The role of body mass index, physical activity, and diet in colorectal cancer recurrence and survival: a review of the literature 1 4

Transcription

1 Review Article The role of body mass index, physical activity, and diet in colorectal cancer recurrence and survival: a review of the literature 1 4 Alina Vrieling and Ellen Kampman ABSTRACT The role of dietary and other lifestyle factors in colorectal cancer recurrence and survival is largely unknown. We conducted a review to summarize the evidence from epidemiologic studies that examined the association of body mass index (BMI), physical activity, and nutrition with colorectal cancer recurrence and survival. We searched MEDLINE and EMBASE for relevant epidemiologic studies published up to March 2010 by using MeSH terms and related key words. We identified 36 articles that were based on 31 independent studies on BMI (n ¼ 21), physical activity (n ¼ 6), or nutrition (n ¼ 12) in relation to colorectal cancer recurrence and survival. Studies were generally based on follow-up of cases in existing patient series, case-control or cohort studies, or chemotherapy trials. BMI, physical activity, and nutrition mostly referred to the time at or before. Only 10 studies assessed BMI (n ¼ 1), physical activity (n ¼ 4), or nutrition (n ¼ 5) after. There may be an association between higher BMI and body fatness before or at the time of and a higher all-cause mortality or colorectal cancer specific mortality or recurrence, although results may differ by sex, tumor location, and molecular subtype. There may be a relation between higher leisure-time physical activity after and a lower all-cause or colorectal cancer specific mortality. For dietary factors, statistically significant associations were only shown for single foods, nutrients, and dietary patterns in single studies. In conclusion, only a paucity of data is available on the effect of dietary and other lifestyle factors on colorectal cancer recurrence and survival. Thus far, no clear conclusions can be drawn. Future studies are warranted, particularly on post BMI and diet. Am J Clin Nutr 2010;92: INTRODUCTION Colorectal cancer is the third most common cancer worldwide (1). The number of patients who survive colorectal cancer is increasing, partly because of earlier detection and more-effective treatment methods. The age-adjusted 5-y relative survival rate over the period increased from 48% to 54% in Europe (2), and a period analysis over 2000 to 2002 showed an age-adjusted 5-y relative survival rate of 56% for Europe and 66% for the United States (3). Worldwide, colon and rectum cancers accounted for 1 million new cases in 2002, and an estimated 2.8 million persons diagnosed with colorectal cancer were alive 5 y of (1). After, individuals with colorectal cancer remain at an increased risk of colorectal cancer recurrence, secondary cancer, other chronic diseases, and cancer death. Their clinical outcome is dependent on several tumor characteristics, such as tumor stage, grade, lymph node involvement, and metastases (4). Also, differences in methods of treatment influence the cancer prognosis. However, there is still a large variability in the clinical outcome of individuals with morphologic identical colorectal cancers who receive comparable therapeutic strategies. This variability may be due to differences in nutritional or other lifestyle factors. Dietary and lifestyle factors in relation to colorectal cancer incidence have been extensively investigated. The second World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) expert report (5) showed that higher body fatness, abdominal fatness, adult attained height, red meat and processed-meat intake, and lower physical activity are associated with a higher colorectal cancer risk. Higher alcohol intake is associated with a higher colorectal cancer risk in men, and probably also in women. Higher dietary fiber, garlic, milk, and calcium intakes are probably associated with a lower colorectal cancer risk. Further, there is limited suggestive evidence for a protective effect of nonstarchy vegetables, fruit, fish, and foods that contain folate, vitamin D, or selenium. In contrast, higher intake of cheese and foods containing iron, animal fats, or sugars may lower colorectal cancer risk. Compared with the vast amount of literature with respect to colorectal cancer incidence, there is only a paucity of data about the effect of dietary and lifestyle factors on colorectal cancer recurrence and survival. This information is important because many individuals who have been diagnosed with colorectal cancer are motivated and have been shown to adjust their dietary and physical activity patterns (6 8). This review will summarize the evidence for the role of body mass index (BMI; in kg/m 2 ), physical activity, and diet before, at the time of, and after co- 1 From the Division of Human Nutrition, Wageningen University, World Cancer Research Fund Project Group, Wageningen, Netherlands. 2 Supported by the World Cancer Research Fund. 3 Current address for AV: Division of Cancer Epidemiology, German Cancer Research Center, Heidelberg, Germany. 4 Address correspondence to E Kampman, Division of Human Nutrition, Wageningen University, PO Box 8129, 6700 EV, Wageningen, Netherlands. ellen.kampman@wur.nl. Received November 30, Accepted for publication June 24, First published online July 14, 2010; doi: /ajcn Am J Clin Nutr 2010;92: Printed in USA. Ó 2010 American Society for Nutrition 471

2 472 VRIELING AND KAMPMAN lorectal cancer in colorectal cancer recurrence and survival. METHODS A search in the MEDLINE ( and EMBASE ( databases was conducted up to March 2010 by using a combination of MeSH terms (colorectal neoplasms, recurrence, disease-free survival, mortality, motor activity, body weight, BMI, waist circumference, waist-hip ratio, body fat distribution, diet, diet therapy, food and beverages, and nutritional sciences) and related key words in titles or abstracts. Retrieved references were checked independently by the 2 investigators, and discordances were discussed and resolved. Relevant articles were obtained and included in the review if they reported on epidemiologic studies in colorectal cancer survivors with respect to BMI, physical activity, or diet in relation to allcause mortality, colorectal cancer specific mortality, or colorectal cancer recurrence. Reference lists of relevant articles were checked to identify any additional epidemiologic studies. Only articles written in the English language were included. Studies were excluded from this review if they solely investigated associations of BMI, physical activity, or diet with intermediate endpoints of disease or with tumor characteristics and not with allcause mortality, colorectal cancer specific mortality, or colorectal cancer recurrence. Weight loss and underweight were not taken into consideration because these factors are more likely to be of a disease-related cause than a lifestyle-related cause of disease outcome. Information extracted from each study included the first author, year,, sample (size and key characteristics), followup time, variable of interest, assessment method, outcome, contrast, risk estimates with 95% CIs, and factors that were adjusted for in the analysis. Because of the relatively small number of relevant studies and the large heterogeneity in type of exposure, timing of exposure assessment, and disease outcomes, a formal meta-analysis could not be conducted. The studies on BMI, physical activity, or diet in relation to all-cause mortality, colorectal cancer specific mortality, or colorectal cancer recurrence were described separately according to the time period these factors were assessed [ie, before (pre), at, or after (post)]. RESULTS The initial search retrieved 2801 articles. We identified a total of 36 relevant articles that were based on 31 independent epidemiologic studies in colorectal cancer survivors that reported on BMI, physical activity, or dietary factors in relation to all-cause mortality, colorectal cancer specific mortality, or colorectal cancer recurrence (Figure 1). These studies are discussed in more detail in this review. BMI The association between BMI and recurrence or survival of colorectal cancer has been investigated in 21 observational studies described in 24 articles (9 32) (Table 1). Six studies evaluated BMI before (9 16), 14 studies investigated BMI at (17 31), and one study focused on BMI during and after chemotherapy (32). Nine studies investigated selfreported BMI (9, 10, 13 18, 22, 28, 32), 7 studies evaluated FIGURE 1. Overview of the number of retrieved relevant studies on BMI, physical activity (PA), and diet in relation to colorectal all-cause mortality, cancer-specific mortality, or recurrence by time of exposure assessment. measured BMI (11, 12, 19 21, 24, 25), and 4 studies obtained BMI from medical records (23, 26, 27, 29, 31). One study did not provide any information about BMI assessment (30). Some studies additionally reported on measured body weight (12, 17), waist circumference (12), percentage body fat (12), and visceral and subcutaneous fat areas (24, 30). BMI before cancer Pre BMI and all-cause or cancer-specific mortality was investigated by 6 studies that were based on existing casecontrol (9, 10) or cohort studies (11 16) that were originally designed to investigate dietary and lifestyle factors in relation to cancer incidence. Except for one study (11), all other studies were adjusted for the tumor stage. In 3 studies, including 410 colon cancer cases (9), 1882 colorectal cancer cases (11), and 526 colorectal cancer cases (12), no association of higher BMI with all-cause mortality (9, 11, 12) or colorectal cancer specific mortality (12) was observed. Higher body fat, weight, and waist circumference were associated with a higher colorectal cancer specific mortality but not with all-cause mortality (12). One study in 633 colorectal cancer cases reported a higher risk of colon cancer-specific mortality [hazard ratio (HR): 2.1: 95% CI: 1.1, 3.8] and all-cause mortality (HR: 1.5: 95% CI: 1.0, 2.1) in obese compared with normal weight postmenopausal women (10). The risk of rectal cancer-specific mortality was not increased, and no association in overweight women was shown (10). Two studies in 647 colon cancer cases (13 15) and 546 colorectal cancer cases (16), respectively, showed that the association between pre BMI and cancer-specific or allcause mortality may depend on molecular characteristics of the tumor. Higher BMI was associated with higher all-cause mortality in individuals with tumors positive for fatty acid synthase expression (13), p27 (cyclin-dependent kinase inhibitor 1B) expression (14), p21 (cyclin-dependent kinase inhibitor 1A) expression (15), or stathmin or oncoprotein-18 expression (16). The association between higher BMI and higher cancer-specific mortality only reached statistical significance for individuals with tumors positive for p21 expression (15) and stathmin or oncoprotein-18 expression (16).

3 LIFESTYLE FACTORS IN COLORECTAL CANCER SURVIVAL 473 TABLE 1 Epidemiologic studies of BMI (in kg/m 2 ) and colorectal cancer recurrence or survival 1 characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment BMI before Slattery, 1989, United States (9) Doria-Rose, 2006, United States (10) Park, 2006, Korea (11) Haydon, 2006, Australia (12) Ogino, 2009, United States (13) 410 CC cases (M, F), aged y, identified through the Utah Cancer Registry and part of 2 case-control studies CRC cases (F, post), aged y, identified through the Wisconsin Cancer Reporting System and part of a casecontrol study 280 and 147 CRC 1882 CRC patients (M) NHICS 555 and 524 CRC 526 CRC patients (M, F; stages I IV), aged y, diagnosed within the MCCS cohort study 208 and 181 CRC 647 CC patients (M, F; stages I IV) NHS and HPFS cohort studies 279 and 160 CC Diagnosis: Self-reported BMI 5 y Median: 83 mo (range: mo) (study 1) or 2 y (study 2) before Diagnosis: Self-reported BMI 5 y Mean: 9.4 y (range: before interview, y) which was conducted a median of 13 mo (range: 4 28 mo) after Diagnosis: Measured BMI at FU ended in 2004 baseline (median: 3.8 y) Diagnosis: Measured percentage Median: 5.5 y body fat, body weight, waist circumference, and BMI at baseline ( ) Diagnosis: 1976/ FU up to June 2006 Self-reported BMI from biennial questionnaire immediately preceding CC BMI, Q4 vs Q1 (cutoffs: NA) CC-specific mortality Obesity (BMI 30) and overweight (BMI of ) vs normal weight (BMI of ) RC-specific mortality Obesity Overweight Obesity Overweight 0.99 (0.90, 1.08) Age, sex, stage, and year of 2.1 (1.1, 3.8) Age, stage, PMH use, 1.5 (0.9, 2.5) and smoking 0.6 (0.2, 1.6) 1.0 (0.5, 2.1) 1.5 (1.0, 2.1) 1.2 (0.9, 1.6) BMI 25 vs, (0.81, 1.24) Age, alcohol consumption, fasting serum glucose concentration, cholesterol concentration, physical activity, food preference, blood pressure, and other comorbidities CRC-specific mortality Body fat per 10% 1.33 (1.04, 1.71) Age, sex, and tumor Body weight per 10 kg 1.15 (1.02, 1.29) stage Waist circumference 1.20 (1.05, 1.37) per 10 cm BMI per 5 kg/m (0.98, 1.35) CC-specific mortality BMI 27.5 vs, (0.48, 18.0) for Age, year of, FASN1 2 sex, tumor location, BMI 27.5 vs,27.5 BMI 27.5 vs,27.5 BMI 27.5 vs, (0.63, 1.30) for FASN (1.14, 14.8) for FASN (0.65, 1.14) for FASN2 P int ¼ stage, tumor grade, K-ras, BRAF, p53, MSI, and CIMP (Continued)

4 474 VRIELING AND KAMPMAN TABLE 1 (Continued) Ogino, 2009, United States (14) Ogino, 2009, United States (15) Ogino, 2009, United States (16) BMI at Tartter, 1984, United States (17) characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment 630 CC patients (M, F; stages I IV) NHS and HPFS cohort studies 272 and 160 CC 647 CC patients (M, F; stages I IV) NHS and HPFS cohort studies 279 and 162 CC 546 CRC patients (M, F; stages I IV) NHS and HPFS cohort studies 236 and 149 CRC Diagnosis: 1976/ FU up to June 2006 Diagnosis: 1976/ FU up to June 2006 Diagnosis: 1976/ FU up to June 2006 Self-reported BMI from biennial questionnaire immediately preceding CC Self-reported BMI from biennial questionnaire immediately preceding CC Self-reported BMI from biennial questionnaire immediately preceding CC CC-specific mortality BMI 30 vs, (0.84, 6.01) for p27-nuclear1 BMI 30 vs, (0.72, 2.14) for p27-altered BMI 30 vs, (1.49, 6.32) for p27-nuclear1 BMI 30 vs, (0.71, 1.65) for p27-altered Pint ¼ CC-specific mortality BMI 30 vs, (2.28, 15.0) for p211 BMI 30 vs, (0.60, 1.83) for p212 BMI 30 vs, (1.65, 6.98) for p211 BMI 30 vs, (0.62, 1.43) for p212 P int ¼ CRC-specific mortality BMI 30 vs, (1.18, 4.69) for STMN11 BMI 30 vs, (0.24, 1.07) for STMN12 Pint ¼ BMI 30 vs, (1.13, 3.32) for STMN11 BMI 30 vs, (0.40, 1.28) for STMN12 Age, year of, sex, family history CRC, tumor location, stage, grade, K-ras, BRAF, PIK3CA, p53, cyclin D1, b-catenin, COX-2, FASN, LINE-1 methylation, MSI, and CIMP Age, year of, sex, family history CRC, tumor location, stage, grade, K-ras, BRAF, PIK3CA, p53, cyclin D1, LINE-1 methylation, MSI, and CIMP Age, year of, sex, family history CRC, tumor location, stage, grade, K-ras, BRAF, PIK3CA, p53, p21, p27, cyclin D1, b-catenin, COX- 2, FASN, LINE-1 methylation, MSI, and CIMP 279 CC patients (M, F; Dukes stage B2, C1, C2) consecutively diagnosed and undergoing curative resection 65% disease-free at 5 y Diagnosis Self-reported preoperative weight 5-y recurrence-free rate and BMI Recurrence Weight greater than the median vs less than the median (cutoffs: NA) BMI greater than the median vs less than the median (cutoffs: NA).1, P ¼ for M P ¼ for F P ¼ for M.1, P ¼ for F (Continued)

5 LIFESTYLE FACTORS IN COLORECTAL CANCER SURVIVAL 475 TABLE 1 (Continued) Dray, 2003, France (18) Meyerhardt, 2003, United States (19) Meyerhardt, 2004, United States (20) Dignam, 2006, United States (21) characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment 148 CRC patients (M, F; Dukes stage A D), aged y, identified in the Cote d Or area and part of a case-control study 46 and 70 at 5 and 10 y, respectively 3759 CC patients (M, F; stages II and III) treated within a randomized adjuvant chemotherapy trial (INT-0089) 1600 and 1289 CC recurrences 1688 RC patients (M, F; stages II and III) treated within a randomized adjuvant chemotherapy and radiotherapy trial (INT-0114) Deaths and RC recurrences: NA 4288 CC patients (M, F; Dukes stage B and C) treated within 2 randomized adjuvant chemotherapy trials (ABP C-04 and C-05) 1697, 1159 CC, and 1286 CC recurrences or secondary primary tumors Diagnosis: Self-reported BMI at 5- or 10-y survival time of interview; 3 mo of Inclusion: Measured BMI at day 1 Median: 9.4 y (max: of chemotherapy 12.7 y) Inclusion: Measured BMI at day 1 Median: 9.9 y (max: of chemotherapy 11.8 y) Inclusion: Measured BMI at day 1 of chemotherapy BMI (data not shown) Age, sex, tumor stage, tumor location, and interaction terms Obesity (BMI 30) vs normal weight (BMI of ) Recurrence Recurrence Overweight (BMI of and ) vs normal weight Obesity (BMI 30) vs Local recurrence normal weight (BMI of ) Recurrence Recurrence or death Local recurrence Median: 11.2 y Overweight (BMI of ) and overweight (BMI of ) vs normal weight CC-specific mortality Severe obesity (BMI 35) vs normal Recurrence weight (BMI of Recurrence or death ) Obesity (BMI of ) Overweight (BMI of ) 1.34 (1.07, 1.67) for F, P int ¼ 0.02 for M 1.24 (0.98, 1.59) for F, P int ¼ (1.23, 5.42) for F,50 y for M for M and F for M and F Age, race, baseline performance status, bowel obstruction, bowel perforation, Duke stage, presence of peritoneal implants, predominant macroscopic pathologic feature, and completion chemotherapy for M and F Age, race, baseline 1.61 (1.00, 2.59) for M, performance status, P for trend ¼ 0.11 bowel obstruction, for F for M and F for M and F for M and F 1.66 (1.07, 2.56) for M for F extent of bowel-wall invasion, no. of positive lymph nodes, and operation type 1.36 (1.06, 1.73) Age, sex, race, 1.28 (1.04, 1.57) performance status, 1.38 (1.10, 1.73) number of positive 1.27 (1.05, 1.53) lymph nodes, presence of bowel obstruction, and treatment (Continued)

6 476 VRIELING AND KAMPMAN TABLE 1 (Continued) Zell, 2007, United States (22) Moghimi-Dehkordi, 2008, Iran (23) Moon, 2008, Korea (24) Tian, 2008, China (25) Asghari-Jafarabadi, 2009, Iran (26) characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment 141 familial (at least one first- degree relative) and 358 sporadic CRC patients (M, F); part of the UCI 199 and 95 CRC (of 106 recorded ) 1127 CRC patients (M, F; stage I IV; average age at : 54 y), registered in a cancer registry 21% of 1127 patients with survival info ( 237 patients) 161 CRC patients (M, F; average age: 66 y) selected from a surgical database Deaths: NA 39 CRC patients (M, F; stages I IV) who underwent surgery in 4 provincial-level hospitals in Fuzhou, China CRC patients (M, F; stages I IV; average age at : 54 y), registered in a cancer registry Deaths: NA Diagnosis: Self-reported BMI at 7 9 y Included: 2002 January 2007 Median: 7.9 y BMI at from medical records BMI per 1-kg/m 2 increase BMI per 1-kg/m 2 increase Overweight (BMI of ) and obesity (BMI.30) vs normal weight (BMI of ) 1.01 (0.94, 1.08) for familial CRC 0.96 (0.93, 0.99) for sporadic CRC Age, sex, treatment with surgery, radiation, and chemotherapy 1.41 (P ¼ 0.65) Family history of 1.80 (P ¼ 0.15) cancer, first treatment, tumor grade, tumor size, extent of wall penetration, regional lymph node metastasis, distant metastasis, and pathologic stage Surgery: Preoperative BMI from Overweight (BMI 25).1.00 (P ¼ 0.210) Not clearly stated (age,.1 y database; visceral Recurrence or death vs normal weight.1.00 (P ¼ 0.064) sex, tumor size, and obesity before (BMI,25) lymph node surgery Ratio of visceral.1.00 (P ¼ 0.241) metastasis) Recurrence or death fat area and 1.98 (1.02, 3.87) subcutaneous fat area greater than the median vs less than the median (cutoff values: NA) Diagnosis: January June y survival Measured BMI 20 d after surgery Included: BMI at from medical records BMI.25 vs (0.08, 19.4) Age, sex, and tumor stage Overweight (BMI of ) and obesity (BMI.30) vs normal weight (BMI of ) 2002 October (0.14, 0.73) Mean: 9.3 y (CC) and 11.3 y (RC) 239 CC: CC: sex, alcohol, inflammatory bowel 0.71 (0.25, 2.03) 291 RC: 0.40 (0.16, 0.97) 0.76 (0.16, 3.57) disease, tumor stage, grade, and size RC: personal history of cancer, treatment, and tumor stage (Continued)

7 LIFESTYLE FACTORS IN COLORECTAL CANCER SURVIVAL 477 TABLE 1 (Continued) Hines, 2009, United States (27) You, 2009, Taiwan (28) Ballian, 2010, United States (29) Guiu, 2010, France (30) characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment 496 CC patients (M, F; stages I IV; average age: 67 y) who underwent surgery at the University of Alabama at Birmingham Hospital RC patients (M, F; stages I IV) registered in the Colorectal Section Tumor Registry, Chang-Gung Memorial Hospital, consecutively undergoing resection 230 RC patients (M, F; stages I III) who underwent protectomy 44 recurrences; : NA 80 CRC patients who received bevacizumab-based chemotherapeutic treatment and 40 CRC patients who received chemotherapy alone (M, F; metastatic CRC) 58 and 18, respectively Surgery: BMI at day of surgery FU up to 2008 from medical records (minimum: 5 y) Included: Self-reported BMI at Median: 74.5 mo day of admission (range: mo) Overweight/obese (BMI 25) vs normal weight (,18.5) Local recurrence Obesity (BMI 30) and overweight (BMI of ) vs underweight Recurrence or death (BMI,18.5) 0.77 (0.61, 0.97) Age, race, comorbidity, for stages I III only 0.58 (0.37, 0.90) for stage IV tumor stage, grade, and bowel obstruction 4.71 (1.01, 21.9) Age, sex, resection 3.24 (0.78, 13.4) for margin, tumor size, lower RC grade, and stage for upper RC for lower and upper RC Surgery: BMI at day of surgery Recurrence BMI 30 vs,30 P ¼ 0.49 Age, sex, tumor stage at P ¼ 0.63 Median: 23 mo (range: mo) Included: Visceral fat area, subcutaneous fat area, and BMI before Median: 24 mo (range: 3 70) and 30 mo (range: 4 84), respectively chemotherapy initiation Visceral fat area greater than or equal to the median vs less than the median Subcutaneous fat area greater than or equal to the median vs less than the median 2.88 (1.13, 7.32) (bevacizumabbased), (chemotherapy alone) (bevacizumabbased, chemotherapy alone) BMI 23.6 vs,23.6 (bevacizumabbased, chemotherapy alone), tumor stage after resection, postoperative complications, margin status, tumor grade, and preoperative chemoradiotherapy Performance status, visceral fat area, and subcutaneous fat area (Continued)

8 478 VRIELING AND KAMPMAN TABLE 1 (Continued) characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment Pathak, 2010, United Kingdom (31) BMI after Meyerhardt, 2008, United States (32) 73 CRC patients with liver metastases (M, F; median age: 65 y) who underwent hepatectomy in the Aintree University Hospital Deaths: NA 1053 CRC patients (M, F; stage III) treated within a postoperative adjuvant chemotherapy trial (CALGB 89803) 261, 369 recurrences and, and 338 recurrences Surgery: BMI from medical records Inclusion: Self-reported BMI Median: 5.3 y midway through and 6 mo after completion of adjuvant chemotherapy (14 and 4 mo after surgical resection, respectively); weight change Overweight (BMI 25) vs normal weight (BMI,25) Recurrence or death Class II to III obesity (BMI 35) or class I obesity (BMI of ) or overweight Recurrence (BMI of ) vs normal weight (BMI of ) 5 kg weight gain vs 62 kg weight gain No adjustment 1.24 (0.84, 1.83) Age, sex, depth of 1.00 (0.72, 1.40) invasion through 0.81 (0.59, 1.11) bowel wall, no. of positive lymph nodes, clinical perforation at time of surgery, bowel obstruction at time of surgery, baseline performance status, treatment group, time between first and second questionnaire, timevarying BMI, smoking status at second questionnaire, and physical activity 1 RR, relative risk; CC, colon cancer; Q, quartile; NA, not available; CRC, colorectal cancer; post, postmenopausal; PMH, postmenopausal hormone; NHICS, National Health Insurance Corporation Study; FU, follow-up; MCCS, Melbourne Collaborative Cohort Study; NHS, Nurses Health Study; HPFS, Health Professionals Follow-Up Study; FASN, fatty acid synthase; 1, positive; 2, negative; K-ras, Kirsten rat sarcoma viral oncogene homolog; BRAF, v-raf murine sarcoma viral oncogene homolog B1; MSI, microsatellite instability; CIMP, CpG island methylator phenotype; PIK3CA, phosphoinositide-3-kinase, catalytic, alpha polypeptide; COX-2, cyclooxygenase-2; LINE-1, long interspersed nucleotide element-1; STMN1, stathmin 1; max, maximum; RC, rectal cancer; ABP, National Surgical Adjuvant Breast and Bowel Project; UCI, University of California Irvine CRC gene-environment study; CALGB, Cancer and Leukemia Group B; Pint, P for interaction. 2 Only adjusted for stage because of 7 total.

9 LIFESTYLE FACTORS IN COLORECTAL CANCER SURVIVAL 479 BMI at cancer Fourteen studies investigated BMI at the time of colorectal cancer (17 31). The first study on BMI in relation to colon cancer recurrence was published in 1984 and conducted in 279 consecutively diagnosed colon cancer patients who were undergoing curative resection (17). This study showed a higher risk of recurrence in women but not in men above compared with below the median BMI and did not adjust for any other relevant factors. One study that was based on an existing casecontrol study that included 148 colorectal cancer patients showed no relation between BMI at and 5 or 10 y of survival (18). In a study initially set up as a case-case study, higher BMI was associated with lower all-cause mortality in 358 sporadic colorectal cancer cases but not in 141 familial colorectal cancer cases (22). One study conducted in Iran was reported twice (23, 26) and first showed analyses for colorectal cancer patients combined (23) and later showed analyses for colon and rectal cancer cases, separately (26). Although the follow-up time for the colon and rectal cancer cases was somewhat longer than for the colorectal cancer cases combined, the number of colon and rectal cases included in the analyses was lower than the number of colorectal cancer cases, partly because of differential adjustment for confounding factors. Overweight and obesity were nonstatistically significantly associated with higher all-cause mortality in colorectal cancer cases combined (23). was significantly lower in overweight and nonsignificantly lower in obese colon and rectal cancer cases separately compared with normal weight colon and rectal cancer cases (26). In 39 colorectal cancer (25) and 230 rectal cancer patients (29), overweight (25) and obesity (29) were not associated with all-cause mortality (25, 29) or recurrence (29). No association for overweight was shown in 73 colorectal cancer patients with liver metastases (31), whereas in another study in stage IV colon cancer patients, a protective effect on all-cause mortality was shown (27). In a small study in 161 colorectal cancer patients, a high compared with low ratio of visceral fat and subcutaneous fat area was associated with a higher risk of recurrence or all-cause mortality (HR: 1.98; 95% CI: 1.02, 3.87), whereas no association for overweight was shown (24). A recent study in 80 colorectal cancer patients who received a bevacizumab-based chemotherapeutic treatment confirmed that a visceral fat area above the median compared with below the median was associated with a higher risk of allcause mortality (HR: 2.88; 95% CI: 1.13, 7.32) (30). No associations with the subcutaneous fat area and BMI were observed, and none of these 3 factors were associated with all-cause mortality in 40 colorectal cancer patients who received chemotherapy alone (30). A study in 1873 rectal cancer patients reported a higher risk of local recurrence in obese compared with underweight patients with lower rectal cancer (HR: 4.71: 95% CI: 1.01, 21.9) but not in patients with upper rectal cancer (28). No association of obesity with recurrence and death was observed for both patient groups (28). Three large adjuvant chemotherapy trials investigated BMI measured at day 1 of chemotherapy (19 21) in relation to cancer recurrence and survival in 3759 (19) and 4288 (21) patients with colon cancer and 1688 patients with rectal cancer (20). Analyses were extensively adjusted for several prognostic variables, including the tumor stage, bowel obstruction at the time of surgery, and functional and physical performance status. In one study, obese women (BMI 30) were at a higher risk of all-cause mortality (HR: 1.34; 95% CI: 1.07, 1.67) and at a nonsignificantly higher risk of colon cancer recurrence (HR: 1.24; 95% CI: 0.98, 1.59) compared with normal weight women, particularly those,50 y of age (HR: 2.58; 95% CI 1.23, 5.42 for colon cancer recurrence) (19). No such associations were observed in men (19). In the other study, only severe obesity (BMI 35) compared with normal weight was associated with a higher risk of recurrence (HR: 1.38; 95% CI: 1.10, 1.73), colon cancer-specific mortality (HR: 1.36; 95% CI: 1.06, 1.73), and all-cause mortality (HR: 1.28; 95% CI: 1.04, 1.57), and similar results were obtained for men and women (21). In patients with rectal cancer, obese men (BMI 30) were at higher risk of local rectal cancer recurrence compared with normal weight men (HR: 1.61; 95% CI: 1.00, 2.59) (20). The risk of all-cause mortality was not increased, and no significant associations were observed in women. Local rectal cancer recurrence was also higher in overweight men (BMI of : HR: 1.66; 95% CI: 1.07, 2.56). Further, no significant associations with overweight were shown in these studies. BMI after cancer Only one study investigated BMI during chemotherapy and weight change from the time of chemotherapy through 6 mo after completion of chemotherapy (32). This study was conducted in 1053 patients with colon cancer. Obese (BMI of ) or severely obese (BMI 35) subjects compared with normal weight subjects at the time of chemotherapy were not at higher risk of recurrence and/or all-cause mortality. Also, weight gain from the time of chemotherapy through 6 mo after chemotherapy was not associated with recurrence and/or death. Physical activity The association between physical activity and recurrence or survival of colorectal cancer was investigated in 6 studies (12, 18, 33 36) (Table 2). In 4 of these studies, previously validated physical activity questionnaires were used, and leisure-time physical activity was assessed in metabolic equivalent task (MET) hours per week (33 36). One study recorded physical activity as low (totally sedentary lifestyle), moderate, or high (recreational or occupational activity that required strenuous physical activity), but no further details were given (18). In the other study, 3 questions were asked about the frequency per week of vigorous exercise, less vigorous exercise, and walking (12). There were no results of randomized clinical trials on physical activity and recurrence or survival of colorectal cancer. Physical activity before cancer In 526 colorectal cancer patients who were identified within a cohort study, nonoccupational physical activity was assessed before (12). Exercisers were shown to be at lower risk of colorectal cancer specific mortality than were nonexercisers (HR: 0.73; 95% CI: 0.54, 1.00). In another study in 573 female colorectal cancer patients, colorectal cancer specific and allcause mortality did not differ for women in the highest quartile compared with the lowest quartile of leisure-time physical activity that was assessed a median of 6 mo before (33).

10 480 VRIELING AND KAMPMAN TABLE 2 Epidemiologic studies of physical activity and colorectal cancer recurrence or survival 1 characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment Physical activity before Haydon, 2006, Australia (12) Meyerhardt, 2006, United States (33) Meyerhardt, 2009, United States (36) Physical activity at time of Dray, 2003, France (18) 526 CRC patients (M, F; stages I IV; aged y) diagnosed within the MCCS cohort 208 and 181 CRC 573 CRC patients (F; stages I III) NHS 132 and 80 CRC 661 CRC patients (M; stages I III) HPFS 258 and 88 CRC 148 CRC patients (M, F; Dukes stage A D; aged y) identified in the Cote d Or area and part of a case-control study 46 and 70 5 and 10 y, respectively Diagnosis: Self-reported Median: 5.5 y nonoccupational physical activity at baseline (median: 5.3 y before ) Inclusion: Self-reported leisuretime physical activity (median: 6 mo before ) CRC-specific mortality Exercisers vs nonexercisers CRC-specific mortality Total MET-h activity pre, 18 vs,3/wk Median: 9.6 y 0.95 (0.57, 1.59) Diagnosis: Self-reported leisuretime CRC-specific mortality Total MET-h activity P for trend ¼ 0.65 Median: 8.6 y physical activity pre P for trend ¼ 0.26 (6 mo before ) Diagnosis Physical activity at time 5- or 10-y survival of interview, 3 mo of 0.73 (0.54, 1.00) Age, sex, and tumor stage 0.86 (0.44, 1.67) Age at, year of, BMI, stage of disease, tumor-differentiation grade, location of primary tumor, chemotherapy, time from to physical activity measurement, change in BMI before and after, and smoking status Age at, year of, BMI, stage of disease, tumordifferentiation grade, location of primary tumor, time from to physical activity measurement, change in BMI before and after, and smoking status Physical activity (data not shown) Age, sex, stage, and tumor location (Continued)

11 LIFESTYLE FACTORS IN COLORECTAL CANCER SURVIVAL 481 TABLE 2 (Continued) Physical activity after Meyerhardt, 2006, United States (33) Meyerhardt, 2006, United States (34) characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment 573 CRC patients (F; stages I III) NHS 132 and 80 CRC 832 CC patients (M, F; stage III) treated within a postoperative adjuvant chemotherapy trial (CALGB 89803) 84 and 159 CC recurrences Inclusion: Self-reported leisuretime physical activity Median: 9.6 y (median: 6 mo before and 1 4 y post) and change in physical activity Inclusion: Median: 3.8 y Self-reported physical activity 6 mo after completion of chemotherapy CRC-specific mortality Total MET-h activity post, 18 vs,3/wk Recurrence Recurrence or death Increase in physical activity vs no change Decrease in physical activity vs no change Total MET-h activity post, 27 vs,3/wk 0.39 (0.18, 0.82) Age at, year of, BMI, stage of disease, 0.48 (0.24, 0.97) tumor-differentiation grade, location of 1.32 (0.74, 2.34) primary tumor, chemotherapy, time Similar results from to physical activity measurement, change in BMI before and after, and smoking status 0.37 (0.16, 0.82) 0.60 (0.36, 1.01) 0.55 (0.33, 0.91) Age, sex, depth of invasion through bowel wall, no. of positive lymph nodes, clinical perforation at time of surgery, bowel obstruction at time of surgery, baseline CEA, tumordifferentiation grade, baseline performance status, treatment arm, weight change between first and second questionnaire, BMI at time of second questionnaire, time between study entry and completion of second questionnaire (Continued)

12 482 VRIELING AND KAMPMAN TABLE 2 (Continued) characteristics) Follow-up time Variable of interest Outcome Contrast RR (95% CI) Adjustment Meyerhardt, 2009, United States (35) Meyerhardt, 2009, United States (36) 484 CC patients (M, F; stages I III) NHS and HPFS 152 and 50 CC 661 CRC patients (M; stages I III) HPFS 258 and 88 CRC Diagnosis: 1976/ FU: 1986 to June 2006 Diagnosis: Median: 8.6 y Self-reported leisuretime physical activity 1 4 y (median: 17 mo) post Self-reported leisuretime physical activity 6 mo to 4 y (median: 15 mo) post CC-specific mortality Total MET-h activity post, 18 vs,18/wk Total MET-h activity post, 18 vs,18/wk CRC-specific mortality Total MET-h activity post, 27 vs,3/wk 0.64 (0.33, 1.23) Age, sex, stage, year of 0.32 (0.12, 0.85) for, histology p271 grade, BMI, and time 1.40 (0.41, 4.72) for p272 Pint ¼ (0.41, 0.86) Pint ¼ 0.37 for p271 vs p (0.24, 0.92) 0.59 (0.41, 0.86) of physical activity assessment No interaction by FASN, K-ras, p53, p21, PIK3CA Age at, year of, BMI, stage of disease, tumor-differentiation grade, location of primary tumor, time from to physical activity measurement, change in BMI before and after, and smoking status 1 RR, relative risk; CRC, colorectal cancer; MCCS, Melbourne Collaborative Cohort Study; NHS, Nurses Health Study; MET-h, metabolic equivalent task-hours; CC, colon cancer; CALGB, Cancer and Leukemia Group B; CEA, carcinoembryonic antigen; HPFS, Health Professionals Follow-Up Study; FU, follow-up; 1, positive; 2, negative; FASN, fatty acid synthase; K-ras, Kirsten rat sarcoma viral oncogene homolog; PIK3CA, phosphoinositide-3-kinase, catalytic, alpha polypeptide; P int, P for interaction; MET-h, metabolic equivalent task hours.

13 LIFESTYLE FACTORS IN COLORECTAL CANCER SURVIVAL 483 Similarly, in 661 male colorectal cancer patients, no significant trend in colorectal cancer specific and all-cause mortality was detected with increasing exercise (36). Physical activity at time of cancer In a study in 148 colorectal cancer patients, physical activity assessed at 3 mo of was not related to the 5- or 10-y survival of colorectal cancer (18). Physical activity after cancer In the study in 573 female colorectal cancer patients (33), cases in the highest quartile for leisure-time physical activity that was assessed 1 4 y after were at lower risk of colorectal cancer specific mortality than were cases in the lowest quartile (HR: 0.39; 95% CI: 0.18, 0.82). Similar results were obtained in the 661 male colorectal cancer survivors (HR: 0.47; 95% CI: 0.24, 0.92) (36). Risk estimates for all-cause mortality were comparable with those for colorectal cancer specific mortality for both studies. Furthermore, female colorectal cancer patients who increased their physical-activity level after were at lower risk of colorectal cancer specific mortality (HR: 0.48; 95% CI: 0.24, 0.97) and all-cause mortality than were patients that did not change their physicalactivity level (33). A randomized adjuvant chemotherapy trial in 832 colon cancer patients showed a lower risk of colon cancer recurrence or death for patients in the highest quintile of physical activity 6 mo after chemotherapy than for patients in the lowest quintile (HR: 0.55; 95% CI: 0.33, 0.91) (34). Another study in 484 colon cancer patients showed a nonsignificantly lower risk of colon cancer-specific mortality (HR: 0.64: 95% CI: 0.33, 1.23) and a lower risk of all-cause mortality (HR: 0.60: 95% CI: 0.41, 0.86) for cases with a leisure-time physical activity of 18 MET-h/wk than cases who engaged in,18 MET-h/wk of exercise after (35). This study also investigated whether this protective association was dependent on molecular characteristics of the tumor. The post leisure-time physical activity of 18 MET-h/wk compared with,18 MET-h/wk decreased the risk of colon cancer-specific mortality for tumors with p27 nuclear expression (HR: 0.32; 95% CI: 0.12, 0.85) but not for tumors with a loss of p27. Physical activity did not differentially affect the all-cause mortality of p27-positive and p27-negative tumors. No modifying effect of fatty acid synthase, Kirsten rat sarcoma viral oncogene homolog; p53; p21; and phospoinositide-3- kinase, catalytic, alpha polypeptide on the association between exercise and colon cancer-specific or all-cause mortality was observed. Dietary factors The association between dietary factors and the recurrence or survival of colorectal cancer has been investigated in 12 observational studies described in 14 articles (9, 11, 18, 22, 25, 26, 37 44) (Table 3). Seven of these studies examined dietary factors that referred to the time before cancer (9, 11, 18, 22, 26, 37 40), and 5 studies investigated post dietary factors (25, 41 44). Eight studies examined habitual diet assessed by a questionnaire (9, 11, 18, 22, 25, 26, 37, 38, 41). These studies used food-frequency questionnaires (9, 22, 25, 37, 38, 41) or a qualitative and quantitative dietary questionnaire (18) or obtained information about alcohol intake from a baseline questionnaire (11) or hospital-document information (26). Four studies evaluated blood concentrations of vitamins (39, 40, 42 44). As also reported by the second WCRF/AICR expert report (5), only 5 small intervention studies conducted in 100 randomly assigned colorectal cancer patients investigated the effects of diet and supplements with respect to all-cause mortality, and no significant results were shown. A more recent flavonoid intervention in 29 colorectal cancer patients did not show any effect on cancer recurrence either (45). The null results can be largely explained by small numbers and low trial quality, and are not further discussed in this review. Dietary factors before cancer Six studies investigated pre diet and were initially set up as case-control (9, 18), case-case (22, 37), or cohort (11, 38) studies to examine the association between dietary factors and colorectal cancer incidence or were based on cancer-registry data (26). Studies included 148 to 1882 colorectal cancer cases. Diet was assessed after a cancer by a questionnaire that referred to habitual diet before the (9, 18, 22, 26, 37) or interview (11, 38). In all but one (11) studies, associations were adjusted for the tumor stage at the, and associations in one study (22) were additionally adjusted for the cancer treatment. In 2 of the 3 studies, high compared with low energy intake before was associated with a lower risk of allcause death at 2 (9) and 5 y (18) of survival, but no association was shown after longer follow-up. In the other study, pre energy intake was not associated with overall survival after 7 9 y of follow-up (22). Higher dietary fiber intake was associated with a higher risk of all-cause mortality in one study (9) but not in the other 2 studies (18, 22). In the case-case study, higher meat consumption was associated with a higher risk of all-cause mortality in familial but not sporadic colorectal cancer cases (37). Further, regular compared with infrequent wine consumption was associated with a lower risk of all-cause mortality in cases with a family history of colorectal cancer only (22). No association was shown for total alcohol consumption with all-cause mortality in 2 studies in colorectal (11) and colon (26) cancer cases, respectively. Whole-year daily use of cod liver oil or daily use of other dietary supplements compared with nonuse was also not associated with all-cause mortality in colorectal cancer cases (38). In a cohort study that included 304 colorectal cancer cases, plasma concentrations of 25-hydroxyvitamin D 3 [25(OH)D] (39) or folate (40) were evaluated. Plasma concentrations were assessed a median of 6 y before the colorectal cancer and were related to colorectal cancer specific and all-cause mortality. Both analyses were adjusted for the stage of disease, and in the analysis on folate, a limited adjustment for the use of chemotherapy was possible. Colorectal cancer specific mortality was nonsignificantly lower for subjects in the highest quartile than in the lowest quartile of plasma 25(OH)D concentrations (HR: 0.61; 95% CI: 0.31, 1.19) and was significantly lower for the highest quintile than for the lowest quintile of plasma folate concentrations (HR: 0.42; 95% CI: 0.20, 0.88). Results were similar for all-cause mortality.

14 484 VRIELING AND KAMPMAN TABLE 3 Epidemiologic studies of diet, blood concentrations of vitamins, and colorectal cancer recurrence or survival 1 characteristics) Follow-up time Dietary assessment method Outcome Contrast RR (95% CI) Adjustments Diet before Slattery, 1989, United States (9) Dray, 2003, France (18) Park, 2006, Korea (11) Zell, 2007, United States (37) 410 CC cases (M, F; aged y) identified through the Utah Cancer Registry and part of 2 case-control studies CRC patients (M, F; Dukes stage A D; aged y) identified in the Cote d Or area and part of a case-control study 46 and 70 5 and 10 y, respectively 1882 CRC patients (M) NHICS 555 and 524 CRC 144 familial (at least one first-degree relative) and 367 sporadic CRC patients (M, F) part of the UCI 203 Diagnosis: FFQ,6 mo after Median: 83 mo (range: mo) Referring to the 5 y (study 1) or 2 y (study 2) before Diagnosis: Qualitative and 5- or 10-y survival quantitative dietary questionnaire,3mo after Referring to the year before Diagnosis: Self-administered FU to end of 2004 baseline (median: 3.8 y) questionnaire including alcohol use Diagnosis: FFQ less than a median 7 9 y of 2 y after Q4 vs Q1 (cutoffs: NA) after 5 y between 5 and 10 y Energy Crude fiber Fat (total, percentage kcal) 0.60 (0.37, 0.98) (particularly for those dying,2 y of ) Fat types Protein (total, percentage kcal) Cholesterol Calcium Vitamin A Vitamin C Energy, high vs moderate and low Carbohydrate Fat Protein Foods, micronutrients Alcohol, vs 0 g/wk Meat: (Q4) vs (Q1 Q3) servings/wk Referring to the year before Meat: (Q4) vs (Q1 Q3) servings/wk Age, sex, stage, year of, religion, and crude fiber 0.18 (0.07, 0.44) Age, sex, stage, tumor location, and Similar Similar Similar interaction terms 0.92 (0.72, 1.19) Age, BMI, fasting serum glucose concentration, cholesterol concentration, physical activity, food preference, blood pressure, and other comorbidities 2.24 (1.25, 4.03) for familial CRC 1.01 (0.67, 1.51) for sporadic CRC Age, sex, and stage (Continued)

15 LIFESTYLE FACTORS IN COLORECTAL CANCER SURVIVAL 485 TABLE 3 (Continued) Zell, 2007, United States (22) Asghari-Jafarabadi, 2009, Iran (26) Skeie, 2009, Norway (38) Blood concentrations before Ng, 2008, United States (39) characteristics) Follow-up time Dietary assessment method Outcome Contrast RR (95% CI) Adjustments 141 familial (at least one first-degree relative) and 358 sporadic CRC patients (M, F) part of the UCI 199 and 95 CRC (of 106 recorded ) 1219 CRC patients (M, F; stages I IV; average age: 54 y) at registered in a cancer registry Deaths: NA 399 CRC patients (F; 65% regional metastases; average age: 61 y) diagnosed within the Norwegian Women and Cancer cohort study 143 Diagnosis: FFQ less than a median 7 9 y of 2 y after Included 2002 to October 2007 Mean: 9.3 y (CC) and 11.3 y (RC) Diagnosis 1996/ Referring to the year before Hospital document information FFQ at baseline ( ) FU up to end 2007 Referring to the year Wine, 1 3 glasses/mo vs,1 glass/mo Beer Liquor Energy Fiber Calcium Fruit and vegetables Alcohol past or current vs never used Whole year daily users vs nonusers of cod liver oil before interview Daily users vs nonusers of other dietary supplements 0.50 (0.25, 0.99) for familial CRC; 0.89 (0.59, 1.33) for sporadic CRC 239 CC: 1.84 (0.89, 3.81) Age, sex, stage, SES, BMI, beer, liquor, treatment with surgery, radiation, and chemotherapy Sex, BMI, inflammatory bowel disease and tumor stage, grade, and size 0.82 (0.48, 1.41) Age at, smoking, stage 1.19 (0.84, 1.69) 304 CRC patients (M, F; stages I IV) NHS and HPFS cohort studies 123 and 96 CRC Diagnosis FU up to 2005 Median: 78 mo (range: mo) Plasma 25(OH)D concentrations assessed a median of 6.0 y (range: y) before CRC-specific mortality Plasma 25(OH)D, Q4 vs Q (0.31, 1.19) 0.52 (0.29, 0.94) Age at, season of blood draw, sex, cancer stage, grade of tumor differentiation, location of primary tumor, year of, BMI at, and post physical activity (Continued)

16 486 VRIELING AND KAMPMAN TABLE 3 (Continued) Wolpin, 2008, United States (40) Diet after Meyerhardt, 2007, United States (41) Tian, 2008, China (25) characteristics) Follow-up time Dietary assessment method Outcome Contrast RR (95% CI) Adjustments 301 CRC patients (M, F; stages I IV) NHS and HPFS cohort studies 122 and 95 CRC 1009 patients (M, F; stage III) treated within a postoperative adjuvant chemotherapy trial (CALGB 89803) 251, 223 CC, and 324 CC recurrences 39 CRC patients (M, F; stages I IV) who underwent surgery in 4 provincial-level hospitals in Fuzhou, China 14 Diagnosis: Plasma folate concentrations CRC-specific mortality Plasma folate, Q5 vs Q (0.20, 0.88) Age at, stage of disease, histologic differentiation, receipt of chemotherapy, tumor location, time period of, BMI, physical activity, smoking status, aspirin use, alcohol consumption, total vitamin D intake, and PMH use (including male sex) FU to (0.24, 0.88) Median: 78 mo (range: assessed a median of mo) 6.0 y (range y) before Inclusion: FFQ midway through Median: 5.3 y adjuvant therapy and 6 mo after completion of adjuvant therapy (4 and 14 mo after surgical resection, respectively) Diagnosis: January to June 2006 Referring to the past 3mo FFQ during 20 d after surgery 1-y survival Referring to the last week Western dietary pattern 2.32 (1.36, 3.96) Age, sex, depth of Recurrence (high consumption of 2.85 (1.75, 4.63) Recurrence or death meat, fat, refined 3.25 (2.04, 5.19) grains, and desserts), Q4 vs Q1 Prudent dietary pattern (high consumption of fruit and vegetables, poultry, and fish) invasion through bowel wall, no. of positive lymph nodes, clinical perforation at time of surgery, bowel obstruction at time of surgery, baseline performance status, treatment group, weight change between first and second questionnaire, and time-varying BMI, physical activity, and total calories Low vs high energy 1.97 (0.21, 18.1) Age, sex, and tumor Low vs high protein 1.56 (0.19, 12.8) stage (Continued)