ‐ 42 III T3–4 N–/+ Oxaliplatin 5‐FU + oxaliplatin versus 5‐FU alone 747 16 (16) – Positive CRM 7 (4) Grade 3–4 toxicity 24 (8) Discontinued owing to toxicity 17 (4) 43 III T3–4 N–/+ Oxaliplatin 5‐FU + oxaliplatin versus 5‐FU alone 495 27·5 (14·0) – Negative nodes 87·4 (80·1) Grade 3–4 haematological toxicity 19 (12·9) 44 I–II T3–4 N–/+ Irinotecan Irinotecan +5‐FU 59 25 – Downstaging 41 3‐year DFS 40 Grade 3–4 toxicity 28·8 45 II T3–4 N–/+ Irinotecan Irinotecan + capecitabine 36 15 – 3‐year OS 80 Grade 3–4 haematological toxicity 25 46 II T3–4 N–/+ Irinotecan Irinotecan + capecitabine 48 25 – 5‐year DFS 75 5‐year OS 94 Grade 3 toxicity 10·5 No grade 4 toxicity 47 II T3–4 N–/+ Irinotecan Irinotecan + 5‐FU versus 5‐FU alone 106 26 (30) Downstaging 75 (74) 5‐year OS 75 (61)

3‐year DFS 40 Grade 3–4 toxicity 28·8 45 II T3–4 N–/+ Irinotecan Irinotecan + capecitabine 36 15 – 3‐year OS 80 Grade 3–4 haematological toxicity 25 46 II T3–4 N–/+ Irinotecan Irinotecan + capecitabine 48 25 – 5‐year DFS 75 5‐year OS 94 Grade 3 toxicity 10·5 No grade 4 toxicity 47 II T3–4 N–/+ Irinotecan Irinotecan + 5‐FU versus 5‐FU alone 106 26 (30) Downstaging 75 (74) 5‐year OS 75 (61) 5‐year DFS 85 (78) Grade 3 toxicity 13 (8) 48 II T3–4 N–/+ Irinotecan Irinotecan + capecitabine 110 21·8 – Negative CRM 89·1 3‐year DFS 96·9 3‐year OS 88·2 Grade 3 GI toxicity 22 No grade 4 toxicity * Results for control group are shown in parentheses. pCR, pathological complete response; cCR, clinical complete response; CRM, circumferential resection margin; 5‐FU, 5‐fluorouracil; DFS, disease‐free survival; LR, local recurrence; OS, overall survival; GI, gastrointestinal. The CAO/ARO/AIO‐04 study39, which included over 1000 participants, was the only trial to find a significantly improved pCR rate with oxaliplatin (from 13 to 17 per cent; P = 0·038). This was also the only trial to report an advantage for oxaliplatin in terms of 3‐year DFS (71·2 to 75·9 per cent). There were no significant differences in grade 3–4 toxicities or postoperative complications. However, the infusional 5‐FU regimen was changed between the control and experimental arms, with oxaliplatin being added to 16 weeks of postoperative adjuvant chemotherapy compared with 16 weeks of bolus 5‐FU alone in the control arm; this means that the relative contributions of oxaliplatin in CRT compared with adjuvant chemotherapy are difficult to define.

changed between the control and experimental arms, with oxaliplatin being added to 16 weeks of postoperative adjuvant chemotherapy compared with 16 weeks of bolus 5‐FU alone in the control arm; this means that the relative contributions of oxaliplatin in CRT compared with adjuvant chemotherapy are difficult to define. The NSABP R‐0440 and PETACC‐641 trials, published only in abstract form, found that addition of oxaliplatin to 5‐FU‐based neoadjuvant therapy led to decreased treatment compliance and increased toxicity, with no associated improvement in pathological tumour downstaging. The STAR‐01 trial42 randomized 747 patients to standard 5‐FU chemotherapy or additional oxaliplatin. Interim analysis detected no difference in pCR and toxicity problems with the addition of oxaliplatin42. Interestingly, initial results from the Chinese FOWARC trial43 showed that use of a modified FOLFOX (oxaliplatin, leucovorin, 5‐FU) 6 regimen in addition to 5‐FU + RT gave significantly improved rates of pCR compared with single‐agent 5‐FU + RT (27·5 versus 14·0 per cent respectively). The trial also showed comparable downstaging and acceptable toxicity in patients with stage II–III disease, and good compliance. Long‐term data are awaited and may still be important for future practice. The evidence at present, including subsequent meta‐analyses49, 50, still supports the use of a single‐agent fluoropyrimidine as the standard of care because of a lack of consistent improvement in pCR and 3‐year DFS rates51 with the combined regimen, and the greater toxicity due to oxaliplatin52.

Interestingly, initial results from the Chinese FOWARC trial43 showed that use of a modified FOLFOX (oxaliplatin, leucovorin, 5‐FU) 6 regimen in addition to 5‐FU + RT gave significantly improved rates of pCR compared with single‐agent 5‐FU + RT (27·5 versus 14·0 per cent respectively). The trial also showed comparable downstaging and acceptable toxicity in patients with stage II–III disease, and good compliance. Long‐term data are awaited and may still be important for future practice. The evidence at present, including subsequent meta‐analyses49, 50, still supports the use of a single‐agent fluoropyrimidine as the standard of care because of a lack of consistent improvement in pCR and 3‐year DFS rates51 with the combined regimen, and the greater toxicity due to oxaliplatin52. Irinotecan Irinotecan, a topoisomerase (TOPO) 1 inhibitor, inhibits religation of single‐strand DNA breaks through the formation of camptothecin 11–TOPO‐1–DNA complexes53. A preclinical study54 has demonstrated irinotecan to be not only a feasible addition to 5‐FU chemotherapy, but also a potent radiosensitizing agent in colorectal cancer, even under hypoxic conditions.

bitor, inhibits religation of single‐strand DNA breaks through the formation of camptothecin 11–TOPO‐1–DNA complexes53. A preclinical study54 has demonstrated irinotecan to be not only a feasible addition to 5‐FU chemotherapy, but also a potent radiosensitizing agent in colorectal cancer, even under hypoxic conditions. In a small phase I trial in 2008, Choi et al.55 examined the addition of weekly irinotecan to traditional 5‐FU neoadjuvant CRT in 16 patients with locally advanced T3–T4 rectal cancers. Some 94 per cent of patients were eligible to progress with surgical resection at the time of restaging, with 93 per cent achieving a R0 resection. In eight patients the disease was downstaged based on the TMN classification, with a pCR rate of 25 per cent. Although the numbers were too small to draw any firm conclusions, the evidence was promising in terms of combination potential. Mehta and colleagues56 conducted a phase II trial with the same dosing strategy in a cohort of 32 patients, of whom 38 per cent achieved a pCR and 71 per cent TNM downstaging. However, 56 per cent experienced acute toxicity with an initial dose of 50 mg/m2, requiring dose alteration or delay in administration. Overall, small phase II studies44, 45, 46, 56 focusing on this approach have achieved pCR rates of 14–37 per cent and tumour downstaging in 24–71 per cent. A summary of these trials is provided in Table 2.

experienced acute toxicity with an initial dose of 50 mg/m2, requiring dose alteration or delay in administration. Overall, small phase II studies44, 45, 46, 56 focusing on this approach have achieved pCR rates of 14–37 per cent and tumour downstaging in 24–71 per cent. A summary of these trials is provided in Table 2. Mohiuddin et al.47 reported 5‐year outcomes on 106 patients randomized to either basic 5‐FU CRT or additional 4‐week doses of 50 mg/m2 irinotecan. They reported an increase in overall survival of 14 per cent with the addition of irinotecan; however, the locoregional recurrence rate was 17 per cent, compared with 16 per cent without irinotecan, and respective distal recurrence rates were 21 and 16 per cent. There was no significant difference between the treatment arms in terms of pCR or downstaging, but an increased rate of acute toxicity was reported in the irinotecan group. Gollins and colleagues48 reported on 110 patients with MRI‐defined locally advanced rectal cancer threatening or involving the surgical CRM treated with a regimen of irinotecan 60 mg/m2 weekly for the first 4 weeks of a 5‐week course of capecitabine CRT. In total, 24 patients (21·8 per cent) had a pCR and 98 (89·1 per cent) a negative CRM. A further study focusing on long‐term outcome in 115 patients57 found no significant difference between the two treatment arms in terms of pCR, with a higher overall survival rate of 87 per cent and DFS rate of 79 per cent in the irinotecan group (median follow‐up 60 months).

pCR and 98 (89·1 per cent) a negative CRM. A further study focusing on long‐term outcome in 115 patients57 found no significant difference between the two treatment arms in terms of pCR, with a higher overall survival rate of 87 per cent and DFS rate of 79 per cent in the irinotecan group (median follow‐up 60 months). Despite the promise of the above studies, no phase III trial of concurrent irinotecan has yet been reported58. This will be rectified in the future by the ongoing UK ARISTOTLE trial, which will complete accrual (target 600 patients) in mid‐2018. In MRI‐defined high‐risk rectal cancer, ARISTOTLE will compare CRT with concurrent capecitabine with or without irinotecan. Epidermal growth factor receptor inhibitors Epidermal growth factor receptor (EGFR), a member of the ErbB family of receptors, is relevant in colorectal cancer because overexpression or upregulation of the EGFR gene occurs in 60–80 per cent of cases59, 60, 61. Expression of the gene is also associated with poor survival62, 63, 64. The anti‐EGFR monoclonal antibodies cetuximab and panitumumab are already approved for the treatment of RAS wild‐type metastatic colorectal cancer65, but their role in LARC remains unclear.

the EGFR gene occurs in 60–80 per cent of cases59, 60, 61. Expression of the gene is also associated with poor survival62, 63, 64. The anti‐EGFR monoclonal antibodies cetuximab and panitumumab are already approved for the treatment of RAS wild‐type metastatic colorectal cancer65, but their role in LARC remains unclear. There have been several clinical trials of EGFR‐targeting monoclonal antibodies as radiosensitizers in neoadjuvant therapy for LARC. These trials are summarized in Table 3. Early efficacy results in terms of pCR rate were around 5–10 per cent74, 75, 76. However, these studies did not investigate tumour RAS status, which is used as a predictive biomarker for anti‐EGFR monoclonal antibody response in metastatic colorectal cancer77, 78. Potentially, optimal ordering of chemotherapy, RT and the EGFR inhibitor might unlock the full radiosensitizing potential of anti‐EGFR monoclonal antibodies79. Table 3 Summary of phase II trials of epidermal growth factor receptor inhibitors Results (%)* Reference Phase Disease stage Test drug Single/combination Cohort size pCR rate cCR rate Other endpoints Toxicity 66 II T3–4 N–/+ Panitumumab Single 19 0 – Downstaging 41 Negative CRM 76 LRC 90 DFS 79 GI disturbance 89 Grade 4 toxicity 21 67 II T3–4 N–/+ Cetuximab Cetuximab + capecitabine 31 0 – Downstaging 42 GI disturbance 13 Grade 4 toxicity 3 68 II T3–4 N–/+ Panitumumab Panitumumab + oxaliplatin +5‐FU 60 21 – Downstaging 58 GI disturbance 39 1 death 69 II T3–4 N–/+ Panitumumab Panitumumab + capecitabine versus capecitabine alone 68 10 (18) – R0 resection 85 (93) Sphincter salvage 69 (70)

Grade 4 toxicity 21 67 II T3–4 N–/+ Cetuximab Cetuximab + capecitabine 31 0 – Downstaging 42 GI disturbance 13 Grade 4 toxicity 3 68 II T3–4 N–/+ Panitumumab Panitumumab + oxaliplatin +5‐FU 60 21 – Downstaging 58 GI disturbance 39 1 death 69 II T3–4 N–/+ Panitumumab Panitumumab + capecitabine versus capecitabine alone 68 10 (18) – R0 resection 85 (93) Sphincter salvage 69 (70) Downstaging 87 (85) GI disturbance 10 (4) 70 II T3–4 Cetuximab Cetuximab + capecitabine + oxaliplatin versus capecitabine + oxaliplatin 165 11 (9) 11 (7) Radiological response 71 (51) GI disturbance 8 (9) 71 II T2–4 N–/+ Cetuximab Cetuximab + capecitabine + irinotecin 82 17 5 R0 resection 82 GI disturbance 25 Grade 4 toxicity 10 72 II T3–4 N–/+ Cetuximab Cetuximab + capecitabine 47 8 – 3‐year DFS 72 3‐year RFS 74 3‐year OS 68 2 of 32 unable to complete treatment owing to GI disturbance and leucopenia 73 I–II T3–4 N–/+ Cetuximab Cetuximab + capecitabine + oxaliplatin 60 8 – 5‐year OS 76 3‐year DFS 88 5 ‐year CSS 78 Grade 2 toxicity 5 * Results for control group are shown in parentheses. pCR, pathological complete response; cCR, clinical complete response; CRM, circumferential resection margin; LRC, locoregional control; DFS, disease‐free survival; GI, gastrointestinal; 5‐FU, 5‐fluorouracil; RFS, relapse‐free survival; OS, overall survival; CSS, cancer‐specific survival.

cal or pathological response. Using next‐generation sequencing, 46 per cent of matched biopsy–resection specimens were discrepant for EGFR pathway mutations. Intratumoral heterogeneity was suggested as a possible explanation, manifesting as a geographical biopsy miss or chemoradiation‐driven emergence of new mutations. Phase II studies so far have failed to suggest a benefit in terms of pCR rate and DFS, and have shown no consistent correlation with KRAS status72, 73, 80. There is currently no role for the addition of EGFR‐targeted therapy as a radiosensitizer in the treatment of LARC81. However, a pilot study82 of RT with personalized chemotherapy and biological therapy, based on molecular markers among 16 patients with T3 or N1 rectal cancers, showed a pCR rate of 50 per cent, which may be the basis for future molecular guided studies. Antiangiogenesis therapy Bevacizumab Bevacizumab is a monoclonal antibody that targets vascular epithelial growth factor (VEGF). In combination with cytotoxic chemotherapy, it has shown potential for rectal cancer treatment; the evidence is, however, currently limited to phase I–II trials83. Salazar et al.84 undertook a multicentre randomized phase II trial in 90 patients with LARC, of capecitabine with or without bevacizumab. The pCR rate was 16 per cent in the bevacizumab arm, compared with 11 per cent in the control arm, and an additional 20 per cent of tumours were downstaged. However, the predefined efficacy endpoint of a difference in treatment arms of 10 per cent was not met, despite these encouraging results.

or without bevacizumab. The pCR rate was 16 per cent in the bevacizumab arm, compared with 11 per cent in the control arm, and an additional 20 per cent of tumours were downstaged. However, the predefined efficacy endpoint of a difference in treatment arms of 10 per cent was not met, despite these encouraging results. Landry et al.85 performed a phase II trial of the addition of bevacizumab therapy with 5‐year follow‐up. Of 57 patients included in the data analysis, 17 per cent achieved a pCR, with an overall 5‐year survival rate of 80 per cent and relapse‐free survival rate of 81 per cent. The pCR endpoint of 30 per cent was not reached and, owing to substantial side‐effects (1 death was attributed to study therapy), the regimen was not considered worthy of further work by the authors.

cent achieved a pCR, with an overall 5‐year survival rate of 80 per cent and relapse‐free survival rate of 81 per cent. The pCR endpoint of 30 per cent was not reached and, owing to substantial side‐effects (1 death was attributed to study therapy), the regimen was not considered worthy of further work by the authors. Sorafenib Sorafenib is a multikinase inhibitor that blocks the receptor tyrosine kinase of VEGF, platelet‐derived growth factor and the RAF serine–threonine kinases along the RAF–mitogen‐activated protein kinase kinase–extracellular signal‐related kinase pathway. Jeong and colleagues86 assessed its potential as a radiosensitizer using three colorectal cell lines, and a xenograft animal model. They were able to demonstrate a scientific rationale for combination therapy, with enhanced radiosensitivity being shown in all three cell lines and the xenograft model, and delayed DNA damage repair caused by the radiation treatment. Van Moos et al.87 evaluated its effect in a cohort of 54 patients with KRAS‐mutated rectal tumours in combination with capecitabine‐based CRT. The pCR rate was 60 per cent, with downstaging in 82 per cent. A second phase I study88 also produced encouraging results, with a pCR of 36 per cent.

by the radiation treatment. Van Moos et al.87 evaluated its effect in a cohort of 54 patients with KRAS‐mutated rectal tumours in combination with capecitabine‐based CRT. The pCR rate was 60 per cent, with downstaging in 82 per cent. A second phase I study88 also produced encouraging results, with a pCR of 36 per cent. Poly(ADP‐ribose) polymerase inhibition PARPs, particularly PARP‐1, play a critical role in the recognition and repair of DNA single‐ and double‐strand breaks. Higher PARP activity has been noted in cancer cells with increased proliferation and chemoradioresistance, and this has led to the development of PARP inhibitors, which reduce the cancer cell's ability to repair single‐ and double‐strand breaks generated by RT and lead to cell death. Preclinical trials have demonstrated radiosensitizing effects in multiple colorectal cell lines89, although this effect is potentially largely dependent on the status of oncogenes such as BRCA1/BRCA2 90 generating defective DNA double‐strand break repair (termed synthetic lethality), and dysregulation of P5391.

death. Preclinical trials have demonstrated radiosensitizing effects in multiple colorectal cell lines89, although this effect is potentially largely dependent on the status of oncogenes such as BRCA1/BRCA2 90 generating defective DNA double‐strand break repair (termed synthetic lethality), and dysregulation of P5391. Veliparib (ABT‐888), a potent orally bioavailable PARP‐1/2 inhibitor, has been shown to enhance the antitumor activity of chemotherapy and RT in preclinical colorectal cancer models92. In in vitro and in vivo experiments in colorectal cancer, veliparib had independent radiosensitization effects and was synergistic with chemotherapy, especially with irinotecan. Final results from a phase Ib dose‐escalation study of veliparib plus capecitabine‐based CRT and surgery were published in 201793, demonstrating a pCR rate of 28 per cent. As with the EGFR monoclonal antibodies, this class of potential radiosensitizer remains an area of interest and future studies are needed to elucidate its role in rectal cancer. Potential predictive biomarkers have not been identified. A recent report94 has described pharmacodynamic assays that are able to measure the low levels at which PARP inhibitors are active.

ass of potential radiosensitizer remains an area of interest and future studies are needed to elucidate its role in rectal cancer. Potential predictive biomarkers have not been identified. A recent report94 has described pharmacodynamic assays that are able to measure the low levels at which PARP inhibitors are active. Immunotherapy The immune system plays an intricate and complex role in all aspects of cancer from carcinogenesis to treatment95. Over the past 10 years, a great deal of work has been done to better understand that role, with the development of therapies such as programmed cell death protein 1 (PD‐1)/programmed death ligand 1 (PD‐L1) inhibitors, cancer vaccines and adoptive cell therapy. In a phase I–II trial96 of adjuvant immunotherapy involving sentinel lymph node T lymphocytes in 55 patients with metastatic colorectal cancer, there was no treatment‐related toxicity, and 24‐month survival rates were 56 and 18 per cent in the treatment and control groups respectively.

ccines and adoptive cell therapy. In a phase I–II trial96 of adjuvant immunotherapy involving sentinel lymph node T lymphocytes in 55 patients with metastatic colorectal cancer, there was no treatment‐related toxicity, and 24‐month survival rates were 56 and 18 per cent in the treatment and control groups respectively. Use of cytokine therapy, although very early in terms of research into colorectal cancer, has been approved by the US Food and Drug Administration for melanoma and renal cell carcinoma (interleukin 2). Although much of the work focusing on colorectal malignancy is in its early phases (I–II), there is evidence to suggest the potential use of these therapies in a combination role for a correctly selected cohort97. Specifically, the PD‐1 immune checkpoint inhibitors pemrolizumab and nivolumab have shown promising activity in DNA mismatch repair‐deficient (dMMR)/microsatellite instability – high colorectal cancers, which carry a high mutation load and an active immune microenvironment98, 99. A small proportion of rectal cancers are dMMR, but one possible area of research is to determine whether the proinflammatory properties of RT might enhance the response of microsatellite‐stable tumours to PD‐1 blockade. The R‐IMMUNE phase II study100 is currently recruiting to compare the use of atezolizumab as a radiosensitizer with 5‐FU‐based neoadjuvant CRT. More studies are in the pipeline, with a recent UK proposal aiming to assess the effect of the PD‐L1 inhibitor durvalumab in combination with RT.

able tumours to PD‐1 blockade. The R‐IMMUNE phase II study100 is currently recruiting to compare the use of atezolizumab as a radiosensitizer with 5‐FU‐based neoadjuvant CRT. More studies are in the pipeline, with a recent UK proposal aiming to assess the effect of the PD‐L1 inhibitor durvalumab in combination with RT. Novel agents With a clear focus of research on optimizing neoadjuvant therapy, several novel agents ranging from cyclo‐oxygenase 2 inhibitors to nanoparticles have been investigated in the preclinical setting (Table 4). Further phase I studies are in preparation to examine both prostaglandin E2 receptor inhibitors (PRAER 1 trial) and Ad3/Ad11p chimeric adenoviruses (CEDAR trial). Table 4 Summary of novel radiosensitizing agents Reference Study design Findings COX‐2 inhibitors Cox‐2 is an inducible enzyme that regulates prostaglandin synthesis and is overexpressed at sites of inflammation and in epithelial malignancy tumours101. It is involved in the regulation of apoptosis, angiogenesis and tumour cell invasiveness. Preclinical studies suggest the potential of COX‐2 inhibitors as selective radiosensitizers102 Debucquoy et al.103 Double‐blind randomized phase II; in addition to 5‐FU; 35patients Improved downstaging No increased toxicity Nanoparticles Aim to improve the therapeutic index of chemoradiotherapy and overcome potential systemic excess toxicity. Focus on particle size sub‐50 nm Caster et al.104 Particles 50, 100 and 150 nm in size loaded with 2 DNA repair inhibitor model drugs in colorectal cancer cell lines All sizes potent radiosensitizers

Nanoparticles Aim to improve the therapeutic index of chemoradiotherapy and overcome potential systemic excess toxicity. Focus on particle size sub‐50 nm Caster et al.104 Particles 50, 100 and 150 nm in size loaded with 2 DNA repair inhibitor model drugs in colorectal cancer cell lines All sizes potent radiosensitizers Good toxicity tolerance Tian et al.105 CRLX101 in combination with oxaliplatin and 5‐FU Increased efficacy of chemoradiotherapy Early stage; needs expansion Histone deacetylase inhibitors Emerging therapeutic concept attempting to target epigenetic regulatory mechanisms and act as a radiosensitizer in combination therapy. SAHA approved as a single agent for refractory cutaneous T‐cell lymphoma Folkvord et al.106 Preclinical study of SAHA using 2 xenograft models In vitro: improved radiosensitivity (P ≤ 0·050) across cell lines at all radiation doses less than 6 h after exposure In vivo: pCR achieved in 1 model Saelen et al.107 Vorinostat assessed under hypoxic conditions in vitro Enhanced radiosensitivity across cell lines Warrants further research Small molecular inhibitors Low molecular weight; able to target both extracellular and intracellular proteins Kleiman et al 108 Preclinical Focus on radiosensitizers for KRAS mutant tumours 28 known radiosensitizers assessed 6 effective; AZD7762 most highly potent Suggested investigation into role of CHK2 inhibitors Nelfinavir HIV protease inhibitor; inhibits Akt at standard clinical doses and results in radiosensitivity Hill et al.109

Warrants further research Small molecular inhibitors Low molecular weight; able to target both extracellular and intracellular proteins Kleiman et al 108 Preclinical Focus on radiosensitizers for KRAS mutant tumours 28 known radiosensitizers assessed 6 effective; AZD7762 most highly potent Suggested investigation into role of CHK2 inhibitors Nelfinavir HIV protease inhibitor; inhibits Akt at standard clinical doses and results in radiosensitivity Hill et al.109 Non‐randomized SONATINA clinical trial focusing on safety in 10 patients with T3–4 N0–2 M1 rectal cancers recruited over 2 years 14 days total oral treatment (7 days preoperative) 2 discontinued owing to toxicity 5 grade 3 toxicity Warrants further research Buijsen et al.110 Phase I trial including 12 patients Escalating doses with capecitabine Primary endpoint: dose‐limiting toxicity 4 of 6 experienced toxicity, precluding further dose escalation pCR 27% Further toxicity concerns Zerumbone Cyclic sesquiterpene from rhizomes of edible ginger plant; emerging evidence of potential for inhibition of proliferation of human colonic adenocarcinoma cells, with minimal toxicity111 Deorukhkar et al.112 3 colorectal cancer cell lines Inhibition of proliferation identified in dose‐dependentmanner Marked radiosensitizer in clonogenic survival curves Little effect on normal fibroblasts

Further toxicity concerns Zerumbone Cyclic sesquiterpene from rhizomes of edible ginger plant; emerging evidence of potential for inhibition of proliferation of human colonic adenocarcinoma cells, with minimal toxicity111 Deorukhkar et al.112 3 colorectal cancer cell lines Inhibition of proliferation identified in dose‐dependentmanner Marked radiosensitizer in clonogenic survival curves Little effect on normal fibroblasts Warrants further research Bortezomib Modified dipeptidyl boronic acid derived from leucine and phenylalanine that acts as a 26S proteasome inhibitor. The ubiquitin–proteasome pathway is involved in intracellular protein degradation in eukaryotic cells O'Neil et al.113 10 patients with stage II or III rectal cancer received 5‐FU‐based chemoradiotherapy plus bortezomib twiceper week pCR 10% High toxicity – diarrhoea Study not progressed COX, cyclo‐oxygenase; 5‐FU, 5‐fluorouracil; SAHA, suberoylanilide hydroxamic acid; pCR, pathological complete response; CHK2, serine–threonine kinase 2; HIV, human immunodeficiency virus. Alternatives to standard radiotherapy strategies Dose escalation An alternative potential method of enhancing the effectiveness of CRT is by increasing the radiation dose, via an increased external‐beam dose or endocavitary brachytherapy. There is evidence to suggest that a dose–response relationship with pCR exists114.

Alternatives to standard radiotherapy strategies Dose escalation An alternative potential method of enhancing the effectiveness of CRT is by increasing the radiation dose, via an increased external‐beam dose or endocavitary brachytherapy. There is evidence to suggest that a dose–response relationship with pCR exists114. A prospective single‐centre study115 from Denmark in patients with T2–3 cancers within 6 cm of the anal verge used radiation dose intensification to the primary tumour delivered with intensity‐modulated external‐beam RT to 60 Gy in 30 fractions over 6 weeks, with 50 Gy to the pelvic nodes, combined with an endorectal brachytherapy tumour boost to 5 Gy and tegafur/uracil on treatment days. Of the 51 patients treated, 78 per cent achieved a cCR and organ preservation; the local recurrence rate was 26 per cent at 2 years. Grade 3 diarrhoea occurred in 8 per cent, and long‐term rectal bleeding was of concern during follow‐up. Gerard and colleagues116 demonstrated improved clinical (24 versus 2 per cent) and pathological (57 versus 34 per cent) responses using the 50‐Kv Papillon technique for contact X‐ray brachytherapy (CXB). Patients with a clinical incomplete response to external‐beam CRT have been shown to achieve a cCR after a CXB boost, with only 11 per cent developing recurrence117.

inical (24 versus 2 per cent) and pathological (57 versus 34 per cent) responses using the 50‐Kv Papillon technique for contact X‐ray brachytherapy (CXB). Patients with a clinical incomplete response to external‐beam CRT have been shown to achieve a cCR after a CXB boost, with only 11 per cent developing recurrence117. With both approaches, there is a lack of randomized data. The recently funded UK APHRODITE study will randomize patients with T1–T3b rectal adenocarcinomas with a maximum diameter of 4 cm, considered unsuitable for radical TME surgery, to standard CRT versus RT dose‐escalated CRT. The OPERA trial will randomize patients with early cT2–T3a–b tumours smaller than 5 cm in diameter, treated with external‐beam CRT, to either an external‐beam CRT boost or a CXB boost.

arcinomas with a maximum diameter of 4 cm, considered unsuitable for radical TME surgery, to standard CRT versus RT dose‐escalated CRT. The OPERA trial will randomize patients with early cT2–T3a–b tumours smaller than 5 cm in diameter, treated with external‐beam CRT, to either an external‐beam CRT boost or a CXB boost. Delivery modification An alternative strategy to dose escalation is the development of novel delivery methods that reduce toxicity, particularly to the small bowel. Intensity‐modulated RT is one such technique that has been proposed owing to its highly conformal dose distribution. There are currently few published prospective data to support its routine use; however, a recent meta‐analysis118 of retrospective studies has suggested that it has a significantly lower toxicity profile than routine three‐dimensional CRT. Future developments may ultimately lead to traditional photon irradiation being replaced with charged particles such as protons or carbon ions, which may have even greater biological effectiveness while maintaining a favourable toxicity profile. At the present time, further clinical studies and access to treatment facilities are required to assess the applicability of these techniques fully119, 120.

placed with charged particles such as protons or carbon ions, which may have even greater biological effectiveness while maintaining a favourable toxicity profile. At the present time, further clinical studies and access to treatment facilities are required to assess the applicability of these techniques fully119, 120. Preoperative chemotherapy given sequentially with (chemo)radiotherapy The twofold rationale for giving neoadjuvant chemotherapy sequentially, either before or after (C)RT, followed by surgery, is to improve the response of the primary tumour and to reduce the distant metastasis rate. Owing to morbidity from RT and pelvic surgery, individuals who have undergone preoperative CRT then surgery may fail to start adjuvant chemotherapy or tolerate it poorly, resulting in dose reductions121. A meta‐analysis122 of four trials including preoperative RT, however, has questioned the benefit of postoperative chemotherapy (hazard ratio for DFS 0·91, 95 per cent c.i. 0·77 to 1·07; P = 0·230), possibly for this reason. Giving chemotherapy before surgery allows an increased dose intensity to be delivered, potentially increasing the response rate.

s including preoperative RT, however, has questioned the benefit of postoperative chemotherapy (hazard ratio for DFS 0·91, 95 per cent c.i. 0·77 to 1·07; P = 0·230), possibly for this reason. Giving chemotherapy before surgery allows an increased dose intensity to be delivered, potentially increasing the response rate. However, although the concept of ‘total neoadjuvant therapy’ is gaining traction123, there is currently very little randomized phase II (and no phase III) evidence specifically examining the benefit of neoadjuvant chemotherapy. Grupo Cancer de Recto (GCR) 3124 was a randomized phase II study of preoperative CAPOX followed by CRT then surgery versus CRT then surgery then postoperative CAPOX in 108 patients. Less toxicity (P < 0·001) and better compliance (P < 0·001) were demonstrated for the same regimen used as neoadjuvant chemotherapy compared with adjuvant chemotherapy, although the pCR rate was no different (13 versus 14 per cent respectively). A non‐randomized US–Canadian trial125 examined four sequential study groups of patients with LARC, examining CRT followed by chemotherapy then surgery. Group 1 had CRT followed by TME 6–8 weeks later. Groups 2, 3 and 4 had two, four and six 2‐weekly cycles of modified FOLFOX delivered between CRT and TME. The pCR rate was 18, 25, 30 and 38 per cent for groups 1–4 respectively. Although promising, it is not clear whether the increased downstaging occurred because of a greater gap between CRT and surgery (6, 8, 12 and 16 weeks for groups 1–4 respectively).

2‐weekly cycles of modified FOLFOX delivered between CRT and TME. The pCR rate was 18, 25, 30 and 38 per cent for groups 1–4 respectively. Although promising, it is not clear whether the increased downstaging occurred because of a greater gap between CRT and surgery (6, 8, 12 and 16 weeks for groups 1–4 respectively). Randomized studies are urgently needed to examine the efficacy of intensified neoadjuvant CRT regimens for rectal cancer, including the sequential addition of preoperative chemotherapy, in comparison to standard neoadjuvant CRT alone. Neoadjuvant chemotherapy alone In the modern TME era, local recurrence rates have fallen to as low as 5 per cent. However, CRT has not affected distant metastatic relapse, which affects up to 30 per cent of patients. Although surgery is associated with long‐term sexual, bowel and bladder dysfunction, preoperative RT can exacerbate this morbidity126. Consideration should be given to whether chemotherapy alone can be as effective as CRT in terms of DFS, thereby avoiding some acute and long‐term toxicity.

p to 30 per cent of patients. Although surgery is associated with long‐term sexual, bowel and bladder dysfunction, preoperative RT can exacerbate this morbidity126. Consideration should be given to whether chemotherapy alone can be as effective as CRT in terms of DFS, thereby avoiding some acute and long‐term toxicity. Several small single‐arm studies using mainly oxaliplatin‐based chemotherapy have reported promising DFS rates. In addition, studies127, 128 examining neoadjuvant CAPOX followed by CRT have clearly shown the substantial downstaging efficacy of chemotherapy, using MRI after chemotherapy but before CRT. The FOWARC Chinese phase III study43 randomized 495 patients with LARC to either standard neoadjuvant CRT using concurrent 5‐FU, CRT with concurrent 5‐FU and oxaliplatin, or FOLFOX chemotherapy alone. Although tumour downstaging was comparable between the standard CRT and chemotherapy‐alone arms (37·1 and 35·5 per cent respectively), the pCR rate was inferior with chemotherapy alone (14·0 versus 6·6 per cent). It was reported recently that there was no difference in DFS or overall survival between the three arms129. At present, there is more evidence to support the replacement of neoadjuvant CRT with chemotherapy using DFS as the primary endpoint, than for a cCR/organ preservation endpoint.

alone (14·0 versus 6·6 per cent). It was reported recently that there was no difference in DFS or overall survival between the three arms129. At present, there is more evidence to support the replacement of neoadjuvant CRT with chemotherapy using DFS as the primary endpoint, than for a cCR/organ preservation endpoint. Discussion The ideal radiosensitizing agent would be one that could target cancer cells selectively130, 131, enhancing the efficacy of treatment with minimal local and systemic toxicity. Exploiting the benefits of neoadjuvant therapy, accurately staging and assessing cCR could open up the era of increasingly personalized medicine and the avoidance of resection altogether132. It is an area of research that could bring significant patient benefits including improvements in health‐related quality of life. However, future clinical trials of radiosensitizers, with the aim of organ preservation, need carefully to consider the endpoints that are used to assess efficacy132.

esection altogether132. It is an area of research that could bring significant patient benefits including improvements in health‐related quality of life. However, future clinical trials of radiosensitizers, with the aim of organ preservation, need carefully to consider the endpoints that are used to assess efficacy132. Although there is much emerging evidence with regard to potential new radiosensitizing agents, the current standard treatment alongside RT remains 5‐FU or capecitabine chemotherapy. The addition of any second systemic agent has yet to show a consistent increase in efficacy in randomized studies. Many promising radiosensitizers have failed to progress beyond the preclinical and early clinical phases (I–II) owing to systemic toxicity and varying rates of pCR. Unfortunately, the quality of phase II studies of potential intensifying agents has been poor. A systematic review133 of 92 phase II trials showed that only eight were randomized.

itizers have failed to progress beyond the preclinical and early clinical phases (I–II) owing to systemic toxicity and varying rates of pCR. Unfortunately, the quality of phase II studies of potential intensifying agents has been poor. A systematic review133 of 92 phase II trials showed that only eight were randomized. There remains the fundamental question of the optimal primary endpoint. In virtually all studies in this review, the pCR rate was employed as the determinant of success. At present, there is no predetermined set definition of what constitutes a pCR. It may be defined as the absence of neoplastic cells in the surgical resection specimen as a result of neoadjuvant treatment (ypT0 and ypN0)134 and indeed may still occur even in the presence of mucosal abnormalities following treatment135. Published rates of pCR range widely from 15 to 40 per cent136, 137. However, despite small cohort sizes being accounted for, very few trials have noted the potential introduction of bias due to lack of standardization of pathologist reporting. The Royal College of Pathologists138 specifies that pathologists should embed all of the tumour‐associated scar and examine three deeper levels on each block before calling a pCR. The lack of reliable lymph node involvement status, dependence on pathologist block and level sampling intensity, and varying time points between the end of RT and surgery affecting tumour regression, could potentially lead to inflated pCR results. The associated benefits of a true pCR include a reduced recurrence rate and enhanced overall survival136, 139, 140.

status, dependence on pathologist block and level sampling intensity, and varying time points between the end of RT and surgery affecting tumour regression, could potentially lead to inflated pCR results. The associated benefits of a true pCR include a reduced recurrence rate and enhanced overall survival136, 139, 140. Tumour regression grading is a semiquantitative assessment of residual tumour cells versus fibroinflammatory tissue in the rectal wall, and has been shown to be able to stratify tumour response to CRT and predict prognosis on an individual‐patient level in two large prospective phase III trials141, 142. Identifying patients who have achieved a cCR following CRT, and who could be followed prospectively with an active surveillance or watch‐and‐wait strategy, is gathering increasing interest. Of 183 patients with T2–T4 N0–2 M0 distal rectal cancers receiving neoadjuvant CRT in a trial published in 2014 by Habr‐Gama and colleagues143, 49 per cent were deemed to have achieved a cCR; 31 per cent of these patients went on to develop local recurrence and the salvage rate was 93 per cent. The rate of local disease control was 94 per cent, with 78 per cent organ preservation. The Habr‐Gama protocol involved clinical, endoscopic, radiological and serological reassessment of patients 8 weeks after completion of neoadjuvant therapy. A cCR was defined as the absence of residual ulceration, stenosis or mass lesion within the rectum on digital palpation and endoscopic imaging. MRI was performed, and the carcinoembryonic antigen level was measured.

, radiological and serological reassessment of patients 8 weeks after completion of neoadjuvant therapy. A cCR was defined as the absence of residual ulceration, stenosis or mass lesion within the rectum on digital palpation and endoscopic imaging. MRI was performed, and the carcinoembryonic antigen level was measured. The International Watch & Wait Database Consortium144 recently published the long‐term outcomes of the largest series of patients managed by this strategy, reporting a 2‐year cumulative regrowth rate of 25·2 per cent among 1009 patients. Surgical treatment data were available for only 148 of the 213 patients who experienced regrowth; 115 proceeded to TME resection, with histologically clear margins in 88 per cent. Overall 5‐year survival rates of 84·7 per cent in this group are comparable to those of major resection. However, before this approach can be established as a standard of care, standardized definitions of cCR and surveillance protocols need to be developed. Criteria for shared decision‐making with the patient for this approach also need to be addressed. An increasing consensus views a two‐stage assessment as optimum for identifying a cCR, at 3 and then 6 months following CRT, allowing enough time for a cCR to develop in initially good responders145. MRI tumour regression grade following CRT has been shown to be predictive of DFS in a cohort of 66 patients from the MERCURY study, suggesting the value of MRI assessment after CRT as part of the protocol for selecting patients for a non‐operative approach146. Although there are still many questions surrounding watch and wait147, it clearly has an increasingly important place in modern rectal cancer management and strategies to intensify CRT need further exploration.

e of MRI assessment after CRT as part of the protocol for selecting patients for a non‐operative approach146. Although there are still many questions surrounding watch and wait147, it clearly has an increasingly important place in modern rectal cancer management and strategies to intensify CRT need further exploration. For patients in whom there has been a good response but not an apparent cCR, local transanal excision is an alternative to major resection148, 149. MRI can be useful in guiding patient selection for such treatments150. However, the recently published GRECCAR 2 study151, which used a composite endpoint of surgical complications and recurrence, failed to show a difference between the two approaches in this setting, suggesting that more prospective studies are needed in this area.

ful in guiding patient selection for such treatments150. However, the recently published GRECCAR 2 study151, which used a composite endpoint of surgical complications and recurrence, failed to show a difference between the two approaches in this setting, suggesting that more prospective studies are needed in this area. It is imperative that studies employ standardized pathological reporting to ensure that the pCR rates quoted are both realistic and comparable. In view of this, use of the pCR as a primary endpoint for research studies and/or clinical trials should perhaps be questioned. If the ultimate goal is organ preservation regardless of whether the patient has undergone a cCR or pCR, perhaps organ preservation should be the primary endpoint. Against this is the morbidity associated with rectal surgery in terms of bowel, urinary and sexual dysfunction. However, there are few data on long‐term toxicity and health‐related quality of life for an active surveillance approach, which clearly needs to be addressed in future prospective studies. A recent patient consultation exercise revealed that, even in the context of cancer care, patients regarded quality of life and presence of a stoma as more important than overall survival152. Future trials of neoadjuvant therapy for rectal cancer need to ensure that patient experience and reported endpoints are addressed.

ecent patient consultation exercise revealed that, even in the context of cancer care, patients regarded quality of life and presence of a stoma as more important than overall survival152. Future trials of neoadjuvant therapy for rectal cancer need to ensure that patient experience and reported endpoints are addressed. Although not available at the present time, it is hoped that the development of biomarker‐based stratified treatment will be used to guide neoadjuvant therapy on a personalized basis in the future153. Such biomarkers may be purely molecular (DNA alterations, gene expression, protein expression, epigenetic or circulating) or a combination of molecular markers and imaging findings154. Reliable pretreatment biomarkers do not currently exist, although ongoing research is attempting to identify pretreatment markers that are predictive of response155, 156. It is essential that future neoadjuvant trials incorporate a translational element to further develop biomarker‐guided therapy. As such, it is critical that such translational arms adhere to a robust biopsy protocol to ensure that enough appropriate biological material is available for downstream analysis in addition to the routine histopathological biopsies taken for diagnostic purposes. Factors that need to be considered include the person taking the biopsies, the quantity of material and timing. Patients who undergo a cCR or pCR will have little or no tumour to access at either clinical follow‐up or at the time of surgical resection; this must be considered in the translational design, which may need to include liquid biopsies157.

red include the person taking the biopsies, the quantity of material and timing. Patients who undergo a cCR or pCR will have little or no tumour to access at either clinical follow‐up or at the time of surgical resection; this must be considered in the translational design, which may need to include liquid biopsies157. Acknowledgements The authors acknowledge Yorkshire Cancer Research for funding the academic programmes at North Wales Cancer Treatment Centre and the University of Leeds. Disclosure: The authors declare no conflict of interest.

Introduction The volume of surgical procedures performed worldwide is large1 and, although many advances have been made in the past several decades, surgical care exposes patients to substantial risk of morbidity and mortality. The safety of surgical care has gained traction within the global health landscape, yet it remains a pressing concern in both resource‐rich and ‐poor settings2. In 2009, the WHO released a 19‐item surgical safety checklist for implementation in countries around the world3. The checklist was designed to promote understanding and cohesive communication, and to ensure good practice among all surgical team members at three specific intervals: before induction of anaesthesia, before skin incision, and before the patient leaves the operating theatre3. Although the improvement in outcomes was dramatic4, uptake of this safety tool remains to be quantified on a large scale. Attention surrounding use of the checklist in resource‐limited settings is of particular relevance to surgical care and outcomes4.

kin incision, and before the patient leaves the operating theatre3. Although the improvement in outcomes was dramatic4, uptake of this safety tool remains to be quantified on a large scale. Attention surrounding use of the checklist in resource‐limited settings is of particular relevance to surgical care and outcomes4. Process‐related discrepancies, such as lack of a particular safe practice protocol, are chief contributors to adverse surgical events5. Despite the potential benefits of checklist use, there are numerous barriers to implementation6, 7. Dynamic educational and social factors, such as ambiguity and confusion around the purpose of the checklist and negative attitudes to checklist adoption among some team members, contribute to poor checklist uptake8, 9. In addition, in settings where resources are limited, completion of a checklist that focuses on unavailable items can seem pointless, for instance in the absence of pulse oximetry or antibiotics10, 11. Despite these barriers to implementation and completion, it has been suggested5 that absence of the checklist itself may serve as a major contributor to adverse surgical events.

etion of a checklist that focuses on unavailable items can seem pointless, for instance in the absence of pulse oximetry or antibiotics10, 11. Despite these barriers to implementation and completion, it has been suggested5 that absence of the checklist itself may serve as a major contributor to adverse surgical events. Although data supporting the effectiveness of the checklist in fostering improved surgical outcomes are encouraging, studies in globally representative populations are uncommon. Furthermore, checklist outcomes have been studied largely within elective general surgery and subspecialty settings, with only a few studies examining checklist use in emergency care12, 13. Attitudes to checklist use by providers working in emergency settings can be negative12. A survey of obstetric care providers found that one‐third believed a checklist would be an inconvenience in emergencies14. Despite this, the benefit of checklist use does extend to emergency surgical care, as shown in an analysis of the original WHO checklist study in urgent operations across eight countries13. Importantly, it is also possible that the benefit of the checklist may be greatest in emergency situations, given the increased risks15. Using a large, validated, global data set, this study aimed to compare reported use of the WHO Surgical Safety Checklist in patients undergoing emergency laparotomy and elective gastrointestinal surgery. Associations were sought between checklist use and perioperative mortality, accounting for country developmental level as well as patient and disease factors.

this study aimed to compare reported use of the WHO Surgical Safety Checklist in patients undergoing emergency laparotomy and elective gastrointestinal surgery. Associations were sought between checklist use and perioperative mortality, accounting for country developmental level as well as patient and disease factors. Methods Study design and participants Data were collected prospectively within two international, multicentre, observational cohort studies: GlobalSurg 116 and GlobalSurg 217. Both studies were performed by the GlobalSurg Collaborative group using prespecified published protocols (NCT0217911218, NCT0266223119). This collaborative methodology has been described elsewhere20. A UK National Health Service Research Ethics review considered both GlobalSurg 1 and GlobalSurg 2 exempt from formal research registration (South East Scotland Research Ethics Service, references NR/1404AB12 and NR/1510AB5). Individual centres obtained their own audit, ethical or institutional approval. Details of data validation have been described and published previously16, 17. Results of this analysis are reported according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines21.

2 and NR/1510AB5). Individual centres obtained their own audit, ethical or institutional approval. Details of data validation have been described and published previously16, 17. Results of this analysis are reported according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines21. In both contributing studies, investigators from healthcare facilities worldwide that fulfilled the inclusion criteria were invited to participate and submit data using an online platform22. Small teams of investigators were recruited via email, social media and through personal contacts. Investigators collected data during at least one 2‐week interval during study windows. In GlobalSurg 1, investigators identified consecutive patients between 1 July 2014 and 21 December 201416. This study included any patient undergoing emergency intraperitoneal surgery, defined in the study methods16. In GlobalSurg 2, investigators included consecutive patients between 4 January 2016 and 31 July 201617. This study included all patients undergoing elective or emergency gastrointestinal surgery17. In both studies, patients were followed to day 30 after surgery, or for the duration of their inpatient stay in locations where follow‐up was not feasible. Local investigators uploaded records to a secure online database, using the Research Electronic Data Capture (REDCap) system23.

mergency gastrointestinal surgery17. In both studies, patients were followed to day 30 after surgery, or for the duration of their inpatient stay in locations where follow‐up was not feasible. Local investigators uploaded records to a secure online database, using the Research Electronic Data Capture (REDCap) system23. Data from both GlobalSurg 1 and GlobalSurg 2 were combined to create a multicentre data set. Variables were cross‐referenced and streamlined for coding consistency. Patients were then selected as having undergone emergency laparotomy or elective gastrointestinal surgery. Emergency laparotomy was captured using the definition from the UK National Emergency Laparotomy Audit24, adapted for global settings. Trauma laparotomy was included, whereas laparoscopic procedures and individuals aged less than 18 years were not included in the analysis.

parotomy or elective gastrointestinal surgery. Emergency laparotomy was captured using the definition from the UK National Emergency Laparotomy Audit24, adapted for global settings. Trauma laparotomy was included, whereas laparoscopic procedures and individuals aged less than 18 years were not included in the analysis. Variables The primary outcome measure was 30‐day perioperative mortality, expressed as a proportion. Perioperative mortality was defined as ‘any death, regardless of cause, occurring within 30 days of surgery in or out of the hospital’25. The metric was calculated by dividing the number of perioperative deaths by the total number of included operations performed26. The primary explanatory variable was reported use of the WHO Surgical Safety Checklist. Checklist use was recorded as ‘no, not available’, ‘no, but available’, ‘yes’, or ‘unknown’ for each patient in the study. Reported use of the checklist was calculated as a proportion recorded as ‘yes’ of the total number of patients included. Countries were stratified into three tertiles according to the Human Development Index (HDI) rank27. This is a composite statistic of life expectancy, education and income indices published by the United Nations. HDI was chosen over purely economic measures of country development on the principle that ‘people and their capabilities should be the ultimate criteria for assessing the development of a country, not economic growth alone’27.

e statistic of life expectancy, education and income indices published by the United Nations. HDI was chosen over purely economic measures of country development on the principle that ‘people and their capabilities should be the ultimate criteria for assessing the development of a country, not economic growth alone’27. Patient‐level variables included age, sex, diabetes, smoking status and ASA fitness grade. For simplification, ASA was grouped as a score of less than III and one of III or more. Disease‐level variables included six major diagnostic groups: abdominal wall, benign foregut, benign midgut/hindgut, malignancy, trauma/injury and other. Operative characteristics, including the requirement for a bowel resection and the level of contamination, were also included. Power considerations GlobalSurg 1 and GlobalSurg 2 both included a priori sample size calculations that accounted for the uncertainty in the data coming from collaborating countries18, 19. In making between‐group comparisons of checklist use by HDI country or urgency, a difference from 10 to 5 per cent can be shown with α = 5 per cent and β = 10 per cent (90 per cent power) with group sizes of 582.

size calculations that accounted for the uncertainty in the data coming from collaborating countries18, 19. In making between‐group comparisons of checklist use by HDI country or urgency, a difference from 10 to 5 per cent can be shown with α = 5 per cent and β = 10 per cent (90 per cent power) with group sizes of 582. Statistical analysis Differences between HDI tertiles were tested with Pearson's χ2 test and Kruskal–Wallis test for categorical and continuous variables respectively. Multivariable logistic regression models were used to adjust for confounding in analyses of checklist use and 30‐day perioperative mortality. Coefficients are expressed as odds ratios (ORs) with 95 per cent confidence intervals and P values derived from percentiles of 10 000 bootstrap replications. Models were constructed using the following principles: variables associated with outcome measures in previous studies were accounted for; demographic variables were included in model exploration; all first‐order interactions were checked and included in final models if found to be influential; final model selection was informed using a criterion‐based approach minimizing the Akaike information criterion and discrimination determined using the c‐statistic (area under the receiver operating characteristic (ROC) curve). Hierarchical models accounting for clustering within countries were not used in this analysis owing to colinearity between explanatory variables of interest and country. Model residuals were checked (residuals versus fitted values; normality plot of standardized deviance residuals; and residuals versus leverage), and goodness of model fit determined using the Hosmer–Lemeshow test.

were not used in this analysis owing to colinearity between explanatory variables of interest and country. Model residuals were checked (residuals versus fitted values; normality plot of standardized deviance residuals; and residuals versus leverage), and goodness of model fit determined using the Hosmer–Lemeshow test. To help translate model outputs to real‐life quantities of interest, bootstrapped simulations of model predictions were performed at specified co‐variable levels. This enables the results of regression analyses to be expressed as probabilities, with the intention of these being more intuitive to interpret than ORs. All analyses were undertaken using the R Foundation statistical program R 3.4.14, with finalfit28 and dplyr packages (R Project for Statistical Computing, Vienna, Austria). Data sharing The data set can be explored using an online visualization application at http://data.globalsurg.org.

To help translate model outputs to real‐life quantities of interest, bootstrapped simulations of model predictions were performed at specified co‐variable levels. This enables the results of regression analyses to be expressed as probabilities, with the intention of these being more intuitive to interpret than ORs. All analyses were undertaken using the R Foundation statistical program R 3.4.14, with finalfit28 and dplyr packages (R Project for Statistical Computing, Vienna, Austria). Data sharing The data set can be explored using an online visualization application at http://data.globalsurg.org. Results A total of 26 228 patient records were sourced from the GlobalSurg 1 (10 745, 41·0 per cent) and GlobalSurg 2 (15 483, 59·0 per cent) data sets (Fig. 1). For 552 patients (2·1 per cent) 30‐day mortality and/or checklist use was not known; these were removed. Patterns of missing data were examined and not considered to influence final results (Tables S1 and S2, supporting information). For 13 380 patients (51·0 per cent) the inclusion criteria were not fulfilled, and these were removed; reasons included undergoing an emergency procedure other than emergency laparotomy and age less than 18 years. The included procedures by country HDI are shown in Table S3 (supporting information). The final data set represents 12 296 patients from 76 countries, with 326 centres from GlobalSurg 116 and 356 centres from GlobalSurg 217. Figure 1 Flow chart of study population. HDI, Human Development Index

Results A total of 26 228 patient records were sourced from the GlobalSurg 1 (10 745, 41·0 per cent) and GlobalSurg 2 (15 483, 59·0 per cent) data sets (Fig. 1). For 552 patients (2·1 per cent) 30‐day mortality and/or checklist use was not known; these were removed. Patterns of missing data were examined and not considered to influence final results (Tables S1 and S2, supporting information). For 13 380 patients (51·0 per cent) the inclusion criteria were not fulfilled, and these were removed; reasons included undergoing an emergency procedure other than emergency laparotomy and age less than 18 years. The included procedures by country HDI are shown in Table S3 (supporting information). The final data set represents 12 296 patients from 76 countries, with 326 centres from GlobalSurg 116 and 356 centres from GlobalSurg 217. Figure 1 Flow chart of study population. HDI, Human Development Index BJS-11051-FIG-0001-cDemographics Extensive data relating to patient and operative characteristics by HDI group and surgery type are provided in Tables S4–S6 (supporting information) to allow a full understanding of the population. Patients were distributed across HDI groups as shown in Fig. 1. Patients undergoing emergency laparotomy were older, more likely to be men, had a higher ASA score, were less likely to have cancer, and had a higher rate of wound contamination (Table S4, supporting information). Those undergoing emergency laparotomy in low‐HDI compared with high‐HDI countries were younger, more likely to be men, had a lower ASA score, were less likely to have had a bowel resection or cancer, and more likely to have wound contamination (Table S5, supporting information). Similar differences were seen in the elective surgery group (Table S6, supporting information).

ed with high‐HDI countries were younger, more likely to be men, had a lower ASA score, were less likely to have had a bowel resection or cancer, and more likely to have wound contamination (Table S5, supporting information). Similar differences were seen in the elective surgery group (Table S6, supporting information). Use of WHO Surgical Safety Checklist Reported WHO Surgical Safety Checklist use across the entire cohort was 8881 of 12 296 (72·2 per cent). There was little difference in checklist use overall when comparing emergency laparotomy with elective surgery (73·7 versus 71·2 per cent respectively). Checklist use differed in high‐ (84·5 per cent) compared with middle‐ (59·4 per cent) and low‐ (47·3 per cent) HDI country groups (Table 1). Table 1 Patient and operative characteristics by reported WHO Surgical Safety Checklist use Checklist used No Yes Total P † Urgency 0·003 Emergency laparotomy 1272 (26·3) 3571 (73·7) 4843 Elective surgery 2143 (28·8) 5310 (71·2) 7453 HDI tertile < 0·001 High 1095 (15·5) 5953 (84·5) 7048 Middle 1496 (40·6) 2189 (59·4) 3685 Low 824 (52·7) 739 (47·3) 1563 Age (years)* 47·7(19·2) 53·8(19·7) – < 0·001‡

Table 1 Patient and operative characteristics by reported WHO Surgical Safety Checklist use Checklist used No Yes Total P † Urgency 0·003 Emergency laparotomy 1272 (26·3) 3571 (73·7) 4843 Elective surgery 2143 (28·8) 5310 (71·2) 7453 HDI tertile < 0·001 High 1095 (15·5) 5953 (84·5) 7048 Middle 1496 (40·6) 2189 (59·4) 3685 Low 824 (52·7) 739 (47·3) 1563 Age (years)* 47·7(19·2) 53·8(19·7) – < 0·001‡ Sex 0·729 M 1612 (28·2) 4095 (71·8) 5707 F 1682 (28·0) 4334 (72·0) 6016 Missing 121 (21·1) 452 (78·9) 573 ASA fitness grade < 0·001 < III 2598 (30·2) 6006 (69·8) 8604 ≥ III 743 (21·2) 2766 (78·8) 3509 Missing 74 (40·4) 109 (59·6) 183 Smoker 0·009 No 2594 (29·2) 6302 (70·8) 8896 Yes 670 (26·5) 1858 (73·5) 2528 Missing 151 (17·3) 721 (82·7) 872 Diabetes < 0·001 No 3076 (28·4) 7744 (71·6) 10 820 Yes 339 (23·0) 1137 (77·0) 1476 Disease classification < 0·001 Abdominal wall 174 (29·8) 409 (70·2) 583 Other 1567 (32·2) 3305 (67·8) 4872 Benign foregut 478 (27·9) 1235 (72·1) 1713 Benign midgut/hindgut 410 (20·8) 1558 (79·2) 1968 Malignancy 569 (22·2) 1994 (77·8) 2563 Trauma/injury 190 (37·3) 320 (62·7) 510 Missing 27 (31) 60 (69) 87 Bowel resection < 0·001 No 1879 (25·7) 5435 (74·3) 7314 Yes 1535 (30·8) 3442 (69·2) 4977 Missing 1 (20) 4 (80) 5 Malignancy < 0·001 No 2846 (29·2) 6887 (70·8) 9733 Yes 569 (22·2) 1994 (77·8) 2563 Contamination 0·061 Clean/contaminated 2564 (27·3) 6840 (72·7) 9404 Contaminated 324 (27·1) 873 (72·9) 1197 Dirty 476 (30·1) 1106 (69·9) 1582 Missing 51 (45·1) 62 (54·9) 113 Values in parentheses are percentages unless indicated otherwise;

lignancy < 0·001 No 2846 (29·2) 6887 (70·8) 9733 Yes 569 (22·2) 1994 (77·8) 2563 Contamination 0·061 Clean/contaminated 2564 (27·3) 6840 (72·7) 9404 Contaminated 324 (27·1) 873 (72·9) 1197 Dirty 476 (30·1) 1106 (69·9) 1582 Missing 51 (45·1) 62 (54·9) 113 Values in parentheses are percentages unless indicated otherwise; * values are mean(s.d.). HDI, Human Development Index. † χ2 test, except ‡ Kruskal–Wallis test (comparisons of available data). A multivariable regression model was used to adjust for confounding and identify predictors of checklist use (Table S7, supporting information). A significant interaction was found between type of surgery and country HDI for checklist use: the patterns of checklist use by surgery type were different across HDI groups. After adjusting for patient and operative characteristics, checklist use continued to be lower in low‐ and middle‐HDI countries (Fig. 2 a). To convey differences in checklist use more intuitively, bootstrapped predicted probabilities of checklist use were determined (Table 2). Absolute risk differences for emergency laparotomy versus elective surgery differed by HDI group. Checklist use was less common for elective surgery than emergency laparotomy in the high‐HDI group (absolute risk difference −9·4 (95 per cent c.i. −11·9 to −6·9) per cent; P < 0·001), no different in the middle‐HDI group (1·9 (−2·3 to 6·5) per cent; P = 0·357) and more common in the low‐HDI group (12·1 (7·0 to 17·3) per cent; P < 0·001).

ss common for elective surgery than emergency laparotomy in the high‐HDI group (absolute risk difference −9·4 (95 per cent c.i. −11·9 to −6·9) per cent; P < 0·001), no different in the middle‐HDI group (1·9 (−2·3 to 6·5) per cent; P = 0·357) and more common in the low‐HDI group (12·1 (7·0 to 17·3) per cent; P < 0·001). Figure 2 Odds ratio plots of WHO Surgical Safety Checklist use and 30‐day mortality. a Use of WHO checklist and b 30‐day mortality for surgery type and Human Development Index (HDI) group from multivariable logistic regression models. Odds ratios are shown with 95 per cent confidence intervals and P values. Checklist use was adjusted for age, ASA score, diabetes status, disease classification, bowel resection and wound contamination. For full models, see Tables S7 and S8 (supporting information). Mortality (b) is adjusted for WHO surgical safety checklist use, age, ASA, disease classification, bowel resection and wound contamination BJS-11051-FIG-0002-cTable 2 WHO Surgical Safety Checklist reported use by country Human Development Index and type of surgery Checklist used* No Yes P§ Adjusted probability of checklist use (%)† Absolute risk difference (%)† P

Figure 2 Odds ratio plots of WHO Surgical Safety Checklist use and 30‐day mortality. a Use of WHO checklist and b 30‐day mortality for surgery type and Human Development Index (HDI) group from multivariable logistic regression models. Odds ratios are shown with 95 per cent confidence intervals and P values. Checklist use was adjusted for age, ASA score, diabetes status, disease classification, bowel resection and wound contamination. For full models, see Tables S7 and S8 (supporting information). Mortality (b) is adjusted for WHO surgical safety checklist use, age, ASA, disease classification, bowel resection and wound contamination BJS-11051-FIG-0002-cTable 2 WHO Surgical Safety Checklist reported use by country Human Development Index and type of surgery Checklist used* No Yes P§ Adjusted probability of checklist use (%)† Absolute risk difference (%)† P High HDI Emergency laparotomy 286 (10·4) 2455 (89·6) 86·0 (83·4, 88·3) – Elective surgery 809 (18·8) 3498 (81·2) < 0·001 76·6 (73·8, 79·3) −9·4 (–11·9, −6·9) < 0·001 Middle HDI Emergency laparotomy 489 (39·4) 753 (60·6) 51·6 (46·4, 56·7) – Elective surgery 1007 (41·2) 1436 (58·8) 0·280 53·6 (49·4, 57·5) 1·9 (–2·3, 6·5) 0·357 Low HDI Emergency laparotomy 497 (57·8) 363 (42·2) 33·3 (28·4, 38·4) – Elective surgery 327 (46·5) 376 (53·5) < 0·001 45·4 (40·2, 50·9) 12·1 (7·0, 17·3) < 0·001 Values in parentheses are *percentages and †95 per cent confidence intervals. A multivariable logistic regression model was specified for checklist use by Human Development Index (HDI) group, type of surgery and confounders (see Fig. 2 and Table S4, supporting information). Bootstrapped adjusted predictions of the probability of checklist use were performed for different HDI groups and surgery type, with other co‐variable levels specified: age 52 years; ASA grade less than III; no diabetes; malignancy disease classification; clean/contaminated wound status. Absolute risk differences for the probability of checklist use were determined and a two‐sided P value was calculated. §χ2 test (within HDI group between surgery type and checklist).

levels specified: age 52 years; ASA grade less than III; no diabetes; malignancy disease classification; clean/contaminated wound status. Absolute risk differences for the probability of checklist use were determined and a two‐sided P value was calculated. §χ2 test (within HDI group between surgery type and checklist). Mortality in emergency laparotomy Overall, 30‐day mortality after emergency laparotomy (621 of 4843, 12·8 per cent) was ten‐times higher than for elective surgery (94 of 7453, 1·3 per cent) (Table S8, supporting information). There was notable variation in mortality after elective surgery by HDI group, but less variation after emergency laparotomy in the unadjusted analysis. However, after adjustment for confounding, significant differences were seen in 30‐day mortality after emergency laparotomy in low‐ (OR 2·43, 95 per cent c.i. 1·81 to 3·25; P < 0·001) and middle‐ (2·80, 2·20 to 3·56; P < 0·001) HDI groups compared with the high‐HDI group (Fig. 2 b; Table S8, supporting information). Thirty‐day mortality after elective surgery in low‐HDI countries was equivalent to 30‐day mortality after emergency laparotomy in high‐HDI countries.

r cent c.i. 1·81 to 3·25; P < 0·001) and middle‐ (2·80, 2·20 to 3·56; P < 0·001) HDI groups compared with the high‐HDI group (Fig. 2 b; Table S8, supporting information). Thirty‐day mortality after elective surgery in low‐HDI countries was equivalent to 30‐day mortality after emergency laparotomy in high‐HDI countries. Use of checklist and 30‐day mortality Overall, reported use of the WHO Surgical Safety Checklist was associated with a lower 30‐day mortality (471 of 8881, 5·3 per cent) compared with reported non‐use (244 of 3415, 7·1 per cent) (Table S8, supporting information). In models adjusting for confounders, reported use of the checklist was still associated with a significantly lower 30‐day mortality (OR 0·60, 95 per cent c.i. 0·50 to 0·73; P < 0·001). Again, to create a more intuitive interpretation of the mortality model, adjusted predicted probabilities of 30‐day mortality were created to allow comparisons across HDI group, surgery type and reported checklist use (Fig. 3; Table S9, supporting information). No interaction between checklist use and 30‐day mortality was seen for HDI or type of surgery. Thus, as expected, a significant checklist effect was seen for each combination of surgery type and HDI group, with magnitudes of effect shown in Fig. 3. The greatest absolute risk difference for checklist use was seen in emergency surgery in low‐HDI (absolute risk reduction 4·6 (95 per cent c.i. 2·7 to 7·0) per cent; P < 0·001) and middle‐HDI (5·1 (2·9 to 7·8) per cent; P < 0·001) countries, by virtue of the higher baseline 30‐day mortality (Table S9, supporting information).

atest absolute risk difference for checklist use was seen in emergency surgery in low‐HDI (absolute risk reduction 4·6 (95 per cent c.i. 2·7 to 7·0) per cent; P < 0·001) and middle‐HDI (5·1 (2·9 to 7·8) per cent; P < 0·001) countries, by virtue of the higher baseline 30‐day mortality (Table S9, supporting information). Figure 3 Adjusted probability of 30‐day mortality by surgery type, Human Development Index group and WHO Surgical Safety Checklist use. a Emergency laparotomy; b elective surgery. The multivariable logistic regression model for 30‐day mortality (Fig. 2; Table S8, supporting information) was used to generate adjusted predicted probabilities of death using bootstrap replication, with other co‐variable levels specified: age 52 years, ASA grade less than III, malignancy disease classification, and contamination. Absolute risk differences for 30‐day mortality are presented with 95 per cent confidence intervals, and two‐sided P values for the absolute risk difference (Table S9, supporting information). HDI, Human Development Index

ified: age 52 years, ASA grade less than III, malignancy disease classification, and contamination. Absolute risk differences for 30‐day mortality are presented with 95 per cent confidence intervals, and two‐sided P values for the absolute risk difference (Table S9, supporting information). HDI, Human Development Index BJS-11051-FIG-0003-cDiscussion In this large multinational prospective cohort study, reported use of the WHO Surgical Safety Checklist was associated with a significant reduction in 30‐day perioperative mortality. This relationship was consistent and independent of key patient and disease‐related variables. Checklist use in low‐HDI countries was half that of high‐HDI countries, and this effect persisted after accounting for differences in patient and disease characteristics. Checklist use was lower for elective surgery than for emergency laparotomy in high‐HDI countries; this finding was unexpected. The association between checklist use and lower mortality was consistent across HDI groups and type of surgery, even after adjustment for case mix. The greatest absolute benefits were found in emergency surgery in low‐ and middle‐HDI countries, owing to the higher baseline mortality rate.

; this finding was unexpected. The association between checklist use and lower mortality was consistent across HDI groups and type of surgery, even after adjustment for case mix. The greatest absolute benefits were found in emergency surgery in low‐ and middle‐HDI countries, owing to the higher baseline mortality rate. Evidence supporting use of the surgical safety checklist in hospital practice is widely positive and supports the promotion of the checklist in patient safety programmes worldwide2. A strength of this study is the breadth of countries and hospitals that contributed prospectively collected data. Use of the WHO Surgical Safety Checklist in low‐ and middle‐income countries was 2928 of 5248 (55·8 per cent), the same as that reported in the recent African Surgical Outcomes Study (6183 of 10 836, 57·1 per cent)29. In the present study, checklist use was found to be significantly lower in some low‐ and middle‐HDI countries compared with high‐HDI countries, yet the association with lower mortality was still seen. This clearly highlights an area for practice change. Fostering awareness to motivate local checklist champions is important, but may be difficult in more remote environments. In regions with less established local organizational infrastructure, the checklist may simply have not come to the attention of providers. Supportive governmental and academic institutions are crucial in facilitating the process, through continued professional development and ministry of health‐accredited programmes30. Successful implementation of the checklist requires careful thought and local adaptation. Avoiding the perception of the checklist as just a ‘tickbox’ exercise is crucial for success9.

ons are crucial in facilitating the process, through continued professional development and ministry of health‐accredited programmes30. Successful implementation of the checklist requires careful thought and local adaptation. Avoiding the perception of the checklist as just a ‘tickbox’ exercise is crucial for success9. This study explored the hypothesis that checklist use is lower in emergency settings due to the particular challenges therein. Focus on emergency laparotomy provided a more homogeneous study group, rather than including all patients undergoing emergency surgery. Checklist use has been studied extensively in elective surgery, but has received less attention in emergency surgery, particularly in global settings13. Narratives that highlight time pressures, low staffing, inflexible hierarchies and lack of resources during times of emergency promote an idea that the checklist should not be prioritized in urgent care. The present study shows that the checklist can be used in emergency settings, and commonly is. Moreover, associations between checklist use and better outcomes are just as evident in emergency laparotomy as in elective surgery. The capacity for checklist implementation in settings providing emergency surgical care should not be undervalued. On the contrary, there may be much to gain for emergency surgery in low‐ and middle‐income settings, given the higher baseline mortality rate.

ust as evident in emergency laparotomy as in elective surgery. The capacity for checklist implementation in settings providing emergency surgical care should not be undervalued. On the contrary, there may be much to gain for emergency surgery in low‐ and middle‐income settings, given the higher baseline mortality rate. A number of weaknesses to the approach taken here may be discussed. In general, the methodology is subject to selection bias at hospital level. Collaborators self‐select to take part, which may reflect better resourced institutions contributing than is seen in the general population. During the study period, consecutive patients must be recruited so that selection bias at patient level is minimized. Data were validated particularly carefully in the GlobalSurg 2 study17. As described previously, patient recruitment together with a subset of variables were re‐collected by an independent team in contributing hospitals. In centres with the fewest resources, this can be a challenge, particularly where there are no formal written records. Some collaborators described the GlobalSurg data as of better quality than what was otherwise available within their hospital.

iables were re‐collected by an independent team in contributing hospitals. In centres with the fewest resources, this can be a challenge, particularly where there are no formal written records. Some collaborators described the GlobalSurg data as of better quality than what was otherwise available within their hospital. With regard to checklist use itself, collaborators simply reported whether a WHO Surgical Safety Checklist had been used before surgery. No review was undertaken of what form the checklist took, what local adaptations may have been introduced, whether there was broad team acceptance of the process, or whether the implementation was deemed successful. Self‐reported checklist use is a poor reflection of meaningful compliance with checklist items31, and partial compliance reduces the positive benefits on outcomes derived from this32. Furthermore, the use and appropriate implementation of a surgical safety checklist may represent a wide range of health service system characteristics, including organizational and management attributes that support good clinical practices. Facilities reporting high use may also be those with significant resources, staffed by individuals familiar with key patient safety concepts who work together to ensure reliable systems are in place to deliver consistent patient care.

luding organizational and management attributes that support good clinical practices. Facilities reporting high use may also be those with significant resources, staffed by individuals familiar with key patient safety concepts who work together to ensure reliable systems are in place to deliver consistent patient care. This study shows that the WHO Surgical Safety Checklist can be used in emergency surgery in resource‐poor settings. The association with lower mortality is likely to reflect broader health system differences that prioritize safe and effective surgical care, yet the checklist plays an important part. Much of the benefit is likely to come from behaviours that can be difficult to measure, such as improved communication, better team work, identification of potential problems before they occur, and empowerment of members of staff at all levels. The checklist likely helps improve surgical safety by providing a framework for focusing teams on specific critical safety standards that are frequently assumed to have occurred but may not be adhered to. It can also help identify specific lapses and process weaknesses that can be the focus of improvement efforts. Where standards are either not known or not clear, the checklist can raise awareness of them and help guide hospital policies and protocols. It can even create a ‘team‐generated Hawthorne effect’, whereby all perioperative personnel are involved in the responsibility of ensuring compliance with standards, and observe completion together using the checklist.

not clear, the checklist can raise awareness of them and help guide hospital policies and protocols. It can even create a ‘team‐generated Hawthorne effect’, whereby all perioperative personnel are involved in the responsibility of ensuring compliance with standards, and observe completion together using the checklist. The data reported here have important implications for policy‐makers. Ten years after the introduction of the checklist, there is much work to be done in promoting its adoption worldwide. Local adaptation and ownership are clearly important in ensuring long‐term sustainable change33. Further studies around the details of implementation in resource‐constrained settings will help tailor checklist procedures to local needs, thereby ensuring greatest effect. Strong compliance and effective implementation are challenging, but have the potential to save many lives and should be a priority for surgical safety. Supporting information Table S1 Pattern of missing data for 30‐day mortality by other patient and operative characteristics. Data are n (%) unless otherwise stated. P‐values are for comparisons across non‐missing data and are chi‐squared or Kruskal‐Wallis (*) tests. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation.

y by other patient and operative characteristics. Data are n (%) unless otherwise stated. P‐values are for comparisons across non‐missing data and are chi‐squared or Kruskal‐Wallis (*) tests. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Table S2 Pattern of missing data for reported WHO Surgical Safety Checklist use by other patient and operative characteristics. Data are n (%) unless otherwise stated. P‐values are for comparisons across non‐missing data and are chi‐squared or Kruskal‐Wallis (*) tests. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Table S3 Procedures by human development index group Table S4 Patient and operative characteristics by type of surgery. Data are n (%) unless otherwise stated. P‐values are for comparisons across non‐missing data and are chi‐squared or Kruskal‐Wallis (*) tests. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Table S5 Patient and operative characteristics by country human development index for emergency laparotomy. Data are n (%) unless otherwise stated. P‐values are for comparisons across non‐missing data and are chi‐squared or Kruskal‐Wallis (*) tests. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation.

human development index for emergency laparotomy. Data are n (%) unless otherwise stated. P‐values are for comparisons across non‐missing data and are chi‐squared or Kruskal‐Wallis (*) tests. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Table S6 Patient and operative characteristics by country human development index for elective surgery. Data are n (%) unless otherwise stated. P‐values are for comparisons across non‐missing data and are chi‐squared or Kruskal‐Wallis (*) tests. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Table S7 Univariable and multivariable logistic regression analysis of reported WHO Surgical Safety Checklist use. Data are n (%) unless otherwise stated. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Number in data frame = 12296, Number in model = 11911, Missing = 385, AIC = 12598·8, C‐statistic = 0·717, H&L = Chi‐sq(8) 67·60 (p<0·001). Table S8 Univariable and multivariable logistic regression analysis of 30‐day mortality. Data are n (%) unless otherwise stated. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Number in data frame = 12296, Number in model = 11916, Missing = 380, AIC = 3947·7, C‐statistic = 0·87, H&L = Chi‐sq(8) 14·28 (p=0·075).

day mortality. Data are n (%) unless otherwise stated. HDI, human development index; ASA, American Society of Anesthesiologists Physical Classification System; SD, standard deviation. Number in data frame = 12296, Number in model = 11916, Missing = 380, AIC = 3947·7, C‐statistic = 0·87, H&L = Chi‐sq(8) 14·28 (p=0·075). Table S9 Probability of 30‐day mortality by surgery type, human development index group and WHO Surgical Safety Checklist use. The multivariable logistic regression model for 30‐day mortality (Figure 2 and supplementary table 5) was used to generate adjusted predicted probabilities of death using bootstrap replication. Appendix S1 Members of the GlobalSurg Collaborative Click here for additional data file. Acknowledgements Organizations that assisted in dissemination and/or translation (alphabetical): Asian Medical Students' Association; Association of Surgeons in Training; College of Surgeons of East, Central and Southern Africa; Cutting Edge Manipal; Egyptian Medical Student Research Association; Ghana College of Physicians and Surgeons; International Collaboration for Essential Surgery; International Federation of Medical Student Associations; Italian Society of Colorectal Surgery; Lifebox Foundation; School of Surgery, Student Audit and Research in Surgery; The Electives Network; United Kingdom National Research Collaborative; West African College of Surgeons; World Society of Emergency Surgery; World Surgical Association.

n of Medical Student Associations; Italian Society of Colorectal Surgery; Lifebox Foundation; School of Surgery, Student Audit and Research in Surgery; The Electives Network; United Kingdom National Research Collaborative; West African College of Surgeons; World Society of Emergency Surgery; World Surgical Association. This work was funded by DFID/MRC/Wellcome Trust Joint Global Health Trial Development Grant (MR/N022114/1). A National Institute for Health Research (NIHR) Global Health Research Unit Grant (NIHR 17‐0799) is supporting the establishment of surgical research units in a subset of contributing low‐income countries. The views expressed are those of the authors and not necessarily those of the National Health Service, NIHR or UK Department of Health and Social Care. The study funder had no role in study design, data collection, data analysis, data interpretation or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Disclosure: The authors declare no conflict of interest.

Introduction The concept of ‘big data’ describes the use of unstructured digital information, usually from multiple sources, that is often collected with no clearly defined purpose for future use1. The volume of data already being produced is vast, with frequent increases in complexity, variety and speed2. Big data in surgery can be defined as the amalgamation and integration of various data sources along the patient pathway to produce a rich matched data set3 (Fig. 1). Figure 1 Conceptualizing big data in healthcare. Health system data are aggregated with data generated by the individual and their environment. Data are transformed and analysed to generate actionable output BJS-11052-FIG-0001-cThe analysis and translation of big data to maximize quality and improve patient care is a priority for healthcare systems4. It is envisaged that measurement and modelling of patient health states and outcomes will quickly become the biggest driver of best practice and healthcare policy5. Continual analysis of patient‐level outcomes has already been demonstrated to significantly reduce morbidity and mortality in high‐income countries6.

ems4. It is envisaged that measurement and modelling of patient health states and outcomes will quickly become the biggest driver of best practice and healthcare policy5. Continual analysis of patient‐level outcomes has already been demonstrated to significantly reduce morbidity and mortality in high‐income countries6. However, discussions around large‐volume patient data frequently place little emphasis on their application in low‐ and middle‐income countries (LMICs), despite the potential for vast gains in patient outcomes and surgical service quality7. Currently, LMICs may lack the ability to gather reliable data6, with an expectation that this situation is unlikely to change in the near future8, 9. Ensuring that LMICs can keep up to date with technological advances will help to prevent future global health inequalities worsening10. The aim of this review was to evaluate the current applications of large‐volume patient‐level data in surgery in LMICs, together with highlighting where further focus is required to improve outcomes, define quality indicators and achieve universally available safe surgery. Methods An electronic systematic search of the PubMed, Embase and Google Scholar databases was performed in accordance with the PRISMA guidelines11, involving all published literature up to the last search on 23 August 2018. The PROSPERO international systematic review registry12 was searched to ensure a similar review had not been performed previously and the protocol was registered accordingly (CRD42018108203).

erformed in accordance with the PRISMA guidelines11, involving all published literature up to the last search on 23 August 2018. The PROSPERO international systematic review registry12 was searched to ensure a similar review had not been performed previously and the protocol was registered accordingly (CRD42018108203). A search of Embase and PubMed was undertaken using the keywords ‘surgery or surg*’, ‘big data’, ‘large data’, ‘informatics’, ‘database’, ‘cohort’ and ‘registry’, combined with LMIC filters as specified by the Cochrane library13. Search terms are listed in Appendix S1 (supporting information). A further supplementary search of Google Scholar was also undertaken. Search limits applied were English language, full text, humans and articles published from 2008 onwards to provide contemporary studies that were likely reflective of current approaches to data capture. The inclusion criteria were: prospectively collected data (or retrospective analysis of such data) on patients undergoing surgery with care being provided, at least in part, in a LMIC, defined according to the World Bank classification14. Studies were excluded if they contained fewer than 100 patients or were RCTs. Conference abstracts were screened to assist in identifying related full‐text articles. Where more than one article was published from a single data set, the article analysing the largest cohort of patients was included.

ank classification14. Studies were excluded if they contained fewer than 100 patients or were RCTs. Conference abstracts were screened to assist in identifying related full‐text articles. Where more than one article was published from a single data set, the article analysing the largest cohort of patients was included. Following the literature search, article titles were screened by four investigators and those meeting the inclusion criteria were screened further by abstract and then full text as appropriate. Any disagreements were resolved by consensus within the group. Bibliographies from included articles were hand‐searched to identify any further relevant articles. Data were extracted independently using a standardized pro forma, including year of publication, countries involved in the study, number of patients for each LMIC, patient‐level data type (cohort, database or registry), surgical specialty and measured outcome(s). In multinational studies where the number of patients for individual countries was not reported, the number of patients in the study was recorded. These studies were excluded from analysis mapping of the global distribution of patients across included studies to avoid data skewing. Individual LMICs where there were fewer than 100 patients in multinational studies were also excluded from analysis mapping. However, studies that did not report patient numbers for individual LMICs and all data from multinational studies were included in all other analyses.

ross included studies to avoid data skewing. Individual LMICs where there were fewer than 100 patients in multinational studies were also excluded from analysis mapping. However, studies that did not report patient numbers for individual LMICs and all data from multinational studies were included in all other analyses. Data types were defined as follows: cohort – collection of patient‐level data over a defined short period; database – concerted and long‐term collection of patient‐level data of consecutive patients over a small geographical area; or registry – studies meeting database classification but performed over a wide geographical area (such as national registries). Definitions were discussed and consensus reached within the group where doubt existed regarding particular studies. Owing to the narrative nature of the review, a qualitative analysis was performed using the R statistical program (https://www.R‐project.org/) and the tidyverse package15. All analyses and graphical representation of the data can be found at https://argoshare.is.ed.ac.uk/bigdata_review.

ed regarding particular studies. Owing to the narrative nature of the review, a qualitative analysis was performed using the R statistical program (https://www.R‐project.org/) and the tidyverse package15. All analyses and graphical representation of the data can be found at https://argoshare.is.ed.ac.uk/bigdata_review. Results The literature search identified 3805 articles, of which 218 full texts were assessed for eligibility (Fig. 2). Following assessment, 68 articles16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, involving 708 032 patients across 71 LMICs, were included in the review (Tables S1 and S2, supporting information). Country‐specific patient numbers were reported in 60 studies but were absent from six50, 51, 52, 57, 62, 83 and two33, 55 provided total LMIC patient numbers only. Figure 2 PRISMA flow chart showing selection of studies for review. *Two of these studies provided total LMIC patient number. LMIC, low‐ and middle‐income country

Results The literature search identified 3805 articles, of which 218 full texts were assessed for eligibility (Fig. 2). Following assessment, 68 articles16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, involving 708 032 patients across 71 LMICs, were included in the review (Tables S1 and S2, supporting information). Country‐specific patient numbers were reported in 60 studies but were absent from six50, 51, 52, 57, 62, 83 and two33, 55 provided total LMIC patient numbers only. Figure 2 PRISMA flow chart showing selection of studies for review. *Two of these studies provided total LMIC patient number. LMIC, low‐ and middle‐income country BJS-11052-FIG-0002-cPatients and studies Studies using big data were well represented across the 10‐year analysis period; however, a dramatic increase in study and patient numbers was seen from 2015 onwards (Fig. 3 a). Relatively few studies were found for the interval 2012–2014 despite no decrease in the total number of studies returned in the initial literature search (339, 358 and 469 studies in 2012, 2013 and 2014 respectively, compared with a median of 222 (range 129–487) for other years). Figure 3 Patient numbers over time in included studies. a Patient numbers in each year; b cumulative count by year

BJS-11052-FIG-0002-cPatients and studies Studies using big data were well represented across the 10‐year analysis period; however, a dramatic increase in study and patient numbers was seen from 2015 onwards (Fig. 3 a). Relatively few studies were found for the interval 2012–2014 despite no decrease in the total number of studies returned in the initial literature search (339, 358 and 469 studies in 2012, 2013 and 2014 respectively, compared with a median of 222 (range 129–487) for other years). Figure 3 Patient numbers over time in included studies. a Patient numbers in each year; b cumulative count by year BJS-11052-FIG-0003-cThe number of patients in the included studies ranged from 335 to 428 346, with a median of 2483 per study. Over 3000 patients were included in 25 of 68 studies; the biggest studies were published in the interval 2015–2018. Studies based on database and registry data were most common and represented 43 of 68 included studies. The majority of data sets identified arose from prospective cohorts of patients. Several of these studies were performed in single centres43, 65 or single nations79, 81, with comparisons made with high‐income countries. The largest cohort of patients originated from the DATASUS registry in Brazil (428 346 patients), which explored outcomes after hysterectomy80. Five multinational observational cohort studies50, 51, 57, 82, 83 were performed in the past 5 years, with the majority conducted over 7 days.

made with high‐income countries. The largest cohort of patients originated from the DATASUS registry in Brazil (428 346 patients), which explored outcomes after hysterectomy80. Five multinational observational cohort studies50, 51, 57, 82, 83 were performed in the past 5 years, with the majority conducted over 7 days. Geographical distribution The studies had a wide geographical LMIC distribution. The majority, however, were from Brazil (12), China (11), India (5) and Thailand (4) (Fig. 4 a). Patient‐level data were collected from 71 LMICs in total; overall, patient representation was particularly low in Africa and Latin America (Fig. 4 b). Figure 4 Global distribution of patients and studies across low‐ and middle‐income countries (LMICs) in included articles (2008 to present). a Number of studies and b number of patients in studies of LMICs. Countries with fewer than 100 patients recruited for a multinational study were excluded from a, as were studies in which LMIC‐specific patient numbers were not specified48, 49, 50, 55, 60, 81 BJS-11052-FIG-0004-cSubject of studies The focus of study varied across included articles (Fig. 5). Short‐term outcomes of surgery were most commonly captured (33 studies) and, of these studies, eight included over 10 000 patients each. Figure 5 Subject area of large‐volume studies of surgery in low‐ and middle‐income countries in relation to number of study patients

BJS-11052-FIG-0004-cSubject of studies The focus of study varied across included articles (Fig. 5). Short‐term outcomes of surgery were most commonly captured (33 studies) and, of these studies, eight included over 10 000 patients each. Figure 5 Subject area of large‐volume studies of surgery in low‐ and middle‐income countries in relation to number of study patients BJS-11052-FIG-0005-cOutcomes following cancer surgery were common topics, including breast19, 31, 38, 45, 46, 47, 77, gastric16, 22, 23, 61, colorectal24, 59, 76, 81 and prostate18, 30 cancer, and hepatocellular carcinoma56, 60. Cardiac surgery34, 43, 65, 70, caesarean section44, 49, 69 and genitourinary fistula27, 33, 74 were also well represented in included articles, whereas clinical presentations included burn management55, trauma66, appendicitis71, groin hernias35 and orthopaedic fracture management48, 73. However, the overall journey of a patient through the surgical care process was poorly represented, with only a single study53 examining access to surgical care and the cost of surgical care to the patient. No study assessed whether the results of big data analyses have resulted in meaningful changes to healthcare systems or had a significant impact on patient outcomes in LMIC settings.

rocess was poorly represented, with only a single study53 examining access to surgical care and the cost of surgical care to the patient. No study assessed whether the results of big data analyses have resulted in meaningful changes to healthcare systems or had a significant impact on patient outcomes in LMIC settings. A number of studies successfully demonstrated the ability to assemble large prospective data sets on patients across multiple nations. The International Surgical Outcomes Study51 included 15 806 patients in eight LMICs, and the African Surgical Outcomes Study82 included 11 422 patients across 25 African countries. These studies captured mortality and complication rates but, as importantly, were able to capture patient risk profiles and patterns of surgical practice. Highlighting differences in surgical outcome by country‐income level, a lack of critical care provision in LMICs was postulated to significantly influence the ability to rescue patients from complications, with implications for resource planning at a governmental level40, 51, 82. Multinational studies also targeted specific disease areas (GlobalSurg 1: emergency abdominal surgery)57 or specific complications of surgery (GlobalSurg 2: surgical‐site infection)83. These two studies57, 83 gathered prospective data on 23 284 patients and demonstrated that low‐income countries carry a disproportionately higher burden of surgical‐site infection and threefold higher mortality rates.

y abdominal surgery)57 or specific complications of surgery (GlobalSurg 2: surgical‐site infection)83. These two studies57, 83 gathered prospective data on 23 284 patients and demonstrated that low‐income countries carry a disproportionately higher burden of surgical‐site infection and threefold higher mortality rates. Discussion The past 5 years has seen an exponential rise in the number of patients included in studies from LMICs, with some very large cohorts in countries such as Brazil, China and India. Geographical disparities are apparent and are particularly obvious in Africa, where far fewer large studies have been published. The focus is predominately on short‐term outcomes after surgery, together with the epidemiology of diseases commonly treated by surgery. Few studies have focused on the specific needs of resource‐poor environments. It is perhaps too early to determine any positive effects of such work on outcomes in populations of individuals receiving surgical care. The use of big data to capture patient‐level outcomes in an LMIC setting has increased exponentially over the past 10 years. However, in global cohort studies the proportion of patients recruited from high‐income countries remains much greater51, 57, 83. This may suggest the limiting role of infrastructure and resources within LMICs in collecting patient data. Huge disparity with big data applications currently exists globally; no included studies used big data algorithms to identify patient management, predict outcome or direct healthcare policy.

ch greater51, 57, 83. This may suggest the limiting role of infrastructure and resources within LMICs in collecting patient data. Huge disparity with big data applications currently exists globally; no included studies used big data algorithms to identify patient management, predict outcome or direct healthcare policy. In high‐income settings, big data are currently the focus of genomewide data analysis84, developing personal omics profiles85 and individualized oncology treatment86. Meanwhile machine‐learning algorithms are being developed to help deliver care, inform health policy and reduce waste87, 88, 89. Technological infrastructure, specialized analytical skills and personal tracking of health statistics using smart phones, particularly in America, is enabling the amalgamation and analysis of big data from multiple sources on an individual level to offer personalized healthcare packages90. However, real‐time mobile technology application to measure infectious disease outbreaks in LMICs has been realised91, 92, and efforts to develop and incorporate multiple levels of patient data should now be a focus.

s of big data from multiple sources on an individual level to offer personalized healthcare packages90. However, real‐time mobile technology application to measure infectious disease outbreaks in LMICs has been realised91, 92, and efforts to develop and incorporate multiple levels of patient data should now be a focus. Combining data from multiple sources to draw population‐level conclusions worldwide is epitomized by the Global Burden of Disease project by the Institute for Health Metrics and Evaluation at the University of Washington. This is a global effort to examine comprehensively the prevalence, incidence and impact of multiple diseases and environmental factors using an extensive network of more than 2500 collaborators from 133 countries93. Recent publications include global predictions on cancer burden94, child mortality95, causes of adult disability96 and alcohol use97. Such projects require accurate national data, which do not exist in many regions. National registries can be expensive to establish and run, but are becoming more common in middle‐income countries, such as the Chinese Guangzhou Occupational Cohort98 and the Brazilian DATASUS registry80.

Combining data from multiple sources to draw population‐level conclusions worldwide is epitomized by the Global Burden of Disease project by the Institute for Health Metrics and Evaluation at the University of Washington. This is a global effort to examine comprehensively the prevalence, incidence and impact of multiple diseases and environmental factors using an extensive network of more than 2500 collaborators from 133 countries93. Recent publications include global predictions on cancer burden94, child mortality95, causes of adult disability96 and alcohol use97. Such projects require accurate national data, which do not exist in many regions. National registries can be expensive to establish and run, but are becoming more common in middle‐income countries, such as the Chinese Guangzhou Occupational Cohort98 and the Brazilian DATASUS registry80. Comprehensive patient‐level databases or registries are yet to be adopted in the majority of LMICs. Barriers limiting broad adoption include a lack of resources and infrastructure, such as electricity and reliable internet connectivity, combined with skill shortages in medical informatics. The advent of the electronic patient record (EPR) may present the best opportunity for routine data analysis at a health‐system level99. Although the costs of set‐up and maintenance can be a barrier, multiple open‐source EPRs now exist which can potentially alleviate some of these100. Recently, Rwanda announced the roll‐out of the OpenMRS system to 250 clinics and hospitals across the country101. This will bring EPRs into national practice and offer the opportunity for real‐time data collection within a healthcare system to be used for infrastructure planning and research.

lleviate some of these100. Recently, Rwanda announced the roll‐out of the OpenMRS system to 250 clinics and hospitals across the country101. This will bring EPRs into national practice and offer the opportunity for real‐time data collection within a healthcare system to be used for infrastructure planning and research. Linked to this is the explosion in mobile phone technology. Three‐quarters of the population of sub‐Saharan Africa already lives in an area with mobile internet connectivity102. On‐board sensors within mobile phones offer the ability to capture data remotely, without the need for specialized equipment. The increasing availability of mobile phone use is already supplementing existing forms of patient data, particularly in high‐income settings. In surgery, this presents exciting avenues for diagnosis and routine follow‐up, particularly in settings where patients cannot easily attend hospitals.

or specialized equipment. The increasing availability of mobile phone use is already supplementing existing forms of patient data, particularly in high‐income settings. In surgery, this presents exciting avenues for diagnosis and routine follow‐up, particularly in settings where patients cannot easily attend hospitals. There are important areas of study that are more specific to resource‐poor areas, such as access to surgical care and the cost of surgical care to patients. Only one study72 was identified that explored the economic consequences of surgery; this reflected previous findings highlighting large‐scale health economic studies in cancer being focused in high‐income countries or heavily modelled using data from high‐income countries103, 104. Evaluating patient cost following surgery is likely to require frequent and long‐term follow‐up, potentially explaining the difficulties in measuring this outcome. Use of mobile technology to circumvent current logistical issues and capture expenditure data following surgery is an exciting avenue.

me countries103, 104. Evaluating patient cost following surgery is likely to require frequent and long‐term follow‐up, potentially explaining the difficulties in measuring this outcome. Use of mobile technology to circumvent current logistical issues and capture expenditure data following surgery is an exciting avenue. The landscape of healthcare data is changing rapidly. Ensuring that LMICs have the resources to keep up to date with technological advances will ensure future global health equality10. New developments, such as artificial intelligence, virtual reality, mobile computing and new molecular techniques, present exciting opportunities for surgeons across the world. Embedding these technologies within ‘learning’ healthcare systems will ensure that data contribute to the incremental development of safe practice. Big data are capable of providing information on safety, complications and survival; however, with the increasing use of big data, care must be taken to account for unknown and unrecognized confounders in order to determine intervention effectiveness and provide strong observational conclusions105. In parallel with future advances, ensuring that electronic data are kept secure is of utmost importance. Respecting an individual patient's rights to confidentiality, autonomy and privacy is fundamental to ensuring public trust in electronic data collection methods. Beyond good data governance practice, technologies such as blockchain may facilitate the safe and secure sharing of healthcare data within increasingly complex interconnected systems.

patient's rights to confidentiality, autonomy and privacy is fundamental to ensuring public trust in electronic data collection methods. Beyond good data governance practice, technologies such as blockchain may facilitate the safe and secure sharing of healthcare data within increasingly complex interconnected systems. There are weaknesses to the approach taken in this review. Pragmatic limitations around the scope of the review search were required and important studies may have been omitted. The synthesis of such a heterogenous group of studies is difficult and conclusions must be made at a high level. This review has demonstrated a significant growth in the use of large‐volume patient‐level data across many surgical specialties and LMICs. At least 71 LMICs currently involved in big data projects were identified, with evidence of an exponential growth in patient numbers totalling more than 700 000. However, to date, the majority of studies using big data have been limited to short‐term outcomes after surgery and few have addressed the needs that are particular to LMICs. Funders, policymakers and specialists in medical informatics urgently need to reorientate this focus if the potential of big data to improve surgical outcomes, particularly in LMICs, is to be realized fully. Disclosure The authors declare no conflict of interest. Supporting information Table S1 Studies included within the systematic review Table S2 Aims of included studies grouped by primary outcome measure and surgical specialty Click here for additional data file.

Introduction Normothermic regional perfusion (NRP) is a technique that aids organ recovery from donors after circulatory death (DCDs), leading to acceptable transplantation outcomes1. NRP was first used in 1989 by Spanish transplant surgeons using a percutaneously placed cardiopulmonary bypass circuit2. Autologous blood from the donor is used as a perfusate and is anticoagulated with heparin. NRP closely resembles other extracorporeal membrane oxygenation (ECMO) circuits that can be used for cardiopulmonary bypass, extracorporeal cardiopulmonary resuscitation (ECPR) or for ventilatory assistance in patients with refractory acute lung injury. ECMO provides mechanical ventilatory and circulatory support via an extracorporeal circuit incorporating a membrane oxygenator, centrifugal pump or roller, heat exchanger and perfusate reservoir. NRP and ECMO have differences, which include the anatomical location of cannulas (peripheral versus central) and flow rates (lower in NRP). NRP improves the viability of organs for kidney and liver transplantation, and leads to better post‐transplantation function with fewer complications1. Recently, NRP has been used to support donation of cardiothoracic organs, permitting recirculation and restoration of cardiac activity leading to transplantation3.

NRP closely resembles other extracorporeal membrane oxygenation (ECMO) circuits that can be used for cardiopulmonary bypass, extracorporeal cardiopulmonary resuscitation (ECPR) or for ventilatory assistance in patients with refractory acute lung injury. ECMO provides mechanical ventilatory and circulatory support via an extracorporeal circuit incorporating a membrane oxygenator, centrifugal pump or roller, heat exchanger and perfusate reservoir. NRP and ECMO have differences, which include the anatomical location of cannulas (peripheral versus central) and flow rates (lower in NRP). NRP improves the viability of organs for kidney and liver transplantation, and leads to better post‐transplantation function with fewer complications1. Recently, NRP has been used to support donation of cardiothoracic organs, permitting recirculation and restoration of cardiac activity leading to transplantation3. NRP works in three ways. First, it acts as a perfusion bridge between asystole and organ procurement, and allows organ procurement without further injury from prolonged ischaemia. Second, it enables rehabilitation at a cellular level by replenishing mitochondrial stores of adenosine 5′‐triphosphate, thereby mitigating against the effects of anaerobic metabolism and mimicking a period of ischaemic preconditioning4, 5, 6. Third, it permits donor organ assessment under non‐ischaemic conditions over a period of time, as well as the opportunity to track physiological responses to reperfusion.

adenosine 5′‐triphosphate, thereby mitigating against the effects of anaerobic metabolism and mimicking a period of ischaemic preconditioning4, 5, 6. Third, it permits donor organ assessment under non‐ischaemic conditions over a period of time, as well as the opportunity to track physiological responses to reperfusion. A number of ethical considerations exist regarding the use of NRP. These include premortem cannulation and systemic heparinization; variable requirements for stand‐off time (the interval between confirmation of death and commencement of the perfusion process); and the potential for reperfusion to lead to return of spontaneous cerebral and cardiac activity (ROSCCA). The latter could invalidate the declaration of death, thereby threatening the fundamental tenet of deceased donation following cardiac arrest. ROSCCA can be assured categorically only by exclusion of the coronary and cerebral circulations from the NRP circuit. In abdominal organ donors, this is achieved by mechanical occlusion of the supracoeliac aorta most commonly using an inflated balloon, or by cross‐clamping. Nonetheless, the potential for ROSCCA to occur while using an inflated aortic occlusion balloon remains poorly understood and characterized. Concerns have been expressed that NRP, with or without inflation of an intra‐aortic occlusion balloon, provides an unsatisfactory assurance that ROSCCA would not occur7, 8.

mping. Nonetheless, the potential for ROSCCA to occur while using an inflated aortic occlusion balloon remains poorly understood and characterized. Concerns have been expressed that NRP, with or without inflation of an intra‐aortic occlusion balloon, provides an unsatisfactory assurance that ROSCCA would not occur7, 8. This study aimed to determine the likelihood of ROSCCA in NRP‐DCDs of abdominal organs, and the impact an inflated intra‐aortic occlusion balloon may have on the likelihood of ROSCCA. The ethical implications of this increasingly used organ preservation technique in an era of donor organ shortage are also explored. Methods Validation cohort and definition of terms A validation cohort of patients who most closely resembled organ donors, but who received ECMO for therapeutic indications without an inflated intra‐aortic occlusion balloon, was identified to determine the likelihood of ROSCCA occurring in NRP‐DCDs. Patients receiving ECMO as part of resuscitative measures following cardiac arrest (ECPR for out‐of‐hospital cardiac arrest (OOHCA)) were considered an appropriate comparator for NRP‐DCDs. This is because ECPR is employed for refractory cardiac arrest where cessation of resuscitative efforts would lead to the inevitable declaration of death. The technical and mechanical components of NRP and ECPR are virtually identical, but two important distinctions exist. First, ECPR has resuscitative intent whereas the intent of NRP is preservation; and, second, there is no intra‐aortic occlusion balloon in ECPR.

Methods Validation cohort and definition of terms A validation cohort of patients who most closely resembled organ donors, but who received ECMO for therapeutic indications without an inflated intra‐aortic occlusion balloon, was identified to determine the likelihood of ROSCCA occurring in NRP‐DCDs. Patients receiving ECMO as part of resuscitative measures following cardiac arrest (ECPR for out‐of‐hospital cardiac arrest (OOHCA)) were considered an appropriate comparator for NRP‐DCDs. This is because ECPR is employed for refractory cardiac arrest where cessation of resuscitative efforts would lead to the inevitable declaration of death. The technical and mechanical components of NRP and ECPR are virtually identical, but two important distinctions exist. First, ECPR has resuscitative intent whereas the intent of NRP is preservation; and, second, there is no intra‐aortic occlusion balloon in ECPR. In OOHCA, a variable period of absent perfusion (no‐flow time) exists between the onset of cardiac arrest and the commencement of cardiopulmonary resuscitation (CPR) with chest compressions, either by a bystander or a trained professional. A period of low flow exists from the commencement of mechanical compressive CPR until the return of perfusion by either return of spontaneous circulation or ECPR.

onset of cardiac arrest and the commencement of cardiopulmonary resuscitation (CPR) with chest compressions, either by a bystander or a trained professional. A period of low flow exists from the commencement of mechanical compressive CPR until the return of perfusion by either return of spontaneous circulation or ECPR. In the organ donation process, stand‐off time refers to the interval after declaration of death up to commencement of the donor perfusion and organ preservation process. A stand‐off time, commonly 5 min, is necessary in most jurisdictions to ensure that autoresuscitation does not take place and that the declaration of death is valid9. In uncontrolled DCDs there is, therefore, a no‐flow time, a low‐flow time and a stand‐off time before return of perfusion by NRP. In controlled DCDs there is only a low‐flow time (following withdrawal of life‐sustaining treatment) and a stand‐off time (following asystole). ECPR‐OOHCA is therefore an appropriate surrogate to determine the likelihood of ROSCCA in NRP‐DCDs.

ime, a low‐flow time and a stand‐off time before return of perfusion by NRP. In controlled DCDs there is only a low‐flow time (following withdrawal of life‐sustaining treatment) and a stand‐off time (following asystole). ECPR‐OOHCA is therefore an appropriate surrogate to determine the likelihood of ROSCCA in NRP‐DCDs. Search methods A search of MEDLINE, Embase and the Cochrane Library electronic databases was performed to identify all articles relating to NRP for DCDs of abdominal organs. The following Medical Subject Heading (MeSH) terms were used and combined using Boolean operators: ‘extracorporeal membrane oxygenation’, ‘donors after cardiac death’, ‘non‐heart‐beating donors’, ‘donors after circulatory death’, ‘normothermic recirculation’, ‘normothermic perfusion’, ‘regional perfusion’, ‘liver transplantation’, ‘kidney transplantation’ and ‘pancreas transplantation’. A further search was carried out in an identical manner to identify all articles relating to ECPR‐OOHCA. The following MeSH terms were used: ‘extracorporeal life support’, ‘extracorporeal resuscitation’, ‘extracorporeal cardiopulmonary resuscitation’, ‘extracorporeal membrane oxygenation’ and ‘out‐of‐hospital cardiac arrest’. References of all identified papers were searched to ensure a comprehensive review. PRISMA guidelines were followed10. Ethical approval was not required.

support’, ‘extracorporeal resuscitation’, ‘extracorporeal cardiopulmonary resuscitation’, ‘extracorporeal membrane oxygenation’ and ‘out‐of‐hospital cardiac arrest’. References of all identified papers were searched to ensure a comprehensive review. PRISMA guidelines were followed10. Ethical approval was not required. Inclusion and exclusion criteria All study designs including cohort studies, case–control studies and case series were considered eligible. Studies performed between January 1997 and June 2016 were included. Case reports, conference abstracts, review articles, animal studies and articles not written in English were excluded. For the NRP‐DCD search, studies of organ donation of non‐abdominal organs and those using an ex situ perfusion process were also excluded. For the ECPR‐OOHCA search, articles that did not describe the relationship between no‐flow time and outcome were excluded. Publications reporting outcomes in ECPR for in‐hospital cardiac arrest were excluded, because the mean delay from cardiac arrest to commencement of CPR is much shorter than in OOHCA. Data extraction Data extraction was undertaken using a standard pro forma. Patient and donor age, perfusion flow rates, and rates of failure in establishing successful perfusion were recorded. The NRP and ECPR protocols were interrogated, and exclusion and inclusion criteria identified and compared. Where recorded, methods of preventing and identifying cerebral and cardiac perfusion in NRP‐DCDs were also identified. In ECPR‐OOHCA, the duration of no flow was recorded.

hing successful perfusion were recorded. The NRP and ECPR protocols were interrogated, and exclusion and inclusion criteria identified and compared. Where recorded, methods of preventing and identifying cerebral and cardiac perfusion in NRP‐DCDs were also identified. In ECPR‐OOHCA, the duration of no flow was recorded. Outcome measures Survival outcomes (discharge from hospital with a favourable neurological outcome) and cause of death following ECPR‐OOHCA were recorded and analysed according to the duration of absence of perfusion. Neurological outcomes were considered favourable if recorded as having a cerebral performance category of 1 or 2 (Table 1)11. In the NRP‐DCD groups, the incidence of ROSCCA was recorded. Survival following ECPR‐OOHCA, where ECPR commenced beyond the critical time of 5 min, was used as a surrogate marker to determine the likelihood of ROSCCA associated with NRP‐DCDs. Table 1 Cerebral performance categories Scale Category Level of function 1 Good cerebral performance Normal living 2 Moderate cerebral disability Sufficient function for independent activities of daily living 3 Severe cerebral disability Limited cognition Fully dependent on others for daily living 4 Coma or vegetative state Cerebral unresponsiveness or any degree of coma without fully meeting the criteria for brain death 5 Brain death Apnoea, areflexia and electroencephalographic silence These provide guidance regarding functional neurological recovery status following brain injury. Adapted from Safar11.

living 4 Coma or vegetative state Cerebral unresponsiveness or any degree of coma without fully meeting the criteria for brain death 5 Brain death Apnoea, areflexia and electroencephalographic silence These provide guidance regarding functional neurological recovery status following brain injury. Adapted from Safar11. Results The database searches identified 410 articles relating to NRP‐DCDs, 12 of which were eligible for inclusion12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, and 947 articles relating to ECPR‐OOHCA, of which eight were eligible for inclusion24, 25, 26, 27, 28, 29, 30, 31 (Fig. 1). This provided a cumulative total of 493 potential DCDs who received NRP organ preservation and 254 patients who received ECPR following OOHCA. Figure 1 PRISMA flow diagram showing selection of articles for review. NRP‐DCD, normothermic regional perfusion for donors after circulatory death; ECPR‐OOHCA, extracorporeal resuscitation for out‐of‐hospital cardiac arrest

, supporting information). Somalia and Democratic People's Republic of Korea were excluded owing to lack of health expenditure data. Based on these estimates, 90 of the 194 WHO member states are below the Commission's recommended target of 5000 operations per 100 000 people (Tables S18 and S19, supporting information). Concerning utility, a major challenge is the numerous definitions of surgical procedure18, 28 as well as different procedure classification systems29, 30, 31. There are particular challenges capturing surgical volume in the private sector. It was also noted that the definition put forward by the Commission does not take into account differing country need28. Indicator 4: perioperative mortality Nationwide data on number of deaths following operations were available from 28 WHO member states (Table 2; Table S20, supporting information). Concerning comparability, there were two different definitions used: nine countries relied exclusively on in‐hospital mortality, seven used 30‐day mortality and 12 provided no definition. Without information on case mix and preoperative patient risk, comparability would remain limited, and potential to game this indicator and achieve targets by doing low‐risk procedures on fit patients would be high. The utility of this metric for benchmarking is limited primarily by data availability; however, comparability would be likely to influence utility even if data were available more widely.

Results The database searches identified 410 articles relating to NRP‐DCDs, 12 of which were eligible for inclusion12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, and 947 articles relating to ECPR‐OOHCA, of which eight were eligible for inclusion24, 25, 26, 27, 28, 29, 30, 31 (Fig. 1). This provided a cumulative total of 493 potential DCDs who received NRP organ preservation and 254 patients who received ECPR following OOHCA. Figure 1 PRISMA flow diagram showing selection of articles for review. NRP‐DCD, normothermic regional perfusion for donors after circulatory death; ECPR‐OOHCA, extracorporeal resuscitation for out‐of‐hospital cardiac arrest BJS-11046-FIG-0001-cProtocols The NRP‐DCD and ECPR‐OOHCA perfusion protocols are reported in Tables 2 and 3 respectively. NRP‐DCD protocols used an inflated occlusion balloon in the supracoeliac or thoracic aorta to prevent reperfusion of supradiaphragmatic organs in all but one series13. Vascular access was obtained via the femoral vessels in all but one case each of NRP and ECPR. Oniscu and colleagues13 performed rapid laparotomy, cannulation of the aorta and vena cava, and cross‐clamping of the thoracic aorta. There was no difference between NRP‐DCD and ECPR‐OOHCA eligibility criteria and cohorts with respect to patient and donor age, perfusion circuit flow rates, and rate of failure to commence the perfusion process. Table 2 Perfusion protocols for normothermic regional perfusion donor after circulatory death programmes

BJS-11046-FIG-0001-cProtocols The NRP‐DCD and ECPR‐OOHCA perfusion protocols are reported in Tables 2 and 3 respectively. NRP‐DCD protocols used an inflated occlusion balloon in the supracoeliac or thoracic aorta to prevent reperfusion of supradiaphragmatic organs in all but one series13. Vascular access was obtained via the femoral vessels in all but one case each of NRP and ECPR. Oniscu and colleagues13 performed rapid laparotomy, cannulation of the aorta and vena cava, and cross‐clamping of the thoracic aorta. There was no difference between NRP‐DCD and ECPR‐OOHCA eligibility criteria and cohorts with respect to patient and donor age, perfusion circuit flow rates, and rate of failure to commence the perfusion process. Table 2 Perfusion protocols for normothermic regional perfusion donor after circulatory death programmes Reference Location Donor type Flow rate Age eligibility (years) Demiselle et al.12 France Uncontrolled 2–3·7 l/min 18–60 Oniscu et al.13 UK Controlled 1·7–4 l/min Variable* Fondevila et al.14 Spain Uncontrolled 1·7 l/min < 65 Jiménez‐Galanes et al.15 Spain Uncontrolled 3·1 l/min < 50 Otero et al.16 Spain Uncontrolled n.r. < 50 Rojas‐Peña et al.17 USA Controlled > 45 ml per kg per min 0·5–65 Sánchez‐Fructuoso et al.18 Spain Uncontrolled n.r. < 60 Valero et al.19 Spain Uncontrolled 1–2 l/min < 65 Reznik et al.20 Russia Uncontrolled 2·5 l/min n.r. Farney et al.21 USA Controlled 4–6 l/min < 60 Lee et al.22 Taiwan Controlled 2 l/min n.r. Koyama et al.23 Japan Controlled 2–3·5 l/min n.r. * Depending on organ to be transplanted. n.r., Not reported.

ntrolled n.r. < 60 Valero et al.19 Spain Uncontrolled 1–2 l/min < 65 Reznik et al.20 Russia Uncontrolled 2·5 l/min n.r. Farney et al.21 USA Controlled 4–6 l/min < 60 Lee et al.22 Taiwan Controlled 2 l/min n.r. Koyama et al.23 Japan Controlled 2–3·5 l/min n.r. * Depending on organ to be transplanted. n.r., Not reported. Table 3 Perfusion protocol for extracorporeal cardiopulmonary resuscitation for out‐of‐hospital cardiac arrest programmes Reference Location Commencement of ECPR in prehospital phase Maximum no‐flow time (min) Flow rate Age eligibility (years) Kagawa et al.24 Japan No < 15 > 2 l/min (target 2·5 l/min) 18–74 Bellezzo et al.25 USA No < 10 Svo 2 > 70% MAP > 65 mmHg n.r. Ferrari et al.26 Germany No < 5 > 3 l/min < 75 Maekawa et al.27 Japan No Variable 50–60 ml per min per kg Variable Lamhaut et al.28 France Yes < 5 2·5–4 l/min < 70 Avalli et al.29 Italy No Variable 2·6 l per min per m2 12–75 Le Guen et al.30 France No < 5 3–4 l/min n.r. Mégarbane et al.31 France No Variable 2·5 l/min n.r. ECPR, extracorporeal cardiopulmonary resuscitation; Svo 2, venous oxygen saturation; MAP, mean arterial pressure; n.r., not reported.

e Yes < 5 2·5–4 l/min < 70 Avalli et al.29 Italy No Variable 2·6 l per min per m2 12–75 Le Guen et al.30 France No < 5 3–4 l/min n.r. Mégarbane et al.31 France No Variable 2·5 l/min n.r. ECPR, extracorporeal cardiopulmonary resuscitation; Svo 2, venous oxygen saturation; MAP, mean arterial pressure; n.r., not reported. Outcomes Demographic data and outcomes from the NRP‐DCD and ECPR‐OHCA programmes are reported in Tables S1 and S2 (supporting information) respectively. There were no reported instances or evidence of reperfusion leading to ROSCCA in any NRP‐DCD programme. No organ donation procedures were abandoned owing to concerns relating to potential cardiac and cerebral reperfusion and ROSCCA. In the ECPR‐OOHCA programmes, there were no survivors with a favourable neurological outcome where the absence of perfusion lasted more than 5 min.

leading to ROSCCA in any NRP‐DCD programme. No organ donation procedures were abandoned owing to concerns relating to potential cardiac and cerebral reperfusion and ROSCCA. In the ECPR‐OOHCA programmes, there were no survivors with a favourable neurological outcome where the absence of perfusion lasted more than 5 min. Discussion There is no evidence to suggest that ROSCCA may occur following the institution of NRP in DCDs where the proximal aorta has been occluded. Survival with a favourable neurological outcome following ECPR for refractory OOHCA does not occur when the no‐flow time exceeds 5 min. Following ECPR‐OOHCA where a no‐flow time was not specified, a small cohort of patients (8 of 254) survived to discharge, either in a persistent vegetative state or with significant neurological disability. It is not possible using the existing data to provide categorical assurance that ROSCCA would not occur in NRP‐DCDs in the absence of a proximally occluded aorta. The periods of no flow and low flow preceding the stand‐off time, in uncontrolled and controlled DCDs respectively, inevitably compound the ischaemic insult and further reduce the likelihood of ROSCCA.

o provide categorical assurance that ROSCCA would not occur in NRP‐DCDs in the absence of a proximally occluded aorta. The periods of no flow and low flow preceding the stand‐off time, in uncontrolled and controlled DCDs respectively, inevitably compound the ischaemic insult and further reduce the likelihood of ROSCCA. Japan presents an interesting situation in that organ donation after brain death, although legal, is not acceptable culturally, but the country also has the largest worldwide practice of ECPR‐OOHCA32, 33. Studies by Morimura and colleagues32 and Sakamoto et al.33 were not eligible for inclusion in this review as the rates of survival and ROSCCA were not reported in relation to the no‐flow time or bystander CPR. Survivors of OOHCA with a favourable neurological outcome (12·3 per cent) and those who remained in a coma or vegetative state (10·4 per cent) were increased in equal proportion following use of ECPR, thereby providing scope for ethical debate on the overall impact of ECPR33. However, in Japan, with a paucity of deceased donor organs, public perception, as well as that of health policymakers, may be swayed in favour of ECPR‐OOHCA given that it could also lead to 60–70 additional organ donors annually.

g use of ECPR, thereby providing scope for ethical debate on the overall impact of ECPR33. However, in Japan, with a paucity of deceased donor organs, public perception, as well as that of health policymakers, may be swayed in favour of ECPR‐OOHCA given that it could also lead to 60–70 additional organ donors annually. Organ donation from DCDs is a sensitive issue, with associated fears of NRP‐related ROSCCA that are both appreciable and justified, especially given the reported incidents of prolonged survival after brain death34. However, such concerns are not supported by the existing evidence based on current practice. Non‐invasive methods of recording cerebral and coronary blood flow exist as an additional safety measure. A Doppler ultrasound probe can be placed over the carotid artery to detect cerebral blood flow and a radial artery catheter is inserted to detect pressure changes35. In Michigan, a lidocaine bolus (1–2 g) was included in the blood perfusate to prevent return of spontaneous cardiac activity17. If cardiac transplantation from NRP‐DCDs is to remain a regular feature of organ donation practice, monitoring of cerebral activity with an electroencephalogram may provide a necessary adjunct, as a safeguarding measure.

us (1–2 g) was included in the blood perfusate to prevent return of spontaneous cardiac activity17. If cardiac transplantation from NRP‐DCDs is to remain a regular feature of organ donation practice, monitoring of cerebral activity with an electroencephalogram may provide a necessary adjunct, as a safeguarding measure. An inflated occlusion balloon in the proximal aorta has significant advantages, as perfusion can be commenced following a minimally invasive approach, and allows organ assessment before a definitive decision on organ procurement is made. In some instances, it is possible that the integrity of the inflated aortic occlusion balloon may have been compromised, but went unrecognized and unreported. However, balloon integrity can be confirmed by the presence/absence of a radial BP signal, both before and after commencing NRP. In the UK, cross‐clamping the thoracic aorta following rapid thoracolaparotomy is the preferred method for NRP‐assisted controlled DCD procurement of abdominal organs. However, thoracotomy ought not be a prerequisite for NRP, and minimally invasive NRP provides particular advantages in uncontrolled donors in societies where a presumed consent process is part of the legislation.

thoracolaparotomy is the preferred method for NRP‐assisted controlled DCD procurement of abdominal organs. However, thoracotomy ought not be a prerequisite for NRP, and minimally invasive NRP provides particular advantages in uncontrolled donors in societies where a presumed consent process is part of the legislation. It has been mooted that NRP‐DCDs without a proximal occlusion of the aorta would be more appropriately termed as donors after brain circulatory death because recirculation of blood occurs following NRP, and that this procedure should therefore require specific consent8. The ethical and physiological boundaries of life and death remain controversial, and it is therefore imperative to characterize the impact of recirculation on the potential likelihood of ROSCCA as this clearly has significant implications for consent in organ donation34. The absence of ROSCCA may imply that no additional specific consent would be required. In countries where ECPR‐OOHCA is practised routinely, the boundaries between ECPR for resuscitative purposes and NRP as an organ preservation technique may become blurred. Where a patient resuscitated by ECPR has a cerebral performance category of 4 or 5 but is dependent on ECMO for circulatory support, management of the potential organ donor should proceed identically to that for ECMO‐independent patients. However, ECMO‐independent patients existing in a vegetative state present a very challenging set of circumstances that requires careful individual consideration by families, clinicians and judicial representatives.

management of the potential organ donor should proceed identically to that for ECMO‐independent patients. However, ECMO‐independent patients existing in a vegetative state present a very challenging set of circumstances that requires careful individual consideration by families, clinicians and judicial representatives. The main limitation of this study is the small number of ECPR‐OOHCA papers that were eligible for inclusion to approximate NRP‐DCD donation. Only eight of 947 potential papers on ECPR‐OOHCA were eligible and a proportion of articles, especially those from Japan, were excluded; this could have contributed to unintended bias. The criteria for the declaration of death, in particular brain death, remain controversial across the spectrum of clinical, scientific and religious communities, including within societies where organ donation practices are well established34. This study has demonstrated that ROSCCA leading to survival with favourable neurological outcome following ECPR‐OOHCA does not occur when the duration of no flow exceeds 5 min. This suggests a similar process in NRP‐DCDs, with a stand‐off time of 5 min. ROSCCA in NRP‐DCDs with occlusion of the proximal aorta has not been reported to date. The presence of an occlusion balloon provides an additional mechanical level of assurance in a very sensitive clinical scenario. This is particularly important as any adverse events would have a catastrophic impact on the public perception of organ donation and transplantation. Supporting information

This study has demonstrated that ROSCCA leading to survival with favourable neurological outcome following ECPR‐OOHCA does not occur when the duration of no flow exceeds 5 min. This suggests a similar process in NRP‐DCDs, with a stand‐off time of 5 min. ROSCCA in NRP‐DCDs with occlusion of the proximal aorta has not been reported to date. The presence of an occlusion balloon provides an additional mechanical level of assurance in a very sensitive clinical scenario. This is particularly important as any adverse events would have a catastrophic impact on the public perception of organ donation and transplantation. Supporting information Table S1 Demographics and outcomes from the normothermic donor after circulatory death (NRP‐DCD) programmes Table S2 Demographics and outcomes from the extracorporeal cardiopulmonary resuscitation for out of hospital cardiac arrest (ECPR‐OOHCA) programmes Click here for additional data file. Acknowledgements This study was supported by the UK Medical Research Council and the Royal College of Surgeons of Edinburgh. Disclosure: The authors declare no conflict of interest.

remain limited, and potential to game this indicator and achieve targets by doing low‐risk procedures on fit patients would be high. The utility of this metric for benchmarking is limited primarily by data availability; however, comparability would be likely to influence utility even if data were available more widely. Indicators 5–6: protection against impoverishing and catastrophic expenditure No data determined to be nationally representative of the out‐of‐pocket cost of surgery were found, a prerequisite for calculating these metrics. Data were provided from five countries on the provider costs of surgery, which showed large variation between types of surgery and between sources (Table S21, supporting information)20, 21. Comparability and utility could not be assessed without available data; indeed, without any data, the indicators are of no use. However, in principle, these indicators are of huge value in assessing progress towards universal health coverage. Unfortunately, there were not enough data points to allow updated modelling of these indicators.

Introduction The tumour microenvironment in colorectal cancer is influenced by somatic mutational and epigenetic events that occur during tumour development, as well as by the host immune system, which exerts negative selection pressures on tumour cells, by recognition of tumour antigens as non‐self1. Immune checkpoints are a series of innate and adaptive regulatory mechanisms to modulate immune activity and promote tolerance to self‐antigens. These can be upregulated in tumours to drive resistance to immune cell‐mediated destruction2, 3. Immunotherapy has been most successful in targeting and blocking these immune checkpoints, leading to effective antitumour responses in some cancers4. The emergence of immunotherapy has transformed the treatment landscape of some cancers, most notably cutaneous melanoma5, 6 and non‐small cell lung cancer (NSCLC)7, 8. So far, the role of immunotherapy in colorectal cancer been limited to the 3–4 per cent of patients with metastatic disease whose tumours demonstrated microsatellite instability (MSI)9, due to germline, somatic or epigenetic inactivation of DNA mismatch repair (MMR) genes10. However, its role could be expanded significantly by drawing on an understanding of the immunogenomic drivers of the response in the tumour environment.

astatic disease whose tumours demonstrated microsatellite instability (MSI)9, due to germline, somatic or epigenetic inactivation of DNA mismatch repair (MMR) genes10. However, its role could be expanded significantly by drawing on an understanding of the immunogenomic drivers of the response in the tumour environment. This review explores current understanding of the relative contributions of innate immune genomic mechanisms and somatic mutations to the immune environment in colorectal cancer, with the implications for potential expansion of the roles of immunotherapy and other targeted therapies in the management of colorectal cancer at all disease stages. Methods Search strategy A literature search was conducted using the PubMed, MEDLINE and Cochrane Library databases, as well as reference lists from appropriate papers. The goal was to provide an overview of published research in the field of colorectal cancer genomics and immunology, with a particular focus on advances since the launch of the genomics era after completion of the Human Genome Project11. The following keywords were used to perform flexible searches within these databases: ‘immunotherapy’, ‘colorectal’ AND ‘cancer’, ‘mutation’, ‘immunity’ and ‘immunologic adjuvants’. Only papers published in English were included.