Alcohol consumption and survival of colorectal cancer patients: population-based study from Germany 1,2

Transcription

1 AJCN. First published ahead of print May 4, 2016 as doi: /ajcn Alcohol consumption and survival of colorectal cancer patients: population-based study from Germany 1,2 Viola Walter, 3 * Lina Jansen, 3 Alexis Ulrich, 8 Wilfried Roth, 4,9 Hendrik Bläker, 10 Jenny Chang-Claude, 5 Michael Hoffmeister, 3 and Hermann Brenner 3,6,7 3 Division of Clinical Epidemiology and Aging Research, 4 Unit of Molecular Tumor Pathology, 5 Division of Cancer Epidemiology, and 6 German Cancer Consortium, German Cancer Research Center, Heidelberg, Germany; 7 Division of Preventive Oncology, National Center for Tumor Diseases, Heidelberg, Germany; Departments of 8 General, Visceral and Transplantation Surgery and 9 Pathology, Institute of Pathology, University Hospital, Heidelberg, Germany; and 10 Institute of Pathology, Charité University Medicine, Berlin, Germany ABSTRACT Background: Studies on the association between alcohol consumption and colorectal cancer (CRC) prognosis have yielded inconsistent results. Objective: The associations of lifetime and 1-y prediagnostic alcohol consumption with relevant prognostic outcomes were evaluated in a large population-based cohort of CRC patients. Design: In , 3121 patients diagnosed with CRC were interviewed on sociodemographic and lifestyle factors, medication, and comorbidities. Cancer recurrence, vital status, and cause of death were documented for a median follow-up time of 4.8 y. With the use of Cox proportional hazard regression, associations between lifetime and recent alcohol consumption and overall, CRC-specific, recurrence-free, and disease-free survival were analyzed. Results: In this patient cohort with a median age of 69 y at diagnosis, lifetime abstainers showed poorer overall [adjusted HR (ahr): 1.25; 95% CI: 1.03, 1.52] and CRC-specific (ahr: 1.37; 95% CI: 1.10, 1.70) survival than lifetime light (women:.0 12 g/d; men:.0 24 g/d). Lifetime heavy showed poorer overall (ahr: 1.37; 95% CI: 1.06, 1.78) and disease-free (ahr: 1.38; 95% CI: 1.09, 1.74) survival. Alcohol abstaining in the year before diagnosis was associated with poorer overall (ahr: 1.42; 95% CI: 1.20, 1.68), CRC-specific (ahr: 1.38; 95% CI: 1.13, 1.68), and disease-free (ahr: 1.23; 95% CI: 1.05, 1.44) survival. Lifetime abstainers with nonmetastatic disease showed poorer CRC-specific (ahr: 1.48; 95% CI: 1.10, 2.00) and recurrence-free (ahr: 1.32; 95% CI: 1.02, 1.70) survival. Wine abstaining but not beer or liquor abstaining was associated with poorer survival. Associations between alcohol consumption and prognosis varied according to presence of diabetes and age. Conclusions: Prediagnostic alcohol abstaining and heavy drinking were associated with poorer survival after a CRC diagnosis than light drinking. The protective effects of light consumption might be restricted to wine, and associations might differ according to age and presence of diabetes mellitus. Am J Clin Nutr doi: /ajcn Keywords: colorectal neoplasms, alcohol consumption, survival, prognosis, recurrence INTRODUCTION Colorectal cancer (CRC) is the third most commonly diagnosed malignancy and the fourth most common cause of cancer-related death worldwide (1). Approximately half of the patients still die within 5 y from diagnosis (2), and further efforts to improve prognosis are needed. In 2010, the International Agency for Research on Cancer added CRC to the list of alcoholrelated cancers (3). Subsequently, meta-analyses have provided extensive evidence on the relation between alcohol consumption and CRC risk (4) and CRC mortality (5). They reported moderate and heavy alcohol consumption to be associated with a higher CRC incidence and heavy drinking to be associated with increased CRC mortality in previously cancer-free individuals. It is assumed these associations result from a complex synergy of various biological mechanisms (6 9). In addition, alcohol consumption might also influence the prognosis of CRC patients after diagnosis. To our knowledge, a potential association of alcohol consumption with CRC prognosis has been evaluated in 6 studies so far (10 15). Categorization of exposure has been quite diverse, as have been the results. Reduced overall survival was found for prediagnostic alcohol consumption (categorized as yes compared with no ) (15) and for postdiagnostic abstaining compared with consumption of 5 15 g/d among women (14). A protective effect of prediagnostic moderate wine consumption (1 3 times/mo compared with less often) or overall alcohol consumption was reported in studies on familial CRC cases (10) or colon cancer cases (12), respectively. 1 Supported by grants from the German Research Council (BR 1704/6-1, BR 1704/6-3, BR 1704/6-4, and CH 117/1-1), the German Federal Ministry of Education and Research (01KH0404 and 01ER0814), and the Interdisciplinary Research Program of the German National Center for Tumor Diseases. 2 Supplemental Figures 1 3, Supplemental Tables 1 and 2, and Supplemental Text are available from the Online Supporting Material link in the online posting of the article and from the same link in the online table of contents at *To whom correspondence should be addressed. v.walter@dkfz. de. Received November 10, Accepted for publication April 1, doi: /ajcn Am J Clin Nutr doi: /ajcn Printed in USA. Ó 2016 American Society for Nutrition 1 of 10 Copyright (C) 2016 by the American Society for Nutrition

2 2of10 WALTER ET AL. In 2 further studies, no effect of prediagnostic alcohol consumption on overall survival was found (11, 13). In addition, 2 studies investigated CRC-specific survival but found no association with alcohol consumption (11, 14). Given these inconclusive results, the heterogeneous assessment of exposure, and a sparse evaluation of survival outcomes, further evidence is much needed to adequately evaluate the association between alcohol consumption and CRC prognosis. We therefore aimed to investigate alcohol consumption in the year up to diagnosis and lifetime alcohol consumption in association with overall, CRCspecific, recurrence-free, and disease-free survival of CRC patients after diagnosis, in a large population-based setting in Germany. METHODS Study design and study population Data were collected from 3146 patients with a first, histologically confirmed diagnosis of CRC. This study population of CRC cases only has been recruited within the DACHS study [Darmkrebs: Chancen der Verhütung durch Screening (colorectal cancer: chances for prevention through screening)] in (International Classification of Diseases, 10th Revision, codes C18 C20). The DACHS study is an ongoing populationbased case-control study with regular follow-up of cases that is carried out in the southwest of Germany in the Rhine-Neckar- Odenwald Region (w2 million inhabitants); further details have been described elsewhere (16 19) (Supplemental Text). All study participants gave written informed consent. The study was approved by the ethics committees of the University of Heidelberg and the state medical boards of Baden-Wuerttemberg and Rhineland-Palatinate. Sociodemographic information, medical, and lifestyle histories of patients were obtained at baseline by trained interviewers through the use of a standardized questionnaire. Details on collection of follow-up data have been described elsewhere (18 20) (Supplemental Text). Follow-up time started at diagnosis, and the vital status of patients was checked regularly with population registers. CRC recurrence information was obtained by standardized questionnaires from patients treating physicians. Death certificates were collected from local public health departments to determine the cause of death. Alcohol consumption assessment At baseline, patients were asked how many portions of either beer (0.33 L), wine (0.25 L), or liquor (0.02 L) they had consumed on average per week at ages 20, 30, 40, 50, 60, 70, and 80 y and in the 12 mo before onset of the medical examinations that led to diagnosis of their CRC. Ethanol content for the different beverage types was derived from food composition tables (21), assuming 4, 8.6, and 33 g of pure ethanol in 100 ml of beer, wine, and liquor, respectively. The total weekly ethanol consumption for each of the reported stages of life was calculated and divided by 7 d, deriving the mean daily dose at the respective time point. For recent mean daily alcohol consumption, we used the mean daily ethanol intake in the 12 mo before onset of the medical examinations that led to diagnosis (patients were interviewed about that at baseline, typically shortly after CRC diagnosis). Mean lifetime alcohol consumption was calculated based on self-reported (recalled) alcohol consumption at 20, 30, 40, 50, 60, 70, and 80 y of age. Both recent and lifetime alcohol consumption were categorized into sexspecific dose groups. Women with consumption levels of 0,.0 12,.12 25, or.25 g/d and men with consumption levels of 0,.0 24,.24 50, or.50 g/d were each classified as abstainers or light, moderate, or heavy, respectively. Statistical analysis The distribution of sociodemographic and clinical variables in total and according to lifetime and recent mean daily alcohol consumption was evaluated in descriptive analyses. Through the use of Cox proportional hazard regression, we evaluated the associations of lifetime and recent alcohol consumption with overall, CRC-specific, recurrence-free, and disease-free survival. Endpoints for overall, CRC-specific, recurrence-free, and disease-free survival were death from any cause, death from CRC, recurrence or death from CRC, and recurrence or death from any cause, respectively. Cox models were computed by using 2 different adjustment settings, defined a priori: 1) adjustment for age at diagnosis (30 59, 60 69, 70 79, or $80 y) and sex and 2) further adjustment for cancer stage (I IV according to the International Union Against Cancer classification); BMI (in kg/m 2 ;,25, 25 to,30, or $30); smoking status (never, former, or current smoker); regular use of b-blockers (yes or no), statins (yes/no), or nonsteroidal antiinflammatory drugs (yes or no); diabetes mellitus (yes or no); history of major cardiovascular disease (including heart failure, myocardial infarction, angina pectoris, or stroke; yes or no); and history of cancer other than CRC (yes or no). Associations of lifetime and recent alcohol consumption with overall and CRCspecific survival were further investigated in dose-response analyses (22) and in Cox models stratified by patient and tumor characteristics. For further details on multiple imputation of missing values and statistical analyses conducted, see Supplemental Text. Data analyses were performed with SAS software, version 9.2 (SAS Institute). Imputation of missing covariables was performed with the use of R, version (23) and the R package mice, version 2.22 (24). Statistical tests were 2-sided, with an a level of No corrections for multiple comparisons were made. RESULTS Study characteristics From the 3146 CRC patients, recruited in , we excluded participants with missing endpoints or follow-up times (n = 15) or with missing International Union Against Cancer stage classification (n = 10). In analyses of recurrence-free and disease-free survival, we excluded patients with recurrence before the interview (n = 87). Of the 3121 patients included in the analyses, 868 died during the follow-up period; of these, the majority (n = 635) died of CRC. A total of 623 patients had a relapse or developed metastases. Median follow-up was 4.8 y for overall and 3.9 y for recurrence-free survival. Descriptive characteristics of the 3121 study participants are shown in Tables 1 and 2. Overall, 59.4% of patients were men, and the median age in our study population was 69 y. Stage I, II, III, and IV CRC was present in 22.2%, 30.6%, 32.1%, and 15% of patients, respectively. Median overall BMI was 26.1 and quite consistent over all alcohol consumption categories. With 55.6% and

3 ALCOHOL CONSUMPTION AND CRC PROGNOSIS 3 of 10 TABLE 1 Distribution of patient characteristics in the total study population of CRC patients and in subgroups according to lifetime mean alcohol consumption behavior 1 Lifetime mean alcohol consumption behavior Total 2 Abstainers Heavy P value 3 Patients 3121 (100) 511 (16.4) 1734 (55.6) 615 (19.7) 248 (7.9) Sex Female 1266 (40.6) 407 (79.6) 703 (40.5) 108 (17.6) 43 (17.3) Male 1855 (59.4) 104 (20.4) 1031 (59.5) 507 (82.4) 205 (82.7) Age at diagnosis, y 69 (30 96) 4 72 (30 92) 69 (33 96) 68 (37 90) 65 (39 94) (19.9) 92 (18.0) 349 (20.1) 117 (19.0) 63 (25.4) (31.9) 116 (22.7) 544 (31.4) 233 (37.9) 102 (41.1) (32.5) 179 (35.0) 572 (33.0) 193 (31.4) 65 (26.2) $ (15.6) 124 (24.3) 269 (15.5) 72 (11.7) 18 (7.3) Tumor location Colon 1847 (59.2) 341 (66.7) 1037 (59.8) 340 (55.3) 120 (48.4) Rectum 1274 (40.8) 170 (33.3) 697 (40.2) 275 (44.7) 128 (51.6) Stage I 694 (22.2) 100 (19.6) 411 (23.7) 126 (20.5) 54 (21.8) II 956 (30.6) 158 (30.9) 513 (29.6) 197 (32.0) 84 (33.9) III 1002 (32.1) 171 (33.5) 566 (32.6) 193 (31.4) 70 (28.2) IV 469 (15.0) 82 (16.0) 244 (14.1) 99 (16.1) 40 (16.1) BMI, kg/m ( ) 26.0 ( ) 26.1 ( ) 26.1 ( ) 26.1 ( ), (38.4) 208 (40.7) 658 (37.9) 235 (38.2) 93 (37.5) $25 to, (42.3) 189 (37.0) 747 (43.1) 257 (41.8) 120 (48.4) $ (19.0) 110 (21.5) 323 (18.6) 122 (19.8) 35 (14.1) Unknown 13 (0.4) 4 (0.8) 6 (0.3) 1 (0.2) 0 (0.0) Smoking status Never smokers 1416 (45.4) 351 (68.7) 823 (47.5) 176 (28.6) 62 (25.0) Former smokers 1266 (40.6) 105 (20.5) 690 (39.8) 344 (55.9) 124 (50.0) Current smokers 428 (13.7) 54 (10.6) 218 (12.6) 93 (15.1) 61 (24.6) Unknown 11 (0.4) 1 (0.2) 3 (0.2) 2 (0.3) 1 (0.4) Schooling, y, (68.2) 385 (75.3) 1158 (66.8) 406 (66.0) 171 (69.0) (16.4) 77 (15.1) 297 (17.1) 100 (16.3) 35 (14.1) (15.2) 46 (9.0) 276 (15.9) 109 (17.7) 42 (16.9) Unknown 6 (0.2) 3 (0.6) 3 (0.2) 0 (0.0) 0 (0.0) Use of b-blockers Yes 817 (26.2) 138 (27.0) 451 (26.0) 170 (27.6) 55 (22.2) No 2259 (72.4) 360 (70.5) 1263 (72.8) 437 (71.1) 191 (77.0) Unknown 45 (1.4) 13 (2.5) 20 (1.2) 8 (1.3) 2 (0.8) Use of statins Yes 424 (13.6) 65 (12.7) 235 (13.6) 87 (14.1) 34 (13.7) No 2692 (86.3) 446 (87.3) 1497 (86.3) 527 (85.7) 214 (86.3) Unknown 5 (0.2) 0 (0.0) 2 (0.1) 1 (0.2) 0 (0.0) Use of NSAIDs Yes 729 (23.4) 122 (23.9) 399 (23.0) 154 (25.0) 51 (20.6) No 2324 (74.5) 377 (73.8) 1300 (75.0) 447 (72.7) 191 (77.0) Unknown 68 (2.2) 12 (2.3) 35 (2.0) 14 (2.3) 6 (2.4) Diabetes mellitus Yes 577 (18.5) 119 (23.3) 310 (17.9) 106 (17.2) 37 (14.9) No 2540 (81.4) 391 (76.5) 1422 (82.0) 508 (82.6) 211 (85.1) Unknown 4 (0.1) 1 (0.2) 2 (0.1) 1 (0.2) 0 (0.0) History of heart failure, myocardial infarction, angina pectoris, or stroke Yes 795 (25.5) 146 (28.6) 439 (25.3) 154 (25.0) 50 (20.2) No 2323 (74.4) 365 (71.4) 1294 (74.6) 461 (75.0) 196 (79.0) Unknown 3 (0.1) 0 (0.0) 1 (0.1) 0 (0.0) 2 (0.8) History of other cancer Yes 353 (11.3) 68 (13.3) 191 (11.0) 66 (10.7) 28 (11.3) No 2763 (88.5) 441 (86.3) 1540 (88.8) 549 (89.3) 220 (88.7) Unknown 5 (0.2) 2 (0.4) 3 (0.2) 0 (0.0) 0 (0.0) 1 Values are n (%) unless otherwise indicated. Alcohol consumption categories use sex-specific cutoffs: light :.0 12 g/d (women) and.0 24 g/d (men); moderate : g/d (women) and g/d (men); and heavy :.25 g/d (women) and.50 g/d (men). CRC, colorectal cancer; NSAID, nonsteroidal anti-inflammatory drug. 2 Includes 13 participants with missing lifetime alcohol consumption information. 3 Pearson x 2 test of independence between covariables and lifetime alcohol consumption. 4 Median; range in parentheses (all such values).

4 4of10 WALTER ET AL. TABLE 2 Distribution of patient characteristics in the total study population of CRC patients and in subgroups according to recent mean daily alcohol consumption behavior 1 Recent mean alcohol consumption behavior Total 2 Abstainers Heavy P value 3 Patients 3121 (100) 917 (29.4) 1303 (41.7) 572 (18.3) 280 (9.0) Sex Female 1266 (40.6) 598 (65.2) 418 (32.1) 153 (26.7) 69 (24.6) Male 1855 (59.4) 319 (34.8) 885 (67.9) 419 (73.3) 211 (75.4) Age at diagnosis, y 69 (30 96) 4 71 (30 94) 69 (33 96) 69 (37 94) 66 (43 88) (19.9) 169 (18.4) 273 (21.0) 109 (19.1) 54 (19.3) (31.9) 250 (27.3) 428 (32.8) 186 (32.5) 120 (42.9) (32.5) 313 (34.1) 411 (31.5) 194 (33.9) 82 (29.3) $ (15.6) 185 (20.2) 191 (14.7) 83 (14.5) 24 (8.6) Tumor location Colon 1847 (59.2) 590 (64.3) 766 (58.8) 330 (57.7) 131 (46.8) Rectum 1274 (40.8) 327 (35.7) 537 (41.2) 242 (42.3) 149 (53.2) Stage I 694 (22.2) 192 (20.9) 296 (22.7) 138 (24.1) 54 (19.3) II 956 (30.6) 292 (31.8) 385 (29.5) 173 (30.2) 97 (34.6) III 1002 (32.1) 302 (32.9) 424 (32.5) 176 (30.8) 81 (28.9) IV 469 (15.0) 131 (14.3) 198 (15.2) 85 (14.9) 48 (17.1) BMI 26.1 ( ) 26.0 ( ) 26.2 ( ) 25.8 ( ) 26.1 ( ), (38.4) 354 (38.6) 483 (37.1) 240 (42.0) 105 (37.5) $25 to, (42.3) 352 (38.4) 577 (44.3) 241 (42.1) 129 (46.1) $ (19.0) 204 (22.2) 240 (18.4) 90 (15.7) 46 (16.4) Unknown 13 (0.4) 7 (0.8) 3 (0.2) 1 (0.2) 0 (0.0) Smoking status Never smokers 1416 (45.4) 527 (57.5) 592 (45.4) 200 (35.0) 72 (25.7) Former smokers 1266 (40.6) 269 (29.3) 546 (41.9) 298 (52.1) 142 (50.7) Current smokers 428 (13.7) 121 (13.2) 162 (12.4) 73 (12.8) 64 (22.9) Unknown 11 (0.4) 0 (0.0) 3 (0.2) 1 (0.2) 2 (0.7) Schooling, y, (68.2) 688 (75.0) 849 (65.2) 376 (65.7) 184 (65.7) (16.4) 131 (14.3) 231 (17.7) 96 (16.8) 44 (15.7) (15.2) 94 (10.3) 221 (17.0) 100 (17.5) 52 (18.6) Unknown 6 (0.2) 4 (0.4) 2 (0.2) 0 (0.0) 0 (0.0) Use of b-blockers Yes 817 (26.2) 265 (28.9) 323 (24.8) 144 (25.2) 74 (26.4) No 2259 (72.4) 636 (69.4) 962 (73.8) 421 (73.6) 204 (72.9) Unknown 45 (1.4) 16 (1.7) 18 (1.4) 7 (1.2) 2 (0.7) Use of statins Yes 424 (13.6) 129 (14.1) 177 (13.6) 79 (13.8) 32 (11.4) No 2692 (86.3) 787 (85.8) 1125 (86.3) 492 (86.0) 248 (88.6) Unknown 5 (0.2) 1 (0.1) 1 (0.1) 1 (0.2) 0 (0.0) Use of NSAIDs Yes 729 (23.4) 242 (26.4) 278 (21.3) 137 (24.0) 61 (21.8) No 2324 (74.5) 655 (71.4) 1001 (76.8) 422 (73.8) 210 (75.0) Unknown 68 (2.2) 20 (2.2) 24 (1.8) 13 (2.3) 9 (3.2) Diabetes mellitus Yes 577 (18.5) 216 (23.6) 243 (18.6) 71 (12.4) 36 (12.9) No 2540 (81.4) 700 (76.3) 1058 (81.2) 500 (87.4) 244 (87.1) Unknown 4 (0.1) 1 (0.1) 2 (0.2) 1 (0.2) 0 (0.0) History of heart failure, myocardial infarction, angina pectoris, or stroke Yes 795 (25.5) 282 (30.8) 323 (24.8) 121 (21.2) 57 (20.4) No 2323 (74.4) 634 (69.1) 979 (75.1) 451 (78.8) 222 (79.3) Unknown 3 (0.1) 1 (0.1) 1 (0.1) 0 (0.0) 1 (0.4) History of other cancer Yes 353 (11.3) 127 (13.8) 138 (10.6) 57 (10.0) 28 (10.0) No 2763 (88.5) 787 (85.8) 1163 (89.3) 515 (90.0) 252 (90.0) Unknown 5 (0.2) 3 (0.3) 2 (0.2) 0 (0.0) 0 (0.0) 1 Values are n (%) unless otherwise indicated. Alcohol consumption categories use sex-specific cutoffs: light :.0 12 g/d (women) and.0 24 g/d (men); moderate : g/d (women) and g/d (men); and heavy :.25 g/d (women) and.50 g/d (men). CRC, colorectal cancer; NSAID, nonsteroidal anti-inflammatory drug. 2 Includes 49 participants with missing recent alcohol consumption information. 3 Pearson x 2 test of independence between covariables and recent alcohol consumption. 4 Median; range in parentheses (all such values).

5 ALCOHOL CONSUMPTION AND CRC PROGNOSIS 5 of %, most CRC patients were lifetime and recent light, respectively. Among abstainers, more patients were women, whereas among light, moderate, and heavy, the majority were men. Lifetime and recent heavy were younger (median: 65 and 66 y) and more likely to be current smokers than patients with lower alcohol consumption. Only in heavy, the majority of CRCs were rectal and not colon cancers. Lifetime and recent abstainers showed the lowest proportions of patients with.10 y of schooling (9.0% and 10.3%, respectively) and the highest prevalence of diabetes mellitus (23.3% and 23.6%, respectively). Recent abstainers included higher proportions of obese patients; nonsteroidal anti-inflammatory drug users; patients with history of heart failure, myocardial infarction, angina pectoris, or stroke; and patients with history of other cancers, compared with recent alcohol consumers. Of patients who were still alive 5 y after cancer diagnosis and who participated in the 5-y follow-up survey (Supplemental Table 1), 30% decreased and 10% increased their alcohol consumption between before and 5 y after diagnosis, with a weighted k value of Prognosis according to lifetime and recent alcohol consumption Table 3 shows the associations between lifetime and recent alcohol consumption and the different endpoints, among stage I IV patients. In Cox models, fully adjusted for all relevant covariables, lifetime abstainers showed significantly reduced overall [adjusted HR (ahr): 1.25; 95% CI: 1.03, 1.52, P value = 0.022] and CRC-specific (ahr: 1.37; 95% CI: 1.10, 1.70, P value = 0.006) survival and nonsignificantly poorer recurrencefree and disease-free survival than lifetime light. Lifetime heavy drinking was associated with significantly poorer overall (ahr: 1.37; 95% CI: 1.06, 1.78, P value = 0.015) and disease-free (ahr: 1.38; 95% CI: 1.09, 1.74, P value = 0.008) survival, with a tendency (P = 0.089) in the same direction also for recurrence-free survival. Recent abstaining was associated with significantly poorer overall (ahr: 1.42; 95% CI: 1.20, 1.68, P value = ), CRC-specific (ahr: 1.38; 95% CI: 1.13, 1.68, P value = 0.001), and disease-free (ahr: 1.23; 95% CI: 1.05, 1.44, P value = 0.010) survival. More details on the doseresponse association between alcohol consumption and survival can be found in Supplemental Figure 1. Table 4 shows the associations between lifetime and recent alcohol consumption and prognosis in nonmetastatic CRC patients (stages I III). Lifetime abstaining was associated with significantly poorer CRC-specific (ahr: 1.48; 95% CI: 1.10, 2.00, P value = 0.010) and recurrence-free (ahr: 1.32; 95% CI: 1.02, 1.70, P value = 0.035) survival and borderline significantly poorer overall and disease-free survival. Lifetime heavy drinking showed an association with poorer disease-free survival. Results for recent alcohol consumption in nonmetastatic CRC patients were similar to associations found for all patients. TABLE 3 HRs for overall, CRC-specific, recurrence-free, and disease-free survival according to lifetime mean alcohol consumption and recent mean alcohol consumption among stage I IV CRC patients 1 Lifetime mean alcohol consumption Recent mean alcohol consumption Abstainers Heavy Abstainers Heavy At risk, n Overall survival Events, n ahr 2 (95% CI) 1.31 (1.09, 1.58) 1.00 (ref) 1.09 (0.91, 1.31) 1.32 (1.03, 1.70) 1.34 (1.13, 1.58) 1.00 (ref) 1.01 (0.83, 1.22) 1.24 (0.97, 1.58) ahr 3 (95% CI) 1.25 (1.03, 1.52) 1.00 (ref) 1.07 (0.90, 1.29) 1.37 (1.06, 1.78) 1.42 (1.20, 1.68) 1.00 (ref) 1.09 (0.90, 1.33) 1.17 (0.91, 1.51) CRC-specific survival Events, n ahr 2 (95% CI) 1.43 (1.16, 1.77) 1.00 (ref) 1.16 (0.94, 1.43) 1.12 (0.82, 1.54) 1.25 (1.03, 1.51) 1.00 (ref) 1.00 (0.80, 1.25) 1.11 (0.83, 1.49) ahr 3 (95% CI) 1.37 (1.10, 1.70) 1.00 (ref) 1.12 (0.91, 1.38) 1.18 (0.85, 1.62) 1.38 (1.13, 1.68) 1.00 (ref) 1.06 (0.84, 1.33) 1.03 (0.76, 1.39) At risk, 4 n Recurrence-free survival Events, n ahr 2 (95% CI) 1.31 (1.08, 1.59) 1.00 (ref) 1.09 (0.91, 1.32) 1.18 (0.91, 1.54) 1.10 (0.93, 1.31) 1.00 (ref) 0.90 (0.74, 1.10) 1.10 (0.85, 1.41) ahr 3 (95% CI) 1.14 (0.94, 1.40) 1.00 (ref) 1.01 (0.84, 1.22) 1.26 (0.97, 1.65) 1.13 (0.94, 1.34) 1.00 (ref) 0.96 (0.78, 1.18) 1.06 (0.82, 1.37) Disease-free survival Events, n ahr 2 (95% CI) 1.28 (1.07, 1.52) 1.00 (ref) 1.13 (0.96, 1.33) 1.32 (1.05, 1.66) 1.22 (1.05, 1.42) 1.00 (ref) 0.96 (0.80, 1.15) 1.20 (0.96, 1.51) ahr 3 (95% CI) 1.13 (0.95, 1.36) 1.00 (ref) 1.06 (0.90, 1.26) 1.38 (1.09, 1.74) 1.23 (1.05, 1.44) 1.00 (ref) 1.03 (0.86, 1.23) 1.16 (0.92, 1.46) 1 Alcohol consumption categories use sex-specific cutoffs: light :.0 12 g/d (women) and.0 24 g/d (men); moderate : g/d (women) and g/d (men); and heavy :.25 g/d (women) and.50 g/d (men). ahr, adjusted HR; Ca, noncolorectal cancer; CRC, colorectal cancer; ref, reference. 2 Cox proportional hazard regression model was adjusted for age and sex; additional adjustment was made for age 3 log(time) to account for timedependent effects. 3 Cox proportional hazard regression model was adjusted for age; sex; stage; smoking status; use of statins; use of nonsteroidal anti-inflammatory drugs; use of b-blockers; diabetes mellitus; history of heart failure, myocardial infarction, angina pectoris, or stroke; and history of Ca. Additional adjustment was made for age 3 log(time) and Ca 3 log(time) to account for time-dependent effects. 4 Excluding belatedly interviewed cases with recurrence before the interview.

6 6of10 WALTER ET AL. TABLE 4 Hazard ratios for overall, CRC-specific, recurrence-free, and disease-free survival according to lifetime mean alcohol consumption and recent mean alcohol consumption, among stage I III (nonmetastatic) CRC patients 1 Lifetime mean alcohol consumption Recent mean alcohol consumption Abstainers Heavy Abstainers Heavy At risk, n Overall survival Events, n ahr 2 (95% CI) 1.27 (1.00, 1.62) 1.00 (ref) 1.04 (0.81, 1.32) 1.29 (0.92, 1.82) 1.50 (1.21, 1.86) 1.00 (ref) 0.96 (0.74, 1.25) 1.33 (0.96, 1.85) ahr 3 (95% CI) 1.26 (0.99, 1.61) 1.00 (ref) 1.07 (0.83, 1.36) 1.33 (0.94, 1.88) 1.46 (1.17, 1.81) 1.00 (ref) 1.04 (0.79, 1.36) 1.38 (0.99, 1.92) CRC-specific survival Events, n ahr 2 (95% CI) 1.43 (1.06, 1.93) 1.00 (ref) 1.08 (0.79, 1.49) 0.87 (0.51, 1.48) 1.38 (1.05, 1.82) 1.00 (ref) 0.94 (0.66, 1.33) 1.11 (0.70, 1.75) ahr 3 (95% CI) 1.48 (1.10, 2.00) 1.00 (ref) 1.12 (0.81, 1.55) 0.89 (0.52, 1.53) 1.39 (1.05, 1.85) 1.00 (ref) 1.02 (0.71, 1.45) 1.13 (0.71, 1.79) At risk, 4 n Recurrence-free survival Events, n ahr 2 (95% CI) 1.32 (1.02, 1.70) 1.00 (ref) 1.01 (0.78, 1.31) 1.06 (0.73, 1.54) 1.15 (0.92, 1.45) 1.00 (ref) 0.84 (0.63, 1.11) 1.08 (0.76, 1.53) ahr 3 (95% CI) 1.32 (1.02, 1.70) 1.00 (ref) 1.02 (0.79, 1.32) 1.10 (0.75, 1.60) 1.16 (0.92, 1.46) 1.00 (ref) 0.91 (0.68, 1.21) 1.11 (0.78, 1.59) Disease-free survival Events, n ahr 2 (95% CI) 1.27 (1.02, 1.58) 1.00 (ref) 1.06 (0.86, 1.32) 1.30 (0.97, 1.75) 1.33 (1.10, 1.61) 1.00 (ref) 0.90 (0.71, 1.14) 1.26 (0.95, 1.68) ahr 3 (95% CI) 1.25 (1.00, 1.55) 1.00 (ref) 1.08 (0.87, 1.34) 1.35 (1.00, 1.83) 1.29 (1.06, 1.57) 1.00 (ref) 0.97 (0.77, 1.23) 1.32 (0.99, 1.77) 1 Alcohol consumption categories use sex-specific cutoffs: light :.0 12 g/d (women) and.0 24 g/d (men); moderate : g/d (women) and g/d (men); and heavy :.25 g/d (women) and.50 g/d (men). ahr, adjusted HR; Ca, noncolorectal cancer; CRC, colorectal cancer; ref, reference. 2 Cox proportional hazard regression model was adjusted for age and sex; additional adjustment was made for age 3 log(time) to account for timedependent effects. 3 Cox proportional hazard regression model was adjusted for age; sex; stage; smoking status; use of statins; use of nonsteroidal anti-inflammatory drugs; use of b-blockers; diabetes mellitus; history of heart failure, myocardial infarction, angina pectoris, or stroke; and history of Ca. Additional adjustment was made for age 3 log(time) and Ca 3 log(time) to account for time-dependent effects. 4 Excluding belatedly interviewed cases with recurrence before the interview. Association tendencies between recent heavy drinking and poorer overall (P = 0.060) and disease-free (P = 0.058) survival were observed in stage I III patients only. In stage IV patients (Table 5), lifetime heavy drinking was significantly associated with reduced recurrence-free and disease-free survival, whereas only an association tendency was found with overall survival (P = 0.058). Recent abstaining was significantly associated with poorer overall and CRC-specific survival. In sensitivity analyses, no relevant difference was observed when comparing survival analysis results based on unimputed data (complete cases; data not shown) with the main survival analysis results based on imputed data. Beverage type specific analyses In sensitivity analyses checking for differences in associations according to beverage type (Tables 6 and 7), lifetime and recent wine abstaining were significantly associated with poorer overall (lifetime ahr: 1.30; 95% CI: 1.11, 1.53) and CRC-specific (lifetime ahr: 1.37; 95% CI: 1.14, 1.65) survival. No significant associations with prognosis were observed for beer or liquor consumption. Subgroup analyses In subgroup analyses, no significant interactions were observed between lifetime alcohol consumption and any of the analyzed covariates. We also checked for differences in associations between recent alcohol consumption and prognosis according to different patient characteristics. For all outcomes, we found a significant interaction between diabetes mellitus presence and recent alcohol consumption (overall survival: P- interaction = 0.039; CRC-specific survival: P-interaction = 0.014; Supplemental Figures 2 and 3). In diabetes patients, abstaining and moderate/heavy drinking were associated with poorer survival (overall survival ahr: 1.90; 95% CI: 1.32, 2.74, and ahr: 1.51; 95% CI: 1.01, 2.25, respectively), whereas in patients without diabetes, associations were less pronounced, and only abstaining showed associations with poorer survival (overall survival ahr: 1.25; 95% CI: 1.03, 1.52). For all outcomes except overall survival, significant interactions between age and recent alcohol consumption were found. In patients aged,70 y, abstaining was generally not associated with prognosis, and moderate/heavy drinking was sometimes associated with poorer survival, whereas in patients $70 y or older, abstaining was associated with poorer survival. For recurrence-free survival only, a significant interaction was found between smoking behavior and recent alcohol consumption (P-interaction = 0.042). Furthermore, tendencies for effect modification were found for sex and on occasion for tumor location (Supplemental Figures 2 and 3). Supplemental Table 2 shows the results of the association between lifetime and recent alcohol consumption and overall

7 ALCOHOL CONSUMPTION AND CRC PROGNOSIS 7 of 10 TABLE 5 Hazard ratios for overall, CRC-specific, recurrence-free, and disease-free survival according to lifetime mean alcohol consumption and recent mean alcohol consumption, among stage IV CRC patients 1 Lifetime mean alcohol consumption Recent mean alcohol consumption Abstainers Heavy Abstainers Heavy At risk, n Overall survival Events, n ahr 2 (95% CI) 1.29 (0.95, 1.76) 1.00 (ref) 1.11 (0.85, 1.45) 1.50 (1.03, 2.17) 1.35 (1.03, 1.76) 1.00 (ref) 1.18 (0.88, 1.57) 0.98 (0.68, 1.42) ahr 3 (95% CI) 1.26 (0.91, 1.73) 1.00 (ref) 1.12 (0.85, 1.47) 1.45 (0.99, 2.12) 1.35 (1.02, 1.78) 1.00 (ref) 1.16 (0.87, 1.56) 0.98 (0.67, 1.43) CRC-specific survival Events, n ahr 2 (95% CI) 1.33 (0.97, 1.84) 1.00 (ref) 1.17 (0.88, 1.54) 1.43 (0.96, 2.13) 1.35 (1.02, 1.78) 1.00 (ref) 1.18 (0.87, 1.59) 0.95 (0.65, 1.40) ahr 3 (95% CI) 1.27 (0.91, 1.77) 1.00 (ref) 1.17 (0.89, 1.55) 1.38 (0.92, 2.08) 1.34 (1.01, 1.79) 1.00 (ref) 1.15 (0.85, 1.55) 0.95 (0.64, 1.41) At risk, 4 n Recurrence-free survival Events, n ahr 2 (95% CI) 1.06 (0.77, 1.45) 1.00 (ref) 1.01 (0.77, 1.32) 1.52 (1.04, 2.21) 1.16 (0.89, 1.51) 1.00 (ref) 1.11 (0.83, 1.48) 1.01 (0.70, 1.46) ahr 3 (95% CI) 1.00 (0.72, 1.39) 1.00 (ref) 1.00 (0.76, 1.31) 1.55 (1.05, 2.28) 1.11 (0.84, 1.46) 1.00 (ref) 1.08 (0.80, 1.45) 1.02 (0.70, 1.48) Disease-free survival Events, n ahr 2 (95% CI) 1.06 (0.78, 1.44) 1.00 (ref) 1.04 (0.80, 1.35) 1.52 (1.05, 2.19) 1.18 (0.91, 1.53) 1.00 (ref) 1.17 (0.88, 1.56) 1.01 (0.70, 1.45) ahr 3 (95% CI) 1.03 (0.75, 1.42) 1.00 (ref) 1.03 (0.79, 1.34) 1.57 (1.07, 2.29) 1.16 (0.88, 1.51) 1.00 (ref) 1.15 (0.86, 1.53) 1.03 (0.71, 1.49) 1 Alcohol consumption categories use sex-specific cutoffs: light :.0 12 g/d (women) and.0 24 g/d (men); moderate : g/d (women) and g/d (men); and heavy :.25 g/d (women) and.50 g/d (men). ahr, adjusted HR; Ca, noncolorectal cancer; CRC, colorectal cancer; ref, reference. 2 Cox proportional hazard regression model was adjusted for age and sex; additional adjustment was made for age 3 log(time) to account for timedependent effects. 3 Cox proportional hazard regression model was adjusted for age; sex; stage; smoking status; use of statins; use of nonsteroidal anti-inflammatory drugs; use of b-blockers; diabetes mellitus; history of heart failure, myocardial infarction, angina pectoris, or stroke; and history of Ca. Additional adjustment was made for age 3 log(time) and Ca 3 log(time) to account for time-dependent effects. 4 Excluding belatedly interviewed cases with recurrence before the interview. survival stratified by molecular pathological features of CRC. Although some molecular pathological features show borderline significant associations between abstaining and poorer prognosis, no significant interactions were found. DISCUSSION Lifetime and recent alcohol abstaining were associated with poorer overall and also disease-specific prognosis, when compared with light alcohol consumption, although not all associations were statistically significant for each of the investigated endpoints and tumor staging groups. Heavy drinking was associated with poorer survival, although for some endpoints and stage subgroups, only borderline significance was reached. Wine abstaining but not beer or liquor abstaining was significantly associated with poorer survival. Associations between alcohol consumption and prognosis varied according to the presence of diabetes and age group. Since 2010, alcohol consumption has been considered an established risk factor for CRC (3). In meta-analyses, associations between moderate or heavy alcohol consumption and increased CRC incidence (4, 25, 26), occurrence of colorectal adenomas (27, 28), and CRC mortality (5) and all-cancer death (29) in previously cancer-free individuals have been reported. It is currently unclear whether similar associations might be present between alcohol consumption and prognosis after CRC diagnosis. Few studies, 5 on prediagnostic alcohol consumption (10 13, 15) and 1 (14) on postdiagnostic alcohol consumption, have investigated this issue before, and the results have been inconclusive (10 15). In previous studies, heavy alcohol consumption has been associated with increased CRC risk (4, 25, 26), but an association with CRC prognosis could not be shown. An increased overall mortality through consumption of alcohol has been reported by Araújo et al. (15), but this study used only a dichotomous classification of alcohol consumption. Other studies on heavy alcohol consumption and prognosis did not find significant associations (11 14). To our knowledge, none of the previous studies considered lifetime alcohol exposure. In our analyses, associations between heavy drinking and prognosis were more pronounced when lifetime rather than recent consumption was considered. Following recommendations by Rehm et al. (30) that light or moderate alcohol represent a better reference category than the usually quite heterogeneous group of abstainers, we used light alcohol as a reference group in our analyses. We found abstaining to be associated with poorer prognosis after CRC diagnosis compared with light alcohol consumption. This is in accordance with previous studies reporting associations between abstaining and poorer survival in women with CRC (14) or associations between moderate (#2 drinks/d in men and #1 drink/d in women) consumption and increased survival in colon cancer patients (12). Biological mechanisms through which alcohol could influence one s health are manifold, but much is still incompletely understood. There are, among other effects, the carcinogenic

8 8of10 WALTER ET AL. TABLE 6 Hazard ratios for overall and CRC-specific survival according to lifetime consumption of different types of beverage, in stage I IV CRC patients 1 Lifetime beer Lifetime wine Lifetime liquor None #12 g/d.12 g/d None #12 g/d.12 g/d None #12 g/d.12 g/d At risk, n Overall survival Events, n ahr (95% CI) 0.89 (0.73, 1.07) 1.00 (ref) 1.06 (0.88, 1.28) 1.30 (1.11, 1.53) 1.00 (ref) 1.18 (0.97, 1.45) 1.06 (0.90, 1.24) 1.00 (ref) 1.48 (0.72, 3.01) CRC-specific survival Events, n ahr (95% CI) 0.86 (0.69, 1.07) 1.00 (ref) 1.03 (0.82, 1.28) 1.37 (1.14, 1.65) 1.00 (ref) 1.18 (0.93, 1.50) 1.12 (0.93, 1.36) 1.00 (ref) 1.52 (0.67, 3.47) 1 Cox proportional hazard regression model was adjusted for age; sex; stage; smoking status; use of statins; use of nonsteroidal anti-inflammatory drugs; use of b-blockers; diabetes mellitus; history of heart failure, myocardial infarction, angina pectoris, or stroke; and history of Ca. Additional adjustment was made for age 3 log(time) and Ca 3 log(time) to account for time-dependent effects. For each model calculation, all 3 beverage type variables (beer, wine, and liquor) were included into the model and analyzed simultaneously. Estimates were calculated on unimputed data. ahr, adjusted HR; Ca, noncolorectal cancer; CRC, colorectal cancer; ref, reference. effects associated with alcohol consumption (6 9), which include carcinogenic effects of alcohol metabolites, such as acetaldehyde, alcohol-induced effects on other carcinogens/chemicals present in, for example, tobacco smoke (facilitated absorption, slowed decomposition), and alcohol-induced malabsorption and deficiencies of nutrients, such as folate. Beneficial effects of moderate alcohol consumption have also been discussed in the literature, especially in studies focusing on potential cardiovascular actions. alcohol consumption has been discussed in terms of potential antioxidant, anti-inflammatory, hypotensive, anti-platelet aggregating, insulin sensitivity increasing, HDL-cholesterol elevating, and fibrinolytic effects (31). It is currently unknown how these different effects potentially associated with alcohol consumption may interplay in cancer patients, especially also regarding interactions with chemotherapeutics. Our study suggests that poorer survival associated with alcohol abstaining might be restricted to diabetes patients. A possible explanation could be the positive effect of moderate alcohol consumption on insulin sensitivity (32, 33). alcohol consumption might be positively associated with a protective effect on the cardiovascular system (34) that might be particularly pronounced among patients with diabetes. Because diabetes patients are in general at a higher risk for cardiovascular mortality than patients without diabetes (35), this finding might prove clinically relevant but demands further research. Although we adjusted for the presence of diabetes in our analyses, the observed interactions with age might be partly explained by the higher diabetes prevalence among older people. In wine abstainers, we observed a poorer survival than in consumers of #12 g ethanol/d from wine. Similar effects were reported by Phipps et al. (11) for consumption of 1 6 units wine/wk and by Zell et al. (10) for regular compared with infrequent wine consumption in familial CRC only. In accordance with our results, they found no association between beer or liquor consumption and CRC prognosis (10, 11). The protective effects of wine consumption could be explained by lifestyle differences between wine and abstainers. Consumption of wine could be correlated with living a healthier lifestyle. Another explanation could be beneficial health effects from certain chemicals, such as polyphenols, present in wine. Anticancer activity has, for example, been observed for resveratrol and anthocyanin in vitro (36 38). Resveratrol, which occurs in grape skin, was observed to inhibit tumor initiation, promotion, and progression (36). An antiproliferative effect was seen in vitro and in vivo (37 39). Anthocyanin, which is present in red wine (and also in lower concentrations in white wine and beer), has been reported to have a suppressing effect on colon cancer cells in vitro (40), and phenolic acid and anthocyanin were shown to inhibit colon cancer cell viability and to increase apoptosis (41, 42). Our study has specific strengths and limitations. Strengths include the large population-based study sample, comprehensive follow-up of patients, thorough assessment of recurrence and causes of death, and comprehensive adjustment for potential confounders. Data collection was conducted by trained interviewers resulting in minor proportions of missing data. Moreover, potential selection bias resulting from missing data was further minimized by multiple imputation. Survivor bias arising through varying intervals between diagnosis and interview was minimized with the use of late-entry models. Stage-specific and further subgroup-specific investigations were conducted that addressed several survival outcomes that are

9 ALCOHOL CONSUMPTION AND CRC PROGNOSIS 9 of 10 TABLE 7 Hazard ratios for overall and CRC-specific survival according to recent consumption of different types of beverage, in stage I IV patients 1 Recent beer Recent wine Recent liquor None #12 g/d.12 g/d None #12 g/d.12 g/d None #12 g/d.12 g/d At risk, n Overall survival Events, n ahr (95% CI) 1.09 (0.91, 1.30) 1.00 (ref) 1.01 (0.81, 1.26) 1.21 (1.02, 1.43) 1.00 (ref) 0.95 (0.78, 1.16) 0.92 (0.75, 1.12) 1.00 (ref) 1.27 (0.62, 2.60) CRC-specific survival Events, n ahr (95% CI) 0.99 (0.80, 1.21) 1.00 (ref) 0.97 (0.75, 1.26) 1.22 (1.00, 1.49) 1.00 (ref) 0.96 (0.76, 1.22) 1.03 (0.81, 1.30) 1.00 (ref) 1.44 (0.63, 3.31) 1 Cox proportional hazard regression model was adjusted for age; sex; stage; smoking status; use of statins; use of nonsteroidal anti-inflammatory drugs; use of b-blockers; diabetes mellitus; history of heart failure, myocardial infarction, angina pectoris, or stroke; and history of Ca. Additional adjustment was made for age 3 log(time) and Ca 3 log(time) to account for time-dependent effects. For each model calculation, all 3 beverage type variables (beer, wine, and liquor) were included into the model and analyzed simultaneously. Estimates were calculated on unimputed data. ahr, adjusted HR; Ca, noncolorectal cancer; CRC, colorectal cancer; ref, reference. commonly used in survival analyses and thereby facilitate comparability with previous studies. Certain limitations also have to be considered. Exposure assessment was restricted to the amount of alcohol consumed before diagnosis, without paying attention to drinking pattern or changes in alcohol consumption after diagnosis. Because exposure assessment was conducted after diagnosis and because of the long recall of past alcohol consumption, there is also potential for recall bias. Analyses comparing alcohol consumption before and after diagnosis among 5-y survivors suggested moderate agreement with a tendency toward lowering alcohol consumption after diagnosis. Apart from this, men show a higher overall alcohol consumption and might engage more often in binge drinking than women. However, no effect modification by sex was observed. Effects in women might further be modified by hormonal differences. In general, heavy alcohol consumption could be underreported because of the stigmatization associated with it. Abstaining from alcohol might be a marker for comorbidity in terms of sick quitters (43), or it might be a marker for a presumably healthier lifestyle. Concerning diabetes patients, a controlled alcohol consumption might be a marker for a better compliance with doctors recommendations and therefore better disease management. Although we carefully controlled for a large number of potential confounders, we cannot exclude the possibility of residual confounding by additional factors potentially related to alcohol consumption and prognosis. Although we minimized potential survivor bias by inclusion of late-entry time, its presence cannot be fully excluded. In conclusion, alcohol abstaining and heavy drinking behavior were associated with poorer survival after a CRC diagnosis. The protective effects of lighter alcohol consumption might be restricted to wine consumers only, and associations might differ according to age and presence of diabetes mellitus. Notwithstanding the need of further clarification of the impact of alcohol consumption on CRC survival, our data suggest that heavy might benefit from restricting their alcohol consumption to light/moderate levels. Potential benefits or harms of abstaining could depend on patients overall health status and comorbidities. In recommendations to cancer patients, potential interactions with chemotherapeutic drugs should be taken into account. More research is required to enhance the basis of evidence for both general and personalized recommendations on alcohol consumption for patients with CRC. We thank Ute Handte-Daub, Ansgar Brandhorst, and Petra Bächer for their excellent technical assistance. We also thank the following hospitals and cooperating institutions that recruited patients for this study: Chirurgische Universitätsklinik Heidelberg, Klinik am Gesundbrunnen Heilbronn, St. Vincentiuskrankenhaus Speyer, St. Josefskrankenhaus Heidelberg, Chirurgische Universitätsklinik Mannheim, Diakonissenkrankenhaus Speyer, Krankenhaus Salem Heidelberg, Kreiskrankenhaus Schwetzingen, St. Marienkrankenhaus Ludwigshafen, Klinikum Ludwigshafen, Stadtklinik Frankenthal, Diakoniekrankenhaus Mannheim, Kreiskrankenhaus Sinsheim, Klinikum am Plattenwald Bad Friedrichshall, Kreiskrankenhaus Weinheim, Kreiskrankenhaus Eberbach, Kreiskrankenhaus Buchen, Kreiskrankenhaus Mosbach, Enddarmzentrum Mannheim, Kreiskrankenhaus Brackenheim, and Cancer Registry of Rhineland- Palatinate, Mainz. We thank the institutes of pathology for their support by providing tumor samples: Institut für Pathologie, Universitätsklinik Heidelberg; Institut für Pathologie, Klinikum Heilbronn; Institut für Angewandte Pathologie, Speyer; Pathologisches Institut, Universitätsklinikum Mannheim; Institut für Pathologie, Klinikum Ludwigshafen; Institut für Pathologie, Klinikum Stuttgart; and Institut für Pathologie, Klinikum Ludwigsburg. We also thank Matthias Kloor, Esther Herpel, and the tissue bank of the National Center

10 10 of 10 WALTER ET AL. for Tumor Diseases, Heidelberg, for microsatellite instability analyses and storage and processing of the tissue samples. The authors responsibilities were as follows VW, JC-C, MH, and H Brenner: designed the research; VW: performed the statistical analysis and wrote the manuscript; VW and H Brenner: had primary responsibility for the final content; and all authors: conducted the research, contributed to the interpretation of data analysis results, and read and approved the final manuscript. None of the authors had a conflict of interest. REFERENCES 1. Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin D, Forman D, Bray F. GLOBOCAN 2012 v1.0, Cancer incidence and mortality worldwide: IARC CancerBase No. 11. International Agency for Research on Cancer. Lyon (France): IARC; De Angelis R, Sant M, Coleman MP, Francisci S, Baili P, Pierannunzio D, Trama A, Visser O, Brenner H, Ardanaz E, et al. Cancer survival in Europe by country and age: results of EUROCARE 5- a population-based study. Lancet Oncol 2014;15: International Agency for Research on Cancer (IARC). IARC monographs on the evaluation of carcinogenic risks to humans: Volume 96. Alcoholic beverage consumption and ethyl carbamate (urethane). Lyon (France): IARC; Fedirko V, Tramacere I, Bagnardi V, Rota M, Scotti L, Islami F, Negri E, Straif K, Romieu I, La Vecchia C, et al. Alcohol drinking and colorectal cancer risk: an overall and dose-response meta-analysis of published studies. Ann Oncol 2011;22: Cai S, Li Y, Ding Y, Chen K, Jin M. Alcohol drinking and the risk of colorectal cancer death: a meta-analysis. Eur J Cancer Prev 2014;23: Eriksson CJ. The role of acetaldehyde in the actions of alcohol (update 2000). Alcohol Clin Exp Res 2001;25(5 Suppl ISBRA):15S 32S. 7. Hamid A, Wani NA, Kaur J. New perspectives on folate transport in relation to alcoholism-induced folate malabsorption: association with epigenome stability and cancer development. FEBS J 2009;276: Pöschl G, Seitz HK. Alcohol and cancer. Alcohol 2004;39: Pöschl G, Stickel F, Wang XD, Seitz HK. Alcohol and cancer: genetic and nutritional aspects. Proc Nutr Soc 2004;63: Zell JA, McEligot AJ, Ziogas A, Holcombe RF, Anton-Culver H. Differential effects of wine consumption on colorectal cancer outcomes based on family history of the disease. Nutr Cancer 2007;59: Phipps AI, Baron J, Newcomb PA. Prediagnostic smoking history, alcohol consumption, and colorectal cancer survival: the Seattle Colon Cancer Family Registry. Cancer 2011;117: Pelser C, Arem H, Pfeiffer RM, Elena JW, Alfano CM, Hollenbeck AR, Park Y. Prediagnostic lifestyle factors and survival after colon and rectal cancer diagnosis in the National Institutes of Health (NIH) AARP Diet and Health Study. Cancer 2014;120: Park SM, Lim MK, Shin SA, Yun YH. Impact of prediagnosis smoking, alcohol, obesity, and insulin resistance on survival in male cancer patients: National Health Insurance Corporation Study. J Clin Oncol 2006;24: Fung TT, Kashambwa R, Sato K, Chiuve SE, Fuchs CS, Wu K, Giovannucci E, Ogino S, Hu FB, Meyerhardt JA. Post diagnosis diet quality and colorectal cancer survival in women. PLoS One 2014;9: e Araújo RF Jr., Lira GA, Guedes HG, Cardoso MA, Cavalcante FJ, Araujo AL, Ramos CC, de Araujo AA. Lifestyle and family history influence cancer prognosis in Brazilian individuals. Pathol Res Pract 2013;209: Brenner H, Chang-Claude J, Seiler CM, Rickert A, Hoffmeister M. Protection from colorectal cancer after colonoscopy: a populationbased, case-control study. Ann Intern Med 2011;154: Jansen L, Hoffmeister M, Arndt V, Chang-Claude J, Brenner H. Stagespecific associations between beta blocker use and prognosis after colorectal cancer. Cancer 2014;120: Hoffmeister M, Jansen L, Rudolph A, Toth C, Kloor M, Roth W, Blaker H, Chang-Claude J, Brenner H. Statin use and survival after colorectal cancer: the importance of comprehensive confounder adjustment. J Natl Cancer Inst 2015;107:djv Walter V, Jansen L, Hoffmeister M, Ulrich A, Chang-Claude J, Brenner H. Smoking and survival of colorectal cancer patients: populationbased study from Germany. Int J Cancer 2015;137: Jansen L, Hoffmeister M, Chang-Claude J, Koch M, Brenner H, Arndt V. Age-specific administration of chemotherapy and long-term quality of life in stage II and III colorectal cancer patients: a population-based prospective cohort. Oncologist 2011;16: Scherz H, Souci SW, editors. Food composition and nutrition tables. 5th ed. Stuttgart (Germany): Medpharm Scientific Publishers; Desquilbet L, Mariotti F. Dose-response analyses using restricted cubic spline functions in public health research. Stat Med 2010;29: R: a language and environment for statistical computing. Version Vienna (Austria): The R Foundation for Statistical Computing; van Buuren S, Groothuis-Oudshoorn K. Mice: multivariate imputation by chained equations in R. J Stat Soft 2011;45: Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F, Fedirko V, Scotti L, Jenab M, Turati F, Pasquali E, et al. Alcohol consumption and sitespecific cancer risk: a comprehensive dose-response meta-analysis. Br J Cancer 2015;112: Zhang C, Zhong M. Consumption of beer and colorectal cancer incidence: a meta-analysis of observational studies. Cancer Causes Control 2015;26: Ben Q, Wang L, Liu J, Qian A, Wang Q, Yuan Y. Alcohol drinking and the risk of colorectal adenoma: a dose response meta-analysis. Eur J Cancer Prev 2015;24: Zhu JZ, Wang YM, Zhou QY, Zhu KF, Yu CH, Li YM. Systematic review with meta-analysis: alcohol consumption and the risk of colorectal adenoma. Aliment Pharmacol Ther 2014;40: Jin M, Cai S, Guo J, Zhu Y, Li M, Yu Y, Zhang S, Chen K. Alcohol drinking and all cancer mortality: a meta-analysis. Ann Oncol 2013;24: Rehm J, Irving H, Ye Y, Kerr WC, Bond J, Greenfield TK. Are lifetime abstainers the best control group in alcohol epidemiology?: on the stability and validity of reported lifetime abstention. Am J Epidemiol 2008;168: Chiva-Blanch G, Arranz S, Lamuela-Raventos RM, Estruch R. Effects of wine, alcohol and polyphenols on cardiovascular disease risk factors: evidences from human studies. Alcohol 2013;48: Baliunas DO, Taylor BJ, Irving H, Roerecke M, Patra J, Mohapatra S, Rehm J. Alcohol as a risk factor for type 2 diabetes: a systematic review and meta-analysis. Diabetes Care 2009;32: Koppes LL, Dekker JM, Hendriks HF, Bouter LM, Heine RJ. alcohol consumption lowers the risk of type 2 diabetes: a meta-analysis of prospective observational studies. Diabetes Care 2005;28: Rajpathak SN, Freiberg MS, Wang C, Wylie-Rosett J, Wildman RP, Rohan TE, Robinson JG, Liu S, Wassertheil-Smoller S. Alcohol consumption and the risk of coronary heart disease in postmenopausal women with diabetes: Women s Health Initiative Observational Study. Eur J Nutr 2010;49: Morrish NJ, Wang SL, Stevens LK, Fuller JH, Keen H. Mortality and causes of death in the WHO Multinational Study of Vascular Disease in Diabetes. Diabetologia 2001;44(Suppl 2):S Jang M, Cai L, Udeani GO, Slowing KV, Thomas CF, Beecher CW, Fong HH, Farnsworth NR, Kinghorn AD, Mehta RG, et al. Cancer chemopreventive activity of resveratrol, a natural product derived from grapes. Science 1997;275: Schneider Y, Vincent F, Duranton B, Badolo L, Gosse F, Bergmann C, Seiler N, Raul F. Anti-proliferative effect of resveratrol, a natural component of grapes and wine, on human colonic cancer cells. Cancer Lett 2000;158: Szende B, Tyihak E, Kiraly-Veghely Z. Dose-dependent effect of resveratrol on proliferation and apoptosis in endothelial and tumor cell cultures. Exp Mol Med 2000;32: Schneider Y, Duranton B, Gosse F, Schleiffer R, Seiler N, Raul F. Resveratrol inhibits intestinal tumorigenesis and modulates hostdefense-related gene expression in an animal model of human familial adenomatous polyposis. Nutr Cancer 2001;39: Kamei H, Hashimoto Y, Koide T, Kojima T, Hasegawa M. Anti-tumor effect of methanol extracts from red and white wines. Cancer Biother Radiopharm 1998;13: Yi W, Fischer J, Akoh CC. Study of anticancer activities of muscadine grape phenolics in vitro. J Agric Food Chem 2005;53: Yi W, Fischer J, Krewer G, Akoh CC. Phenolic compounds from blueberries can inhibit colon cancer cell proliferation and induce apoptosis. J Agric Food Chem 2005;53: Shaper AG, Wannamethee G, Walker M. Alcohol and mortality in British men: explaining the U-shaped curve. Lancet 1988;2: