Differences in the management of patients requiring an emergency resection for colonic cancer in two European populations

Abstract Background Patients with colonic cancer who require emergency colonic cancer surgery often experience poorer outcomes compared with their elective counterparts. In this setting, several treatments approaches are available. In 2009, Danish guidelines recommended treatment with stent for obstruction in left-sided tumours as a bridge to surgery, if expertise is accessible. The aim of this study was to compare the use of elective and emergency resections for colonic cancer and postoperative mortality in two similar demographic populations. Methods All patients who underwent a major resection for colonic cancer, between 2005 and 2016 in Denmark and Yorkshire (UK) were identified. The proportion undergoing emergency surgery, the proportion receiving a stent procedure before their resection, and 30-day postoperative mortality were compared between the populations. Logistic regression was used to determine changes in the proportion of those undergoing emergency surgery and 30-day postoperative mortality. Results Out of 45 397 patients treated during the study interval, 41 880 were selected. Emergency surgery decreased in Denmark from 16.6 per cent in 2005–07 to 12.9 per cent in 2014–16, but increased in Yorkshire (13.5 per cent to 16.8 per cent). Danish patients with left-sided tumours were less likely to undergo emergency surgery (risk ratio 0.90, 95 per cent c.i. 0.82 to 0.99) and an increase in stent use coincided with a statistically significant decrease in emergency surgery in these patients. Thirty-day postoperative mortality in all resections (elective and emergency) decreased in both populations, but a larger decrease was observed in Denmark (7.7 per cent to 3.0 per cent in Denmark and 7.1 per cent to 3.3 per cent in Yorkshire). Conclusion Patients in Denmark experienced a reduction in the use of emergency resection and increase in stenting procedures, following the policy implemented in some departments of converting potential emergency resections into elective resections.


Introduction
Approximately 15-30 per cent of patients with colorectal cancer present as an emergency 1 . Those who require an emergency surgical resection tend to have worse short-term outcomes than those who receive elective care, and these include increased rates of postoperative mortality, postoperative complications, and length of hospital stay [2][3][4][5] . Poorer survival and recurrence rates have also been reported in some, but not all studies [6][7][8] . The introduction of alternative treatments for suitable patients, such as, self-expanding metallic stents (SEMs) as a bridge to surgery, may reduce the number of patients undergoing an emergency resection 9 ; however, because of the expertise required and the lack of consensus in published guidelines, the utilization of treatment options such SEMs are likely to vary both within and across populations 10,11 .
Comparisons in the management of patients with colorectal cancer between similar demographic populations, can help to identify differences in practices, which may have an impact on outcomes 12,13 , but few studies have been undertaken. The Yorkshire Cancer Research Bowel Cancer Improvement Programme (YCR BCIP) is using this philosophy by attempting to identify areas of improvement in the management of patients with colorectal cancer in Yorkshire 14 . Denmark and the UK both have healthcare provision that is largely free at the point of care and financed through taxation, with a similar disease burden and adult life expectancy 15, 16 . This, and given that Denmark has a similar population size to Yorkshire (both 5.7 million), presents the opportunity to use Denmark as a suitable comparator in the management of patients with colorectal cancer.
In 2009, guidelines were issued in Denmark recommending treatment with SEMs for obstruction in left-sided tumours as a bridge to surgery 17 . Given the demanding technical skill, however, it was emphasized that the use of stents in an emergency situation should only be performed in departments with the necessary expertise.
The present study was undertaken to compare the use of elective and emergency resections for colonic cancer and postoperative mortality between Denmark and the region of Yorkshire, UK.

Methods
This was a retrospective population-based study of patients with a first primary colonic cancer diagnosed in Denmark and the region of Yorkshire, UK. All patients aged 18 years or older were diagnosed between 1 January 2005 and 31 December 2016 (ICD-10 C18 and C19) and had undergone a major surgical resection at a Danish or an English NHS hospital, up to 1 year after diagnosis. Patients with a malignant neoplasm of the appendix (ICD-10 C18.1) were excluded as these were not recorded in Denmark from 2014.
Danish patients were obtained from the Danish Colorectal Cancer Group (DCCG) database 18 , which captures all patients with colorectal cancer who have been diagnosed and/or treated at a public hospital. The main surgical procedure, and urgency of that procedure, is recorded in the DCCG and those patients recorded as undergoing a major resection procedure were identified and categorized as either an elective or emergency resection. Major resection was defined as receiving selected radical procedures (Table S1) occurring within 1 month before, and up to 1 year after, the date of diagnosis. No formal definition of emergency surgery is given in the DCCG, except that the indication for emergency surgery (ileus, perforation, bleeding, or otherwise) should be reported. The majority of those classed as an emergency are operated on within 36 h of admission. Surgical procedures, and other interventions, before the main surgery are also recorded and those who had a stent inserted before their main surgical procedure were identified.
Patients in Yorkshire were identified using the UK Colorectal Cancer Intelligence Hub's COloRECTal Repository (CORECT-R) by linkage of the data from the national cancer registry (National Cancer Registration and Analysis Service) and inpatient hospital admissions (Hospital Episode Statistics, HES) 19 . Major surgical resection is defined in CORECT-R using a methodology that matches procedure codes within HES to identify all operations used to surgically treat colorectal cancer. The nature of the patients' admission for the procedure is also recorded. A resection is categorized as an emergency resection if the procedure was undertaken within 2 days of an emergency admission, and an elective resection otherwise. Those who had a recorded HES procedure with OPSC4 codes pertaining to the insertion of a stent (H214, H243, H244, H273, H274, H314, and Y141 to Y149) up to 30-days before, or with the same date as the major surgical resection, were identified as having a stent inserted.

Outcomes of interest
The primary outcome of interest was the rate of emergency resection in the two populations over time. Secondary outcomes included the rate of stent procedures, hospital variation in the use of emergency resection, and 30-day postoperative mortality following elective and emergency resections by study interval, and defined as death of the patient within 30-days of the resection date.

Statistics
The proportion of emergency resections was calculated as a percentage of all resections over the study interval and by To investigate the factors associated with use of emergency resection and to test the statistical significance of changes over time, Danish and Yorkshire populations were modelled separately using multilevel mixed effects. Poisson regression with a robust error variance to estimate the risk ratio (RR), as is recommended for binary outcomes where the probability of the outcome is common 20 . The binary dependent variable was emergency resection or not, independent fixed effects were age group (18-59, 60-69, 70-79, and 80 years or older), sex, tumour site (right, left, or unspecified), stage of disease (1)(2)(3)(4) and study interval, and hospital where the operation was conducted was fitted as a random effect. Tumours located in the caecum, ascending colon, hepatic flexure, and transverse colon were categorized as right-sided tumours, whereas those in the splenic flexure, descending colon, sigmoid colon, and rectosigmoid junction were categorized as left-sided tumours. Stage of disease was missing in 2.4 per cent and 4.0 per cent of Danish and Yorkshire patients respectively. Therefore, an ordered logistic regression was used to impute missing values and estimated model coefficients and standard errors according to Rubin's combination rules.
Variation in the use of emergency resection by hospital was assessed by the median OR (MOR), calculated from the estimated variance of the distribution of random effects after fitting of multilevel logistic models with the same covariates described above and stratified by study interval. The MOR quantifies the variation areas between the second-level variation (hospitals in the models) and allows comparisons with the fixed effects covariates on the OR scale 21,22 . A MOR equal to one would indicate no variation in the use of emergency surgery between hospitals, whereas a MOR more than one would indicate variation. Bootstrapping was performed to calculate biased-corrected 95 per cent confidence intervals for the variance estimate from the multilevel logistic models using 200 replications, which were then used to create confidence intervals for the MORs 23 .
To moderate the effect of colorectal screening programmes on changes in the emergency resection proportion, for all analyses, Danish patients who were registered as being diagnosed based on the Danish Screening Programme and Yorkshire patients who had a screening diagnosis as defined by the Routes to Diagnosis methodology 24 were excluded.
The RR for deaths following emergency resection were calculated and compared with elective resection by fitting a multilevel mixed-effects regression model, with hospital as a random effect. The binary dependent variable was 30-day postoperative mortality, and independent fixed effects were age, sex, tumour site, stage of disease, and surgical urgency (elective or emergency resection), while stratifying by study interval. Adjustments were performed for these covariates, as they have been previously associated with both postoperative mortality 25 and use of emergency surgery 3 so are assumed to be confounders of the relationship between them. The missing stage information were imputed as described above.
Statistical analyses were performed using STATA version 16, (StataCorp, College Station, TX, USA). Statistical significance was set at P < 0.05.

Emergency resections
Out of 45 397 patients treated, a total of 24 828 and 17 052 major resections were included for patients diagnosed with colonic cancer between 2005 and 2016 in Denmark and Yorkshire respectively ( Table 1) The use of emergency resections increased with increasing age and stage of disease in both Denmark and Yorkshire ( Table 2). Patients with a left-sided tumour were significantly less likely to have received an emergency resection than those with a right-sided tumour in Denmark (adjusted RR 0.

Tumour site and use of stents
Use of emergency resection decreased in both patients with rightand left-sided tumours in Denmark; however, the decrease was significant in the left-sided tumours (P = 0.520 and P = 0.007 respectively; Table 3 and Fig. 1). In Yorkshire, no significant change in use of emergency resection was observed in patients with left-sided tumours (P = 0.327), but an increase was seen in those with right-sided tumours (P < 0.001).
The observed proportion of patients receiving a stent procedure increased for patients with a left-sided tumour in Denmark from 6.1 per cent in 2005-2007 to 10.5 per cent in 2014-2016, while remaining less than 2 per cent for patients with a right-sided tumour (Fig. 1). A small increase in the proportion of patients receiving a stent procedure was observed for left-sided Yorkshire patients (0.7 per cent to 1.4 per cent).

Postoperative mortality
Thirty-day postoperative mortality for all patients undergoing an elective or emergency resection decreased over time in both Denmark (7.7 per cent to 3.0 per cent) and Yorkshire (7.1 per cent to 3.3 per cent). For patients who underwent an emergency resection, a larger decrease in 30-day mortality was observed over the study interval in Yorkshire (16.2 per cent to 7.7 per

Discussion
This study has identified differences in the use of emergency resection for patients with colonic cancer between Denmark and Yorkshire. Postoperative mortality in the earliest interval of the study for all patients was higher in Denmark than in Yorkshire. The substantial reduction in use of emergency resections in Denmark over the study interval coincided with a decrease in the overall 30-day mortality rate that was subsequently lower than that in Yorkshire by the latest interval of the study.
The high 30-day mortality in Danish patients undergoing emergency resection compared with those in Yorkshire, indicates that only the patients with a higher morbidity are the undergoing an emergency resection. When the models were stratified by study interval, the risk of postoperative mortality increased over time for Danish patients undergoing an emergency resection compared with those undergoing an elective resection. This provides some evidence in the shifting of Danish patients who would have undergone emergency resection to an elective resection, for those with potentially curable disease; however, further work to investigate whether this is indeed the case is required.
Differential use of SEMs as a bridge to surgery may explain the large decrease observed in the proportion of emergency resection performed in Denmark but not in Yorkshire. Danish guidelines in 2009/2010 recommended treatment with SEMs for obstruction in left-sided tumours without suspicion of perforation where possible, as a bridge to surgery 17 . Some evidence of this was observed in the present study as patients with left-sided tumours were less likely to undergo an emergency resection than those with a right-sided tumour and we observed an increase in the use of stents after 2005-2007; however, the guidelines also highlight the demanding technical skill and set-up for this technique needed in the emergency situation and that stenting should only be performed in departments with the necessary expertise. As a result, the use of SEMs increased from the early 2010s but is likely to vary by surgical department. This, and different interpretation of the existing literature may explain the wide variation of emergency surgery by treating hospital observed in the later interval of this study.
Colonic stents were also considered as a bridge to surgery in the English 2011 NICE guidelines 26 . This was updated in 2014 to state that it should be explained to patients (or family members) that the obstruction can be managed initially either by emergency surgery or colonic stent with no clear evidence that one is better than the other, and/or patients be given the chance to participate in a randomized clinical trial comparing the two treatments 27 . The observed variation in emergency surgery between Yorkshire hospitals seen in this study is also likely to reflect this. Although current evidence may indicate bridge to surgery as the preferred treatment in some cases [28][29][30] , the differences between the Danish and English guidelines and the timing of their releases, together with the evidence available at the time of study may explain the differences observed here. In this study, an increase in the use of stents in Yorkshire was not observed, and patients with a left-sided tumours were not less likely to receive an emergency resection. However, further investigation on the uptake of SEMs in both populations is required to determine the contribution to the amount of emergency surgery performed.
While the total number of emergency resections in Yorkshire remained relatively stable the number of elective resections decreased, resulting in a statistically significant increase in the emergency resection proportion. The English screening programme began in July 2006 but by October 2008 not all of the Yorkshire region had achieved complete roll out 31 . National coverage across all English regions was achieved in 2010. Excluding screened patients here could have impacted the number of elective cases in later intervals, as the population 'at-risk' of diagnosis (and hence resection) will have reduced if diagnosed earlier through screening. The Danish screening programme was introduced at a later time, in March 2014 32 . Additionally, Denmark had a higher resection rate than Yorkshire over the same study time interval 33 . This, in addition to a higher incidence rate in Denmark 34 , may also explain why the overall number of resected patients in this study was much higher in Denmark when the overall populations are of equal size. Comparative data on patient morbidity were not available in this study. Factors such as socioeconomic deprivation, co-morbidities, ASA grade, and histopathological profile have been reported to be associated with emergency patients 1,7,35 . Data on these characteristics would have allowed us to account for their influence on the differences in the use of emergency surgery reported in this study. The potential for these factors to contribute to the differences in practice observed here, should be considered in studies that look to confirm these results.
It is important to consider the limitations of the data used here when considering the findings of this study. The main weakness of the study concerns the definition of emergency resection and how directly comparable the proportions calculated in the two populations are. Emergency resection in Denmark is recorded by surgeons using a surgical proforma to indicate the urgency of the procedure. Although no formal definition of what constitutes an emergency procedure, guidance states emergency patients are those where indication for surgery is suspected due to either ileus, perforation, bleeding, or otherwise and usually occurs within 36 h of admission. This information was not directly available in the Yorkshire data, so we used surgery within 2 days following an emergency admission. We believe this to be an acceptable way to define an emergency resection, as we compared the estimated proportion in England using our definition, with results from the National Bowel Cancer Audit (NBOCA) over the same time period 36 . The NBOCA used the National Confidential Enquiry into Patient Outcomes and Death (NCEPOD) to classify resections in urgent and emergency procedures 37 . For April 2014 to March 2016, the proportion of English resections we estimated to be an emergency resection was 16.0 per cent, and the proportion of resections in NBOCA classified as urgent and emergency resections was also estimated to be 16.0 per cent. Although the definitions of emergency surgery differ between Denmark and Yorkshire, they are defined in a way that are robust to changes over time, and therefore the observed trend observed in this study should remain valid within both populations.
This study has shown both Denmark and Yorkshire have substantially reduced mortality after surgery in both elective and emergency resections. The larger decrease in overall postoperative mortality in Denmark has coincided with a reduction in the use of emergency resection and increase in stenting procedures, following a policy of converting potential emergency resections into elective resections implemented in some departments. Identifying and addressing the reasons for an increase in emergency surgery in Yorkshire could contribute to a further lowering of postoperative mortality.

Funding
This work was funded by Yorkshire Cancer Research (award reference number L394). Prof. Quirke holds an NIHR Senior Investigator award. The UK Colorectal Cancer Intelligence Hub is supported by Cancer Research UK (C23434/A23706).