Differences in outcomes of oesophageal and gastric cancer surgery across Europe

Transcription

1 Original article Differences in outcomes of oesophageal and gastric cancer surgery across Europe J. L. Dikken 1,3, J. W. van Sandick 4,W.H.Allum 6, J. Johansson 8, L. S. Jensen 9, H. Putter 2, V. H. Coupland 7,M.W.J.M.Wouters 1,4, V. E. P. Lemmens 5 and C. J. H. van de Velde 1* Departments of 1 Surgery and 2 Medical Statistics, Leiden University Medical Center, Leiden, Departments of 3 Radiotherapy and 4 Surgery, the Netherlands Cancer Institute Antoni van Leeuwenhoek Hospital, Amsterdam, and 5 Comprehensive Cancer Centre South, Eindhoven, the Netherlands, 6 Department of Surgery, The Royal Marsden NHS Foundation Trust, and 7 King s College London, Thames Cancer Registry, London, UK, 8 Department of Surgery, Lund University Hospital, Lund, Sweden, and 9 Department of Surgery, Aarhus University Hospital, Aarhus, Denmark Correspondence to: Professor C. J. H. van de Velde, Department of Surgery, K6-R, Leiden University Medical Centre, PO Box 9600, 2300 RC Leiden, The Netherlands ( c.j.h.van_de_velde@lumc.nl) Background: In several European countries, centralization of oesophagogastric cancer surgery has been realized and clinical audits initiated. The present study was designed to evaluate differences in resection rates, outcomes and annual hospital volumes between these countries, and to analyse the relationship between hospital volume and outcomes. Methods: National data were obtained from cancer registries or clinical audits in the Netherlands, Sweden, Denmark and England. Differences in outcomes were analysed between countries and between hospital volume categories, adjusting for available case-mix factors. Results: Between 2004 and 2009, oesophagectomies and 9010 gastrectomies were registered. Resection rates in England were 18 2 and21 6 per cent for oesophageal and gastric cancer respectively, compared with and per cent in the Netherlands and Denmark (P < 0 001). The adjusted 30-day mortality rate after oesophagectomy was lowest in Sweden (1 9 percent). After gastrectomy, the adjusted 30-day mortality rate was significantly higher in the Netherlands (6 9 percent) than in Sweden (3 5 percent; P = 0 017) and Denmark (4 3 percent; P = 0 029). Increasing hospital volume was associated with a lower 30-day mortality rate after oesophagectomy (odds ratio 0 55 (95 per cent confidence interval 0 42 to 0 72) for at least 41 versus 1 10 procedures per year) and gastrectomy (odds ratio 0 64 (0 41 to 0 99) for at least 21 versus 1 10 procedures per year). Conclusion: Hospitals performing larger numbers of oesophagogastric cancer resections had a lower 30-day mortality rate. Differences in outcomes between several European countries could not be explained by differences in hospital volumes. To understand these differences in outcomes and resection rates, with reliable case-mix adjustments, a uniform European upper gastrointestinal cancer audit with recording of standardized data is warranted. * Co-authors can be found under the heading Contributors Presented to the Annual Meeting of the Dutch Surgical Society, Veldhoven, the Netherlands, May 2012, and the European Society of Surgical Oncology 32nd Congress, Valencia, Spain, September 2012; published in abstract form as Eur J Surg Oncol 2012; 38: 765 Paper accepted 6 September 2012 Published online in Wiley Online Library ( DOI: /bjs.8966 Introduction Quality assurance is increasingly being acknowledged as a crucial factor for improvement of care for patients with oesophageal and gastric cancer. In Europe, the average 5-year survival rate is 11 per cent for oesophageal cancer and 25 per cent for gastric cancer, but variation between and within countries is considerable 1. The reasons for these variations are difficult to assess. In some countries there are nationally sponsored cancer registries, whereas 2012 British Journal of Surgery Society Ltd British Journal of Surgery 2013; 100: 83 94

2 84 J. L. Dikken, J. W. van Sandick, W. H. Allum, J. Johansson, L. S. Jensen, H. Putter et al. Table 1 Characteristics of participating countries and available data sets The Netherlands Sweden Denmark England Inhabitants ( 10 6 ) Incidence of oesophageal cancer (per )* M F Incidence of gastric cancer (per )* M F Centralization of surgery Oesophagectomy 2006: 10/year No 2003: 5 centres 2001: 40/year 2008: 4 centres Gastrectomy No No 2003: 5 centres 2001: 60/year 2008: 4 centres Registry Registry used Netherlands Cancer Registry National Quality Registry of Oesophageal and Gastric Cancer National Database of Oesophagogastric Cancer, National Pathology Registry, National Registry of Patients, Danish Civil Registration System English Cancer Registries Registry type Cancer registry Clinical audit Clinical audit Cancer registry Registry active since Multiple sources Registry organization 8 regions, 1 central database 6 regions, 1 central database 1 central database 8 regions, 1 central database Data collection Trained registrars Trained doctors and Surgeons treating the Multiple sources nurses patients Years of diagnosis in data set Follow-up until January 2010 April 2011 January 2011 December 2009 Case ascertainment Nationwide Partial Nationwide Nationwide Data availability Patient age and sex Co-morbidity (Charlson/ASA) / /+ +/+ / Tumour location (O/OGJ/S) Tumour histology (AC/SCC/other) TNM stage No. of lymph nodes evaluated Surgery (oesophagectomy/gastrectomy) Surgery hospital (Neo)adjuvant therapy day postoperative mortality In-hospital mortality + 2-year survival after surgery *World Standard Ratio (Karim-Kos et al. 2 ). Minimum annual hospital volume. Case ascertainment was incomplete in certain regions in Sweden; regions with case ascertainment below 90 per cent were excluded. +, Yes;, no. ASA, American Society of Anesthesiologists; O, oesophageal; OGJ, oesophagogastric junction; S, stomach; A, adenocarcinoma; SCC, squamous cell carcinoma; TNM, tumour node metastasis. others have established clinical audits. Data recorded are variable and there are differences in data interpretation. Outcome comparisons are inevitably limited. One of the key elements in any comparison is data completeness, in order to minimize bias. Programmes and processes have been established in the Netherlands, Sweden, Denmark and England that are designed to achieve as comprehensive data collection as possible with the aim of assuring the quality of treatment of oesophageal and gastric cancer. The present study was undertaken to evaluate differences in annual hospital volumes, resection rates and treatment outcomes in these four countries, and to determine where improvements can be made to allow better comparisons between countries.

3 Outcomes of oesophageal and gastric cancer surgery across Europe 85 Methods National data were obtained from cancer registries in the Netherlands and England, and from clinical audits in Sweden and Denmark (Table 1). The cancer registries from the Netherlands and England, and the audit from Denmark, provided national coverage of all patients with a diagnosis of oesophagogastric cancer. Only patients who underwent surgery were included in the Swedish audit, and so no resection rates could be calculated for Sweden. In several Swedish regions, it was apparent that not all patients who had undergone surgical resection had been registered. To reduce the chance of selection bias, only Swedish regions with a case ascertainment above 90 per cent were included. Applying the same criteria, detailed data from patients in the voluntary UK National Oesophago-Gastric Cancer Audit (NOGCA) 3 were not included (case ascertainment 71 per cent). Resection rates were calculated in the cohort of patients with a diagnosis of oesophageal or gastric cancer between 2004 and Postoperative mortality, survival and annual hospital volumes were calculated in the cohort of patients who underwent surgical resection between 2004 and Demographic information was available in all data sets, but co-morbidities were not registered uniformly and could therefore not be used for case-mix adjustments. Tumour location and histology based on the International Classification of Diseases for Oncology (ICD-O) were available in all data sets 4. Tumour location was defined as oesophagus (ICD-O C ) or stomach (ICD-O C ). Staging was performed according to the sixth edition of the International Union Against Cancer tumour node metastasis (TNM) classification 5. Information on TNM stage was not available for the English data, as stage was not recorded routinely by the English registries. The hospital (in England these were described by the registry as trusts that can be responsible for several individual hospitals) where the operation was performed was available in all data sets. Annual hospital volume was defined separately for oesophagectomy and gastrectomy as the number of resections per hospital in each calendar year. Volume categories were defined according to the distribution of number of resections among hospitals (Fig. S1, supporting information). Statistical analysis Data regarding oesophagectomies and gastrectomies were analysed separately. Differences in patient characteristics, resection rates and annual hospital volumes between countries were analysed with the χ 2 test. Postoperative mortality was defined as death from any cause within 30 days after surgery. Differences in 30-day mortality between countries were analysed by means of generalized estimating equations, adjusting for available case-mix factors (sex, age, morphology) and clustering of patients within hospitals using a random hospital effect model 6. Two-year overall survival after surgery was chosen as the long-term outcome because of the relatively short follow-up period owing to the recent nature of the data, and calculated from the day of surgery until death from any cause (event) or alive at last follow-up (censored). Details Diagnosis of oesophageal or gastric cancer The Netherlands ( ) n = Denmark ( ) n = 4283 England ( ) n = Calculation of resection rates Resection for oesophageal or gastric cancer The Netherlands ( ) n = 5791 Sweden ( ) n = 653 Denmark ( ) n = 1420 England ( ) n = Resection for non-metastatic oesophageal or gastric cancer The Netherlands ( ) n = 5153 Sweden ( ) n = 606 Denmark ( ) n = 1334 Calculation of hospital volumes, adjusted 30-day mortality, unadjusted 2-year survival Excluded The Netherlands, Sweden, Denmark: M1 disease n = 771 England: all patients n = Calculation of adjusted 2-year survival Fig. 1 Study profile. *Swedish data include only regions with over 90 per cent case ascertainment; case ascertainment was over 99 per cent in the other countries. No data on tumour node metastasis stage were extracted from the English cancer registries

4 86 J. L. Dikken, J. W. van Sandick, W. H. Allum, J. Johansson, L. S. Jensen, H. Putter et al. of the cause of death were not available. Unadjusted 2- year overall survival rates for each country were calculated by Kaplan Meier analysis. Adjusted differences in 2-year overall survival between countries were analysed by Cox regression, adjusting for case-mix factors as categorical co-variables (sex, age, morphology, stage) and clustering of patients within hospitals. English patients were excluded from the adjusted 2-year survival analyses as stage data were not available. Differences in outcomes between hospital volume categories were evaluated in the same way as differences in outcomes between countries, including the adjustment for clustering of patients within hospitals. An interaction analysis was performed between country and annual Table 2 Characteristics of patients who underwent oesophagectomy and 9010 who underwent gastrectomy for cancer The Netherlands Sweden Denmark England P Oesophagectomy Sex M 2179 (77 3) 185 (80 1) 699 (78 4) 5295 (76 6) F 640 (22 7) 46 (19 9) 193 (21 6) 1617 (23 4) Age (years) < (34 5) 73 (31 6) 299 (33 5) 2171 (31 4) (55 6) 133 (57 6) 530 (59 4) 4001 (57 9) > (9 9) 25 (10 8) 63 (7 1) 740 (10 7) Mean Median Histology < Adenocarcinoma 2141 (75 9) 162 (70 1) 637 (71 4) 5483 (79 3) SCC 615 (21 8) 42 (18 2) 201 (22 5) 1190 (17 2) Other carcinoma 63 (2 2) 27 (11 7) 54 (6 1) 239 (3 5) TNM stage* < (0 4) 15 (6 5) 20 (2 2) I 446 (15 8) 18 (7 8) 67 (7 5) II 977 (34 7) 101 (43 7) 381 (42 7) III 912 (32 4) 71 (30 7) 334 (37 4) IV 339 (12 0) 12 (5 2) 37 (4 1) Unknown 135 (4 8) 14 (6 1) 53 (5 9) 6912 (100) Mean Median II II II Gastrectomy Sex < M 1838 (61 8) 241 (57 1) 305 (57 8) 3304 (64 9) F 1134 (38 2) 181 (42 9) 223 (42 2) 1784 (35 1) Age (years) < < (20 2) 67 (15 9) 141 (26 7) 820 (16 1) (47 4) 193 (45 7) 267 (50 6) 2585 (50 8) > (32 4) 162 (38 4) 120 (22 7) 1683 (33 1) Mean Median Histology < Adenocarcinoma 2929 (98 6) 396 (93 8) 502 (95 1) 4879 (95 9) Other carcinoma 43 (1 4) 26 (6 2) 26 (4 9) 209 (4 1) TNM stage* < (0 5) 13 (3 1) 6 (1 1) I 1015 (34 2) 110 (26 1) 83 (15 7) II 695 (23 4) 105 (24 9) 109 (20 6) III 666 (22 4) 111 (26 3) 159 (30 1) IV 508 (17 1) 54 (12 8) 40 (7 6) Unknown 73 (2 5) 29 (6 9) 131 (24 8) 5088 (100) Mean Median II II III Values in parentheses are percentages. *Sixth edition of the tumour node metastasis (TNM) classification. Mainly T4 N+ M0 and T1 3 N3; some palliative resections. Calculated by excluding unknown stage and considering stage group as a continuous variable. SCC, squamous cell carcinoma. χ 2 test; England excluded.

5 Outcomes of oesophageal and gastric cancer surgery across Europe day mortality (%) day mortality (%) The Netherlands a Oesophagectomy Sweden Denmark England 0 The Netherlands Sweden Denmark England b Gastrectomy Fig. 2 Postoperative 30-day mortality after a oesophagectomy and b gastrectomy, adjusted for sex, age and histology, in the Netherlands, Sweden, Denmark and England. a *P < versus England; b *P < versus the Netherlands (generalized estimated equations) Table 3 Differences in postoperative 30-day mortality and 2-year survival between countries Oesophagectomy Gastrectomy 30-day mortality (%)* 2-year survival (%) 30-day mortality (%)* 2-year survival (%) Absolute adjusted The Netherlands 4 6 (3 3, 5 9) 56 8 (54 5, 59 3) 6 9 (5 1, 8 8) 59 0 (56 8, 61 3) Sweden 1 9 (0 0, 3 8) 61 0 (54 6, 68 0) 3 5 (1 5, 5 6) 59 0 (54 2, 64 3) Denmark 4 6 (2 4, 6 8) 58 2 (54 8, 61 9) 4 3 (2 4, 6 2) 62 8 (58 5, 67 5) England 5 8 (4 7, 6 9) 5 9 (4 3, 7 4) Absolute unadjusted The Netherlands 52 4 (50 2, 54 6) 51 9 (49 9, 53 9) Sweden 56 7 (50 0, 63 4) 51 7 (46 8, 56 6) Denmark 53 3 (50 0, 56 6) 53 7 (49 4, 58 0) England 54 4 (53 2, 55 6) 56 3 (54 9, 57 7) Adjusted odds or hazards ratio Country The Netherlands 1 00 (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) Sweden 0 40 (0 14, 1 16) 0 93 (0 75, 1 15) 0 50 (0 28, 0 88) 0 97 (0 85, 1 11) Denmark 1 00 (0 60, 1 69) 0 96 (0 80, 1 15) 0 60 (0 38, 0 95) 0 89 (0 80, 1 00) England 1 28 (0 96, 1 72) 0 84 (0 65, 1 07) Sex M 1 00 (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) F 0 75 (0 61, 0 93) 0 78 (0 69, 0 89) 0 81 (0 68, 0 96) 0 93 (0 83, 1 03) Age (years) < (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) (1 44, 2 25) 1 40 (1 27, 1 55) 2 58 (1 79, 3 73) 1 29 (1 08, 1 54) > (2 96, 5 05) 1 89 (1 60, 2 24) 5 98 (1 09, 8 75) 1 94 (1 65, 2 29) Histology Adenocarcinoma 1 00 (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) SCC 1 44 (1 16, 1 79) 1 27 (1 14, 1 41) Other carcinoma 1 33 (0 84, 2 11) 1 46 (1 02, 2 09) 1 57 (1 01, 2 45) 0 97 (0 66, 1 43) TNM stage (0 29, 1 15) 0 52 (0 20, 1 37) I 1 00 (reference) 1 00 (reference) II 1 95 (1 46, 2 60) 2 10 (1 81, 2 42) III 3 68 (2 73, 4 95) 3 81 (3 29, 4 41) IV 8 21 (4 42, 15.25) 6 40 (5 37, 7 62) Unknown 1 73 (0 99, 3 02) 2 06 (1 60, 2 64) Values in parentheses are 95 per cent confidence intervals. *The Netherlands, Sweden, Denmark and England; for 30-day mortality, adjustments were made for sex, age, histology and clustering of patients within hospitals. The Netherlands, Sweden and Denmark; for 2-year survival, adjustments were made for sex, age, histology, stage and clustering of patients within hospitals. Odds ratios are shown for mortality and hazard ratios for survival. Sixth edition of the tumour node metastasis (TNM) classification. SCC, squamous cell carcinoma.

6 88 J. L. Dikken, J. W. van Sandick, W. H. Allum, J. Johansson, L. S. Jensen, H. Putter et al Oesophagectomies (%) Gastrectomies (%) The Netherlands Sweden Denmark England The Netherlands Sweden Denmark England a Oesophagectomy b Gastrectomy Fig. 3 Annual hospital volumes (procedures per year) for a oesophagectomy and b gastrectomy in the Netherlands, Sweden, Denmark and England. a,b P < (χ 2 test) day mortality (%) day mortality (%) Annual hospital volume a Oesophagectomy 0 b Gastrectomy Annual hospital volume Fig. 4 Postoperative 30-day mortality after a oesophagectomy and b gastrectomy, adjusted for sex, age, and histology, by annual hospital volume (procedures per year). *P < versus 1 10 (generalized estimated equations) hospital volume. Annual hospital volume was analysed both as a categorical and as a linear variable. Statistical analyses were performed with SPSS version (IBM, Armonk, New York, USA) and R version ( Results The Swedish registries with more than 90 per cent case ascertainment were in Uppsala Örebro ( ), Norra ( ), Sydöstra ( ) and Stockholm Gotland ( ). Not all countries had complete data for each year (Fig. 1). Between January 2004 and December 2009, patients with a diagnosis of oesophageal or gastric cancer were registered in the Netherlands, Denmark or England. Resection rates were similar in the Netherlands and Denmark: per cent for oesophageal cancer and per cent for gastric cancer. Resection rates in England were significantly lower: 18 2 per cent for oesophageal cancer and 21 6 per cent for gastric cancer (both P < 0 001). Characteristics of patients who underwent resection Between 2004 and 2009, patients underwent oesophagectomy or gastrectomy for cancer (Table 2).

7 Outcomes of oesophageal and gastric cancer surgery across Europe 89 Table 4 Multivariable analysis of effect of annual hospital volume on 30-day mortality and 2-year survival Oesophagectomy Gastrectomy 30-day mortality (%)* 2-year survival (%) 30-day mortality (%)* 2-year survival (%) Odds ratio Hazard ratio Odds ratio Hazard ratio Annual hospital volume (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) (0 61, 1 11) 0 92 (0 78, 1 08) 0 84 (0 67, 1 05) 1 04 (0 93, 1 15) or (0 50, 0 93) 0 84 (0 63, 1 11) 0 64 (0 41, 0 99) 1 01 (0 84, 1 22) (0 39, 0 85) 0 77 (0 63, 0 94) (0 42, 0 72) 0 79 (0 66, 0 96) P for trend# < Sex M 1 00 (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) F 0 77 (0 62, 0 95) 0 78 (0 69, 0 90) 0 80 (0 67, 0 95) 0 92 (0 83, 1 02) Age (years) < (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) (1 45, 2 28) 1 40 (1 27, 1 55) 2 58 (1 78, 3 72) 1 30 (1 09, 1 55) > (3 06, 5 21) 1 87 (1 58, 2 23) 5 88 (4 04, 8 58) 1 96 (1 67, 2 30) Histology Adenocarcinoma 1 00 (reference) 1 00 (reference) 1 00 (reference) 1 00 (reference) SCC 1 44 (1 15, 1 79) 1 29 (1 15, 1 44) Other carcinoma 1 28 (0 81, 2 04) 1 45 (1 03, 2 05) 1 50 (0 96, 2 33) 0 97 (0 65, 1 43) TNM stage (0 29, 1 14) 0 52 (0 20, 1 35) I 1 00 (reference) 1 00 (reference) II 1 96 (1 46, 2 62) 2 08 (1 80, 2 40) III 3 71 (2 74, 5 04) 3 75 (3 24, 4 35) IV 8 13 (4 39, 15 08) 6 38 (5 34, 7 62) Unknown 1 77 (1 01, 3 11) 1 94 (1 51, 2 48) Values in parentheses are 95 per cent confidence intervals. *The Netherlands, Sweden, Denmark and England; for 30-day mortality, adjustments were made for sex, age, histology and clustering of patients within hospitals. The Netherlands, Sweden and Denmark; for 2-year survival, adjustments were made for sex, age, histology, stage and clustering of patients within hospitals. Oesophagectomy; gastrectomy. Sixth edition of the tumour node metastasis (TNM) classification. SCC, squamous cell carcinoma. #Cox regression. Median age was 64 years for all patients who underwent oesophagectomy and 71 years for those who underwent gastrectomy. The percentage of patients aged over 75 years undergoing resection was lowest in Denmark, 7 1 per cent for oesophagectomy and 22 7 per cent for gastrectomy, compared with per cent for oesophagectomy and per cent for gastrectomy in the other countries. The highest proportion of patients with stage I tumours (oesophagectomy 15 8 per cent, gastrectomy 34 2 per cent) and the highest proportion of stage IV disease (oesophagectomy 12 0 per cent, gastrectomy 17 1 per cent) were recorded in the Netherlands. Median follow-up for all patients was 37 months. Exact in-hospital mortality data were not available from all four data sources. In all countries, the postoperative 30-day mortality rate was lower after oesophagectomy (4 6 per cent) than after gastrectomy (6 7 per cent), but variation between countries was considerable. The adjusted 30-day mortality rate after oesophagectomy was lowest in Sweden (1 9 per cent) and highest in England (5 8 per cent) (P = 0 028) (Fig. 2a, Table 3). Differences between other countries were not significant. After gastrectomy, the adjusted 30-day mortality rate in the Netherlands (6 9 per cent) was significantly higher than in Sweden (3 5 per cent; P = 0 017) and Denmark (4 3 per cent; P = 0 029) (Fig. 2b). Unadjusted 2-year overall survival estimates were not significantly different between countries, except for 2-year survival after gastrectomy between the Netherlands and England (51 9 versus 56 3 per cent; P < 0 001) (Fig. S2, supporting information). Adjusted 2-year survival rates were not significantly different between the Netherlands, Sweden and Denmark, in either resection group (Table 3; Fig. S3, supporting information) Annual hospital volumes for oesophagectomy were higher than for gastrectomy (Fig. S1, supporting information). The variation between countries is shown in Fig. 3. In Denmark, 65 6 per cent of oesophagectomies were performed in hospitals with an annual volume above 30 procedures per year, whereas a similar proportion

8 90 J. L. Dikken, J. W. van Sandick, W. H. Allum, J. Johansson, L. S. Jensen, H. Putter et al. (63 6 per cent) was performed in Sweden in hospitals with an annual volume of fewer than 11 per year. In Denmark, 58 9 per cent of all gastrectomies for cancer were performed in hospitals with an annual volume above 20 per year, whereas in the Netherlands and Sweden 75 2and 76 1 per cent of gastric resections respectively were performed in hospitals with an annual volume of less than 11 per year. In England, 68 9 per cent of gastrectomies were performed in hospitals with an annual volume of less than 21 per year. Increasing hospital volume was associated with significantly lower postoperative mortality rates, after both oesophagectomy and gastrectomy (Fig. 4, Table 4). The adjusted 30-day mortality rate after oesophagectomy in hospitals with an annual volume of at least 41 per year was lower than in hospitals with an annual volume of less than 11 per year (4 3 versus 7 2 per cent; P < 0 001). The adjusted 30-day mortality rate after gastrectomy in hospitals with an annual volume of at least 21 per year was also lower than in hospitals with an annual volume of less than 11 per year (4 4 versus 6 7 per cent; P = 0 047). Testing for interaction between country and hospital volume category revealed a significant interaction regarding postoperative 30-day mortality after oesophagectomy, which was the result of a stronger volume outcome relation in Denmark than in the other countries (data not shown). No such interaction was found for gastrectomy. High hospital volume was also significantly associated with better 2-year survival after oesophagectomy, with a hazard ratio of 0 79 (95 per cent confidence interval 0 66 to 0 96) for the highest-volume group (at least 41 per year) compared with the lowest-volume group (1 10 per year). Following gastrectomy, there was no significant association between hospital volume and 2-year survival (Table 4; Fig. S4, supporting information) and no interaction between country and hospital volume category regarding 2-year survival. Discussion This study has shown variations in annual hospital volumes for oesophagectomy and gastrectomy across four European countries. The hospital volumes were highest in Denmark. Resection rates were similar in the Netherlands and Denmark, but considerably lower in England. The postoperative 30-day mortality rate was lowest in Sweden, after both oesophagectomy and gastrectomy, and the 30-day mortality rate after gastrectomy was significantly higher in the Netherlands than in Sweden and Denmark. Larger numbers of stage I and stage IV oesophageal and gastric cancers were resected in the Netherlands than in the other countries. Increasing hospital volume was associated with lower postoperative mortality after both oesophagectomy and gastrectomy. Two-year adjusted survival after surgery was similar in each country, with longer overall survival after oesophagectomy compared with gastrectomy. Studies on outcomes after cancer surgery are commonly based on data from clinical trials or patient cohorts from specialized surgical centres. Owing to selection, these series may not reflect practice in general and cannot be used to compare outcomes between countries. Population-based studies, as performed by EUROCARE, provide insight into the differences in mortality and survival patterns between countries 1. In the EUROCARE framework, however, registry data may be incomplete for some countries and recent information is not available. The methodology was intended for incidence and survival trend analyses, not for monitoring clinical practice or providing feedback to individual healthcare providers. Nationwide clinical audits, as currently performed in the UK, Sweden, Denmark and the Netherlands, provide detailed information on patient, tumour, treatment and hospital characteristics, and data are quickly available for comparative analyses. A disadvantage of clinical audits is that data are reported by the healthcare provider and are therefore not always complete. In contrast, cancer registries mostly include all available patients, but the information captured is less detailed. In the present study, information on patient co-morbidities was missing from the Dutch and English data set, and tumour stage from the English data set. This lack of this information may have biased outcome and partly explain some of the differences. In the present study, resection rates for both oesophageal and gastric cancer were lower in England than in the Netherlands and Denmark (and not available in Sweden). The UK NOGCA has confirmed a steady reduction in resection rates over the past decade, describing rates of curative resection for oesophageal junctional and gastric cancer of 33 and 31 per cent respectively in 1998, decreasing to 24 and 23 per cent in This has been attributed to improved preoperative staging and multidisciplinary management, thereby better selecting patients for surgery 8. Comparison of resection rates is also confounded by differences in clinical practice, but with the present data sets no conclusions can be drawn on which country has the optimal resection rate. This should be addressed in future studies if adequate information on preoperative staging is included. A Dutch study published in 2001 showed lower postoperative mortality rates after oesophagectomies in high-volume hospitals, and as of 2006 oesophagectomy in

9 Outcomes of oesophageal and gastric cancer surgery across Europe 91 the Netherlands was centralized with a minimum annual volume of ten per year 9. As of 2011, this was increased to 20 per year. Over the study period, there was no minimum volume standard for gastrectomy in the Netherlands, but gastrectomies will be centralized from In addition, a national oesophagogastric cancer audit was started in This may answer why the resection rate in stage IV disease was higher than elsewhere, perhaps reflecting practice in smaller hospitals where preoperative assessment may be less robust. In Sweden, a national oesophagogastric cancer audit was initiated in Both oesophagectomies and gastrectomies were performed in low volumes, but recently Sweden has also started to centralize upper gastrointestinal surgery. A nationwide oesophagogastric cancer registry has been initiated in Denmark. Upper gastrointestinal surgery was restricted to five centres in 2003, reducing to four centres in This was accompanied by reduced postoperative mortality after gastrectomy and an increase in the number of evaluated lymph nodes, often used as a quality indicator in gastric cancer surgery 12. In the present study, hospital volumes in Denmark were higher than in any other country, with the majority of oesophagectomies being performed in hospitals with a volume of over 40 per year. In England, Wales and Northern Ireland, a National Health Services Cancer Plan became effective in In this plan, recommendations were made to centralize oesophagogastric cancer surgery to centres with a population base of between one and two million, establish specialist treatment teams, and audit all steps in oesophagogastric cancer care 14. By 2008 and 2009, 82 per cent of oesophageal and gastric cancer resections were done in 41 designated centres, with 63 per cent of oesophagectomies and 65 per cent of gastrectomies being performed in high-volume centres (at least 50 resections per year) 15. Centralization of surgery is not unique to Europe. A recent US study described centralization of several surgical procedures, including oesophagectomy for cancer, that resulted in a decrease in postoperative mortality 16. Owing to its population-based nature, the present study provides an accurate comparison of postoperative mortality and long-term survival after oesophagectomy and gastrectomy between several countries in Europe. However, the variability in the recorded data and missing information on patient co-morbidities, multimodality therapy and cause-specific survival makes it impossible to explain properly the differences between countries simply in terms of annual hospital volumes. Sweden had lower postoperative mortality rates than the other countries, even after adjustment for case mix, without performing surgery in high-volume institutions. High-quality healthcare with nationwide quality assurance programmes might have contributed to the these results 17. Differences in selecting patients for surgery between Sweden and the other countries could be a factor and the inclusion only of regions in Sweden with high case ascertainment may also have biased the findings. On the contrary, the postoperative mortality rate after gastrectomy in the Netherlands was high. This might be explained by the absence of a quality assurance programme during the studied period for gastric cancer surgery in the Netherlands. Differences in unadjusted 2-year survival rates between countries should be interpreted with care, as tumour stage distributions in the group of patients who underwent surgical resection might have differed between countries. The relationship between annual hospital volume and postoperative mortality after oesophagectomy and gastrectomy has been investigated extensively 16,18. For oesophagectomy, most studies have found a benefit for high-volume surgery 19. Results regarding hospital volumes for gastrectomies are less uniform, some finding no effect on postoperative mortality Patient numbers in these studies, however, were relatively small (below 5000) compared with those in which a benefit for gastrectomies in high-volume centres was found (up to ) The evidence on hospital volume in relation to long-term survival is limited. Among four studies of oesophagectomy, two were positive 28 31, and five of seven gastrectomy studies were positive 27, In the present study, a significant relationship between annual hospital volume and postoperative mortality was found for both oesophagectomy and gastrectomy. Furthermore, increasing hospital volume for oesophagectomy was associated with improved long-term survival. No such relationship was found for gastrectomy; this might be explained by the low threshold of what was considered high-volume surgery (21or more procedures per year). It could be argued that in the present study individual surgeon volumes should have been analysed as well as hospital volume. Quality of care and outcomes, however, are the result of collaboration between different professionals, including surgeons, anaesthetists, intensive care unit staff and nursing staff. All these disciplines contribute to outcomes 38. The role of the individual surgeon is important, but attribution of results solely to the surgeon is open to misinterpretation. Using hospital volume as the only basis for determining outcome quality has also been criticized 24. There can be low-volume hospitals with excellent outcomes and vice versa. Outcomebased referral avoids this problem, by selecting centres of

10 92 J. L. Dikken, J. W. van Sandick, W. H. Allum, J. Johansson, L. S. Jensen, H. Putter et al. excellence based on case mix-adjusted outcomes. It has been used to centralize oesophagectomy in one part of the Netherlands, which led to a reduction in the postoperative mortality rate from 12 to 3 per cent over a 10-year period 39. There were considerable differences between the four European countries involved in the present study regarding resection rates, postoperative 30-day mortality rates and annual hospital volumes in oesophagogastric cancer surgery. Increasing hospital volume was associated with better outcomes, but differences in outcomes between countries could not be explained by these differences in annual hospital volumes. Nationwide clinical audits aim to identify centres of excellence based on case mix-adjusted outcomes. On an international level, these audits can be used to understand differences in outcomes between countries, but require uniform definitions and registration of data. The present study highlights the need to collect standard clinical data in each country to facilitate international comparisons, analogous to the EURECCA initiative for colorectal cancer 40. A European oesophageal and gastric cancer audit could provide further insight into differences between countries, leading to better quality of care for patients with oesophageal and gastric cancer. Contributors L. G. M. van der Geest (Comprehensive Cancer Centre The Netherlands, Utrecht, The Netherlands), H. J. Larsson (Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark), A. Cats (Department of Gastroenterology, The Netherlands Cancer Institute Antoni van Leeuwenhoek Hospital, Amsterdam, The Netherlands) and M. Verheij (Department of Radiotherapy, The Netherlands Cancer Institute Antoni van Leeuwenhoek Hospital, Amsterdam, The Netherlands). Acknowledgements This study was funded by the Signalling Committee on Cancer of the Dutch Cancer Society (KWF Kankerbestrijding). The study sponsor had no role in the study design; in the collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication. Disclosure: The authors declare no conflict of interest. References 1 Sant M, Allemani C, Santaquilani M, Knijn A, Marchesi F, Capocaccia R; EUROCARE Working Group. EUROCARE-4. Survival of cancer patients diagnosed in Results and commentary. Eur J Cancer 2009; 45: Karim-Kos HE, de Vries E, Soerjomataram I, Lemmens V, Siesling S, Coebergh JW. Recent trends of cancer in Europe: a combined approach of incidence, survival and mortality for 17 cancer sites since the 1990s. Eur J Cancer 2008; 44: Association of Upper Gastrointestinal Surgeons of Great Britain and Ireland. National Oesophago-Gastric Cancer Audit Third Annual Report [accessed 5 October 2012]. 4 World Health Organization (WHO). International Classification of Diseases for Oncology (ICD-O-3) (3rd edn); adaptations/oncology/en/index.html [accessed 5 October 2012]. 5 Sobin LH, Wittekind C (eds). TNM Classification of Malignant Tumours (6th edn). Wiley-Liss: New York, Zeger SL, Liang KY. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73: NHS Information Centre. National Oesophago-Gastric Cancer Audit First Annual Report NHS Information Centre: Leeds, Lyratzopoulos G, Barbiere JM, Gajperia C, Rhodes M, Greenberg DC, Wright KA. Trends and variation in the management of oesophagogastric cancer patients: a population-based survey. BMC Health Serv Res 2009; 9: van Lanschot JJ, Hulscher JB, Buskens CJ, Tilanus HW, ten Kate FJ, Obertop H. Hospital volume and hospital mortality for esophagectomy. Cancer 2001; 91: Dutch Upper GI Cancer Audit. [accessed 5 October 2012]. 11 Jensen LS, Nielsen H, Mortensen PB, Pilegaard HK, Johnsen SP. Enforcing centralization for gastric cancer in Denmark. Eur J Surg Oncol 2010; 36(Suppl 1): S50 S Coburn NG, Swallow CJ, Kiss A, Law C. Significant regional variation in adequacy of lymph node assessment and survival in gastric cancer. Cancer 2006; 107: Department of Health. Guidance on Commissioning Cancer Services. Improving Outcomes in Upper Gastro-intestinal Cancers. The Manual; groups/dh_digitalassets/@dh/@en/documents/digitalasset/ dh_ pdf [accessed 5 October 2012]. 14 Palser TR, Cromwell DA, Hardwick RH, Riley SA, Greenaway K, Allum W et al. Re-organisation of oesophago-gastric cancer care in England: progress and remaining challenges. BMC Health Serv Res 2009; 9: National Cancer Intelligence Network (NCIN). Improving Outcomes: a Strategy for Cancer NCIN Information Supplement. [accessed 5 October 2012]. 16 Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume and operative mortality for high-risk surgery. NEngl JMed2011; 364: Socialstyrelsen. Nationella riktlinjer för bröst-, kolorektal- och prostatacancer Beslutsstöd för prioriteringar;

11 Outcomes of oesophageal and gastric cancer surgery across Europe 93 socialstyrelsen.se/lists/artikelkatalog/attachments/8947/ _ pdf [accessed 5 October 2012]. 18 Gruen RL, Pitt V, Green S, Parkhill A, Campbell D, Jolley D. The effect of provider case volume on cancer mortality: systematic review and meta-analysis. CA Cancer J Clin 2009; 59: Courrech Staal EF, Wouters MW, Boot H, Tollenaar RA, van Sandick JW. Quality-of-care indicators for oesophageal cancer surgery: a review. Eur J Surg Oncol 2010; 36: Baré M, Cabrol J, Real J, Navarro G, Campo R, Pericay C, et al. In-hospital mortality after stomach cancer surgery in Spain and relationship with hospital volume of interventions. BMC Public Health 2009; 9: Damhuis R, Meurs C, Dijkhuis C, Stassen L, Wiggers T. Hospital volume and post-operative mortality after resection for gastric cancer. Eur J Surg Oncol 2002; 28: Reavis KM, Hinojosa MW, Smith BR, Wooldridge JB, Krishnan S, Nguyen NT. Hospital volume is not a predictor of outcomes after gastrectomy for neoplasm. Am Surg 2009; 75: Skipworth RJ, Parks RW, Stephens NA, Graham C, Brewster DH, Garden OJ et al. The relationship between hospital volume and post-operative mortality rates for upper gastrointestinal cancer resections: Scotland Eur J Surg Oncol 2010; 36: Birkmeyer JD, Siewers AE, Finlayson EV, Stukel TA, Lucas FL, Batista I et al. Hospital volume and surgical mortality in the United States. NEnglJMed2002; 346: Kuwabara K, Matsuda S, Fushimi K, Ishikawa KB, Horiguchi H, Fujimori K. Hospital volume and quality of laparoscopic gastrectomy in Japan. Dig Surg 2009; 26: Learn PA, Bach PB. A decade of mortality reductions in major oncologic surgery: the impact of centralization and quality improvement. Med Care 2010; 48: Nomura E, Tsukuma H, Ajiki W, Oshima A. Population-based study of relationship between hospital surgical volume and 5-year survival of stomach cancer patients in Osaka, Japan. Cancer Sci 2003; 94: Gillison EW, Powell J, McConkey CC, Spychal RT. Surgical workload and outcome after resection for carcinoma of the oesophagus and cardia. Br J Surg 2002; 89: Rouvelas I, Lindblad M, Zeng W, Viklund P, Ye W, Lagergren J. Impact of hospital volume on long-term survival after esophageal cancer surgery. Arch Surg 2007; 142: Verhoef C, van de Weyer R, Schaapveld M, Bastiaannet E, Plukker JT. Better survival in patients with esophageal cancer after surgical treatment in university hospitals: a plea for performance by surgical oncologists. Ann Surg Oncol 2007; 14: van de Poll-Franse LV, Lemmens VE, Roukema JA, Coebergh JW, Nieuwenhuijzen GA. Impact of concentration of oesophageal and gastric cardia cancer surgery on long-term population-based survival. Br J Surg 2011; 98: Bachmann MO, Alderson D, Edwards D, Wotton S, Bedford C, Peters TJ et al. Cohort study in South and West England of the influence of specialization on the management and outcome of patients with oesophageal and gastric cancers. Br J Surg 2002; 89: Birkmeyer JD, Sun Y, Wong SL, Stukel TA. Hospital volume and late survival after cancer surgery. Ann Surg 2007; 245: Enzinger PC, Benedetti JK, Meyerhardt JA, McCoy S, Hundahl SA, Macdonald JS et al. Impact of hospital volume on recurrence and survival after surgery for gastric cancer. Ann Surg 2007; 245: Ioka A, Tsukuma H, Ajiki W, Oshima A. Hospital procedure volume and survival of cancer patients in Osaka, Japan: a population-based study with latest cases. Jpn J Clin Oncol 2007; 37: Xirasagar S, Lien YC, Lin HC, Lee HC, Liu TC, Tsai J. Procedure volume of gastric cancer resections versus 5-year survival. Eur J Surg Oncol 2008; 34: Thompson AM, Rapson T, Gilbert FJ, Park KGM. Hospital volume does not influence long-term survival of patients undergoing surgery for oesophageal or gastric cancer. Br J Surg 2007; 94: Aiken LH, Clarke SP, Cheung RB, Sloane DM, Silber JH. Educational levels of hospital nurses and surgical patient mortality. JAMA 2003; 290: Wouters MW, Krijnen P, Le Cessie S, Gooiker GA, Guicherit OR, Marinelli AW et al. Volume- or outcome-based referral to improve quality of care for esophageal cancer surgery in the Netherlands. J Surg Oncol 2009; 99: van Gijn W, van de Velde CJ; members of the EURECCA consortium. Improving quality of cancer care through surgical audit. Eur J Surg Oncol 2010; 36(Suppl 1): S23 S26.

12 94 J. L. Dikken, J. W. van Sandick, W. H. Allum, J. Johansson, L. S. Jensen, H. Putter et al. Supporting information Additional supporting information may be found in the online version of this article: Fig. S1 Distribution of annual hospital volumes: a oesophagectomy (n = ) and b gastrectomy (n = 9010) (Word document) Fig. S2 Unadjusted 2-year survival rates after a oesophagectomy and b gastrectomy in the Netherlands, Sweden, Denmark and England, determined by Kaplan Meier analysis. Survival estimates and 95 per cent confidence intervals are provided in Table 3 (Word document) Fig. S3 Two-year survival after a oesophagectomy and b gastrectomy, adjusted for sex, age, histology and stage, in the Netherlands, Sweden and Denmark. Survival estimates and 95 per cent confidence intervals are provided in Table 3 (Word document) Fig. S4 Two-year survival after a oesophagectomy and b gastrectomy, adjusted for sex, age, histology and stage group, by annual hospital volume category (Word document)