Volume 119, Issue 8 p. 1528-1536
Original Article
Free Access

Body mass index at diagnosis and survival among colon cancer patients enrolled in clinical trials of adjuvant chemotherapy

Frank A. Sinicrope MD

Corresponding Author

Frank A. Sinicrope MD

Division of Oncology, Mayo Clinic and North Central Cancer Treatment Group, Rochester, Minnesota

Fax: (507) 255-6318

Mayo Clinic, 200 First Street SW, Rochester, MN 55905Search for more papers by this author
Nathan R. Foster MS

Nathan R. Foster MS

Division of Biomedical Statistics and Informatics, Mayo Clinic and North Central Cancer Treatment Group, Rochester, Minnesota

Search for more papers by this author
Greg Yothers PhD

Greg Yothers PhD

National Surgery Adjuvant Breast and Bowel Project, Pittsburgh, Pennsylvania

Search for more papers by this author
Al Benson MD

Al Benson MD

Eastern Cooperative Oncology Group, Philadelphia, Pennsylvania

Search for more papers by this author
Jean Francois Seitz MD

Jean Francois Seitz MD

French Society for Digestive Oncology, University of the Mediterranean, Marseilles, France

Search for more papers by this author
Roberto Labianca MD

Roberto Labianca MD

Unit of Medical Oncology, Riuniti Hospital, Bergamo, Italy

Search for more papers by this author
Richard M. Goldberg MD

Richard M. Goldberg MD

Cancer and Leukemia Group B, Chicago, Illinois

Search for more papers by this author
Aimery DeGramont MD

Aimery DeGramont MD

Multidisciplinary Cooperative Oncology Group, Paris, France

Search for more papers by this author
Michael J. O'Connell MD

Michael J. O'Connell MD

National Surgery Adjuvant Breast and Bowel Project, Pittsburgh, Pennsylvania

Search for more papers by this author
Daniel J. Sargent PhD

Daniel J. Sargent PhD

Division of Biomedical Statistics and Informatics, Mayo Clinic and North Central Cancer Treatment Group, Rochester, Minnesota

Search for more papers by this author
for the Adjuvant Colon Cancer Endpoints (ACCENT) Group

for the Adjuvant Colon Cancer Endpoints (ACCENT) Group

Division of Oncology, Mayo Clinic and North Central Cancer Treatment Group, Rochester, Minnesota

Search for more papers by this author
First published: 10 January 2013
Citations: 125

Abstract

BACKGROUND:

Although obesity is an established risk factor for developing colon cancer, its prognostic impact and relation to patient sex in colon cancer survivors remains unclear.

METHODS:

The authors examined the prognostic and predictive impact of the body mass index (BMI) in patients with stage II and III colon carcinoma (N = 25,291) within the Adjuvant Colon Cancer Endpoints (ACCENT) database. BMI was measured at enrollment in randomized trials of 5-fluorouracil–based adjuvant chemotherapy. Association of BMI with the time to recurrence (TTR), disease-free survival (DFS), and overall survival (OS) were determined using Cox regression models. Statistical tests were 2-sided.

RESULTS:

During a median follow-up of 7.8 years, obese and underweight patients had significantly poorer survival compared with overweight and normal-weight patients. In a multivariable analysis, the adverse prognostic impact of BMI was observed among men but not among women (Pinteraction = .0129). Men with class 2 and 3 obesity (BMI ≥35.0 kg/m2) had a statistically significant reduction in DFS (hazard ratio [HR], 1.16; 95% confidence interval [CI], 1.01-1.33; P = .0297) compared with normal-weight patients. Underweight patients had a significantly shorter TTR and reduced DFS (HR, 1.18; 95% CI, 1.09-1.28; P < .0001) that was more significant among men (HR, 1.31; 95% CI, 1.15-1.50; P < .0001) than among women (HR, 1.11; 95% CI, 1.01-1.23; P = .0362; Pinteraction = .0340). BMI was not predictive of a benefit from adjuvant treatment.

CONCLUSIONS:

Obesity and underweight status were associated independently with inferior outcomes in patients with colon cancer who received treatment in adjuvant chemotherapy trials. Cancer 2013. © 2013 American Cancer Society.

INTRODUCTION

The increasing prevalence of obesity and its association with both cancer risk and prognosis have major importance in public health. Approximately 34% of adults in the United States are obese, which is defined as having a body mass index (BMI)1, 2 ≥30 kg/m2. Rates of obesity have increased 2-fold among adults and 3-fold among children in the past 30 years in the United States.3 Although obesity is an established risk factor for colorectal cancer (CRC) incidence and death,4-8 its association with mortality in survivors of CRC is less clear, and inconsistent results have been reported.9-12 Within clinical trials of adjuvant chemotherapy in patients with colon cancer, the obese subgroup has represented 17% to 35% of the study cohorts, which has limited the statistical power for comparison with survival outcomes. Evidence indicates that the time frame during which BMI is estimated or determined (ie, prediagnosis vs postdiagnosis) may also influence its prognostic impact.13 To date, there are inconsistent data as to whether the association of BMI with CRC survival may differ by patient sex.9, 11

In the current study, we determined the association of BMI with colon cancer prognosis and examined its potential predictive impact for the outcome of 5-fluorouracil (5-FU)–based adjuvant chemotherapy. We tested the hypothesis that obese patients have an increased risk of colon cancer recurrence and death compared with normal-weight patients that may be more evident in men compared with women. We used the Adjuvant Colon Cancer Endpoints (ACCENT) database, a pooled resource of >20,000 participants in national and international colon cancer adjuvant chemotherapy trials.14 This database provides a unique opportunity to definitively address the impact of BMI on clinical outcomes in patients with colon cancer. Clarification of the role of BMI in prognosis can influence patient education and management, because body weight represents a modifiable factor that may influence patient outcomes.

MATERIALS AND METHODS

Study Population

We used the ACCENT Group14 database, which contains data on TNM stage II and III colon cancers (N = 25,291) from patients who participated in 21 randomized trials of 5-FU–based adjuvant chemotherapy conducted in North America and Europe (Table 1). We included all trials that had the following data: BMI, age, tumor stage, sex, and censoring variables for clinical outcome. The database does not include data on toxicity or comorbid conditions. All studies were approved by institutional review boards at the respective study sites, and all participants provided written informed consent. This study was conducted under an institutional review board-approved protocol.

Table 1. Study Population of Patients With Stage II and III Colon Cancers (N = 25,291)
Variable No. of Patients (%)
Clinical trial
 INT-0035 780 (3.1)
 NCIC 359 (1.4)
 NSABP C-01 664 (2.6)
 NSABP C-02 678 (2.7)
 NSABP C-03 1042 (4.1)
 NSABP C-04 2083 (8.2)
 NSABP C-05 2136 (8.4)
 NSABP C-06 1556 (6.2)
 NSABP C-07 2434 (9.6)
 CALGB 89803 1236 (4.9)
 SWOG 9415 912 (3.6)
 NCCTG 78-48-52 240 (0.9)
 NCCTG 87-46-51 106 (0.4)
 NCCTG 89-46-51 907 (3.6)
 NCCTG 91-46-53 871 (3.4)
 PETACC-3 3174 (12.5)
 SIENA 225 (0.9)
 GIVIO 820 (3.2)
 GERCOR 869 (3.4)
 MOSAIC 2237 (8.8)
 X-ACT 1962 (7.8)
Study treatment
 Surgery vs 5-FU 3872 (15.3)
 5-FU vs 5-FU variations 8820 (34.9)
 5-FU vs oxaliplatin 4671 (18.5)
 5-FU vs irinotecan 4410 (17.4)
 5-FU vs oral 5-FU 3518 (13.9)
BMI category
 Underweight <20 kg/m2 1853 (7.3)
 Normal, 20-24.9 kg/m2 9887 (39.1)
 Overweight, 25-29.9 kg/m2 9088 (35.9)
 Obese, ≥30 kg/m2 4463 (17.6)
  Class 1, 30-34.9 kg/m2 3203 (12.7)
  Class 2-3, ≥35 kg/m2 1260 (5)
  • Abbreviations: 5-FU, 5-fluororacil; BMI, body mass index; CALGB, Cancer and Leukemia Group B; GERCOR, French Multidisciplinary Clinical Research Group in Oncology; GIVIO, Interdisciplinary Group for Cancer Care Evaluation (Italy); INT, intergroup; MOSAIC, Multicenter International Study of Oxlaliplatin/5-Fluorouracil-Leucovorin in the Adjuvant Treatment of Colon Cancer; NCCTG, North Central Cooperative Treatment Group; NCIC, National Cancer Institute of Canada; NSABP, National Surgical Adjuvant Breast and Bowel Project; PETACC, Pan-European Trial in Adjuvant Colon Cancer; SIENA/University of Siena; SWOG, Southwest Oncology Group; X-ACT, Xeloda in Adjuvant Colon Cancer Therapy.

Measurement and Categorization of Body Mass Index

Body weight and height were measured and recorded at study enrollment by trained staff and were used to calculate the BMI (in kg/m2). BMI categories were created on the basis of World Health Organization classifications and previous reports1 as follows: underweight, BMI <20 kg/m2; normal weight, 20 to 24.9 kg/m2; overweight, 25 to 29.9 kg/m2; and obese, ≥30 kg/m2 (class 1 obese, 30-34.9 kg/m2; class 2 and 3 obese, ≥35.0 kg/m2).

Statistical Analyses

The association of the categorical BMI with clinicopathologic variables was analyzed using the Kruskal-Wallis test (continuous) or the chi-square test (≥3 categories). The Cochran-Armitage test for trend was used across the ordered BMI categories (2-level variables). The association of BMI with clinicopathologic variables was determined using chi-square tests and Wilcoxon rank-sum tests. The time to recurrence (TTR) was calculated as the number of years from random assignment to colon cancer recurrence. Disease-free survival (DFS) was measured from the date of randomization to the first date of local, regional, or distant relapse or death. Overall survival (OS) was calculated with date of death as the outcome. Outcome variables were censored at 8 years, and their distributions were estimated using Kaplan-Meier methodology. Univariate and multivariate Cox proportional hazard models15 were used to explore associations of BMI with outcome variables. Score (univariate) and likelihood-ratio (multivariate) test P values were used to test the significance of each covariate after stratifying by treatment group. Interaction effects were tested in Cox models with the use of the likelihood-ratio test. The prognostic impact of BMI also was modeled using restricted cubic splines.16 All statistical tests were 2-sided. Analyses were performed using SAS software (SAS Institute, Cary, NC).

RESULTS

The study included 25,291 patients with curatively resected, TNM stages II and III colon cancer who participated in 21 randomized trials of 5-FU–based adjuvant chemotherapy (Table 1). BMI was measured at study entry and was categorized as indicated in Table 1: the median BMI was 25.4 kg/m2 (range, 10.0-70.3 kg/m2). After a median follow-up of 7.8 years in living patients, 32% had cancer recurrence, and 32% (N = 7973) had died.

Across BMI categories, we observed statistically significant but clinically modest associations between BMI and tumor stage, the number of metastatic lymph nodes, age, sex, Eastern Cooperative Oncology Group performance status, and T-classification (all P < .0001) (Table 2). Tumors from obese patients versus normal-weight patients were more likely to be, stage III versus II, distal, and T1/T2 versus T3/T4 tumors (all P < .01) (Table 2). Compared with normal-weight patients, obese patients were more likely to have >3 metastatic regional lymph nodes (N2 disease; P = .0001) (Table 2). Similar numbers of surgically removed lymph nodes were examined from obese patients compared with patients in other BMI categories. Compared with normal-weight patients, underweight patients were significantly more likely to be younger (median, 58 years vs 61 years; P < .0001), to be women (69% vs 48%; P < .0001), and to have a performance status of 1 or 2 versus 0 (24% vs 19%; P < .0001) (Table 2).

Table 2. Clinical Characteristics by Body Mass Index Group
BMI Category: No. of Patients (%) P
Variable Underweight, N = 1853 Normal, N = 9887 Overweight, N = 9088 Obese, N = 4463 Total, N = 25,291 Overalla Obese vs Normalb Underweight vs Normalb
TNM Stage < .0001 .0013 .0783
 II 648 (35 3250 (32.9) 2801 (30.8) 1346 (30.2) 8045 (31.8)
 III 1205 (65) 6637 (67.1) 6287 (69.2) 3117 (69.8) 17,46 (68.2)
Histologic gradecd .0948 .4203 .7622
 1-2 921 (80.3) 5149 (80.7) 4821 (82.3) 2198 (81.4) 13,089 (81.4)
 3-4 226 (19.7) 1233 (19.3) 1036 (17.7) 502 (18.6) 2997 (18.6)
Tumor sited .0211 .0068 .6695
 Distal 913 (55.3) 4804 (54.7) 4497 (55.5) 2352 (57.2) 12,566 (55.5)
 Proximal 739 (44.7) 3979 (45.3) 3609 (44.5) 1757 (42.8) 10,084 (44.5)
Sex < .0001 0.0190 < .0001
 Women 1284 (69.3) 4709 (47.6) 3342 (36.8) 2220 (49.7) 11,555 (45.7)
 Men 569 (30.7) 5178 (52.4) 5746 (63.2) 2243 (50.3) 13,736 (54.3)
Performance statusd <.0001 .0138 < .0001
 0 1289 (76) 7580 (81.4) 7067 (82.2) 3352 (79.4) 19,288 (81)
 1 381 (22.5) 1657 (17.8) 1470 (17.1) 840 (19.9) 4348 (18.3)
 2 25 (1.5) 70 (0.8) 58 (0.7) 29 (0.7) 182 (0.8)
T-classificationd < .0001 .0001 .0368
 T1-T2 190 (11.4) 1207 (13.3) 1258 (14.6) 687 (15.9) 3342 (14.1)
 T3-T4 1478 (88.6) 7896 (86.7) 7345 (85.4) 3647 (84.1) 20,366 (85.9)
Age: Median, y 58.0 61.0 62.0 60.0 61.0 < .0001 .0007 < .0001
No. of positive lymph nodesd < .0001 <.0001 .2708
 0 641 (37.9) 3232 (35.9) 2785 (33.5) 1342 (32.2) 8000 (34.5)
 1-3 693 (41) 3831 (42.5) 3679 (44.3) 1819 (43.7) 10,022 (43.2)
 >3 357 (21.1) 1950 (21.6) 1840 (22.2) 1006 (24.1) 5153 (22.2)
  • Abbreviations: BMI, body mass index.
  • a P values were determined with the Kruskal-Wallis test (continuous data), the chi-square test (variables with ≥3 categories), or the Cochran-Armitage test for trend (variables with 2 categories).
  • b P values were determined with the chi-square test (categorical data) or the Wilcoxon rank-sum test (continuous data).
  • c Grade1-2 was defined as well/moderate differentiation, and grade 3-4 was defined as poor differentiation/undifferentiated.
  • d Note that some patients were missing data for this variable.

In univariate analysis, the BMI category was associated significantly with TTR, DFS, and OS across all adjuvant studies (Table 3). Overweight patients were not at increased risk of recurrence or mortality compared with normal-weight patients. Obese patients had shorter TTR and worse DFS and OS (OS: hazard ratio [HR], 1.11; 95% confidence interval [CI], 1.04-1.18; P = .0014) versus normal-weight patients (Table 3). Underweight patients also had shorter TTR and worse DFS rates (HR, 1.11; 95% CI, 1.03-1.20; P = .0093) and OS rates (Table 3). Patients who had stage III tumors versus stage II tumors, poor differentiation, and increased numbers of metastatic lymph nodes all had shorter TTR, DFS, and OS (all P < .0001) (Table 3). Higher T-classification, but not primary tumor site, and worse Eastern Cooperative Oncology Group performance status were associated similarly with worse outcome (data not shown). Analysis by patient sex revealed that the BMI category was significantly prognostic in men for TTR (P = .0015), DFS (P < .0001), and OS (P < .0001), but not in women (all P > .20) (Table 3, Fig. 1A-D).

Details are in the caption following the image

Kaplan-Meier plots illustrate the association of body mass index category with the time- to- recurrence and overall survival for (A,C) men and (B,D) women with resected, stage II and III colon cancer who participated in adjuvant chemotherapy trials.

Table 3. Univariate Survival for Clinicopathologic Variables (N = 25,291)
Time-to Recurrence Disease-Free Survival Overall Survival
Variable HR (95% CI) Pa HR (95% CI) Pa HR (95% CI) Pa
BMI category
 Overall .0208 .0069 .0030
  Normal Reference Reference Reference
  Underweight 1.11 (1.02-1.21) .0169b 1.11 (1.03-1.20) .0093b 1.12 (1.02-1.22) .0132b
  Overweight 1.01 (0.96-1.06) .7992b 1.00 (0.96-1.05) .9373b 1.03 (0.97-1.08) .3283b
 Obese 1.07 (1.01-1.14) .0290b 1.07 (1.01-1.13) .0174b 1.11 (1.04-1.18) .0014b
   Class 1 1.06 (0.99-1.13) .1140b 1.06 (1.00-1.13) .0681b 1.11 (1.04-1.19) .0024b
   Class 2-3 1.11 (1.00-1.22) .0500b 1.10 (1.00-1.20) .0503b 1.09 (0.98-1.21) .1018
Men
 BMI category, overall .0015b < .0001 < .0001
  Normal Reference Reference Reference
  Underweight 1.17 (1.01-1.35) .0401b 1.23 (1.08-1.40) .0019b 1.29 (1.12-1.48) .0004b
  Overweight 0.99 (0.93-1.06) .8171b 0.97 (0.91-1.03) .2641b 0.98 (0.92-1.05) .5213b
  Obese 1.14 (1.05-1.24) .0023b 1.11 (1.03-1.20) .0073b 1.14 (1.04-1.24) .0033b
   Class 1 1.12 (1.02-1.23) .0167b 1.09 (1.00-1.18) .0558b 1.12 (1.02-1.23) .0170b
   Class 2-3 1.21 (1.04-1.40) .0127b 1.20 (1.04-1.37) .0096b 1.19 (1.02-1.38) .0242b
Women
 BMI category, overall .4660 .3969 .2064
  Normal Reference Reference Reference
  Underweight 1.09 (0.98-1.21) .1276b 1.09 (0.98-1.20) .1005b 1.08 (0.96-1.20) .2028b
  Overweight 1.03 (0.95-1.11) .5088b 1.04 (0.96-1.12) .3390b 1.07 (0.99-1.16) .0862b
  Obese 1.00 (0.92-1.10) .9441b 1.03 (0.95-1.12) .4342b 1.08 (0.99-1.19) .0955b
   Class 1 0.98 (0.89-1.09) .7759b 1.03 (0.93-1.13) .6004b 1.10 (0.99-1.22) .0767b
   Class 2-3 1.04 (0.91-1.19) .5738b 1.05 (0.92-1.19) .4578b 1.05 (0.91-1.20) .5327b
Histologic gradec < .0001 < .0001 < .0001
 1-2 Reference Reference Reference
 3-4 1.40 (1.31-1.49) 1.37 (1.29-1.46) 1.54 (1.44-1.65)
Stage < .0001 < .0001 < .0001
 II Reference Reference Reference
 III 2.76 (2.60-2.92) 2.34 (2.22-2.46) 2.43 (2.29-2.57)
Sex .3966 < .0001 < .0001
 Women Reference Reference Reference
 Men 1.02 (0.98-1.06) 1.09 (1.04-1.13) 1.10 (1.06-1.15)
No. of metastatic lymph nodes < .0001 < .0001 < .0001
 0 Reference Reference Reference
 1-3 2.09 (1.97-2.23) 1.83 (1.73-1.94) 1.87 (1.76-1.99)
 >3 4.17 (3.91-4.46) 3.46 (3.26-3.66) 3.75 (3.52-4.00)
Treatment arm < .0001 < .0001 < .0001
 Control Reference Reference Reference
 Experimental 0.88 (0.84-0.92) 0.89 (0.85-0.92) 0.91 (0.87-0.96)
  • Abbreviations: BMI, body mass index.
  • a Value represents the test score from a Cox regression model after stratifying by study treatment group.
  • b P values were determined with the Wald chi-square test.
  • c Grade1-2 was defined as well/moderate differentiation, and grade 3-4 was defined as poor differentiated/undifferentiated. Data were missing for 9205 patients.

A significant interaction between obesity and patient sex was observed in which obese men (HR, 1.14; 95% CI, 1.05-1.24; P = .0023), but not obese women (HR, 1.00; P = .9441), had significantly shorter TTR (Pinteraction = .0345) (Table 3). Furthermore, underweight and obese men had significantly poorer DFS and OS compared with women (Table 3, Fig. 1A-D). Men with class 2 and 3 obesity had significantly inferior outcomes for TTR (P = .0127), DFS (P = .0096), and OS (P = .0242) compared with women (all P > .45) (Table 3). The correlation between BMI and survival outcomes did not differ significantly according to patient age or tumor stage (data not shown).

In a multivariable analysis, the BMI category was associated significantly with TTR, DFS, and OS after adjusting for covariates (see Table 4). Compared with normal-weight patients, obese patients had poorer DFS and OS (HR, 1.10; 95% CI, 1.04-1.17; P = .0023), even after adjusting for age, stage, treatment, and sex. Overweight and normal-weight patients had similar outcomes, whereas underweight patients had significantly worse TTR (P = .0044), DFS (HR, 1.18; 95% CI, 1.09-1.28; P < .0001), and OS (HR, 1.21; 95% CI, 1.11-1.32; P < .0001) after adjusting for covariates (see Table 4).

Table 4. Multivariable Analysis
Time to Recurrence Disease-Free Survival Overall Survival
Variable HR (95% CI) Pa HR (95% CI) Pa HR (95% CI) Pa
All patients, N = 25,291
 BMI category, overall .0073 < .0001 < .0001
  Underweight vs normal 1.13 (1.04-1.24) .0044b 1.18 (1.09-1.28) < .0001b 1.21 (1.11-1.32) < .0001b
  Overweight vs normal 0.99 (0.94-1.04) .7258b 0.97 (0.92-1.02) .1912b 0.99 (0.94-1.04) .6074b
  Obese vs normal 1.06 (1.00-1.13) .0707b 1.06 (1.00-1.13) .0337b 1.10 (1.04-1.17) .0023b
   Class 1 vs normal 1.05 (0.98-1.12) .1797b 1.05 (0.98-1.12) .1526b 1.10 (1.02-1.18) .0084b
   Class 2-3 vs normal 1.08 (0.98-1.20) .1194b 1.10 (1.01-1.21) .0362b 1.11 (1.00-1.23) .0450b
 Age, 1-y increase 1.00 (1.00-1.00) .5483 1.01 (1.01-1.01) < .0001 1.01 (1.01-1.02) < .0001
 Stage, III vs II 2.77 (2.61-2.93) < .0001 2.34 (2.23-2.47) < .0001 2.43 (2.30-2.57) < .0001
 Treatment, experimental vs control 0.87 (0.83-0.91) < .0001 0.88 (0.85-0.92) < .0001 0.91 (0.87-0.95) < .0001
 Sex, men vs women 1.05 (1.01-1.10) .0190 1.13 (1.08-1.18) < .0001 1.14 (1.09-1.19) < .0001
Men, N = 13,736
 BMI category, overall .0009 < .0001 < .0001
  Underweight vs normal 1.22 (1.05-1.42) .0078b 1.31 (1.15-1.50) < .0001b 1.39 (1.21-1.60) < .0001b
  Overweight vs normal 0.97 (0.90-1.03) .3314b 0.94 (0.88-1.00) .0430b 0.95 (0.89-1.02) .1339b
  Obese vs normal 1.10 (1.01-1.20) .0228b 1.09 (1.01-1.17) .0360b 1.11 (1.02-1.21) .0137b
   Class 1 vs normal 1.09 (0.99-1.20) .0659b 1.06 (0.98-1.16) .1562b 1.10 (1.00-1.20) .0517b
   Class 2-3 vs normal 1.14 (0.98-1.33) .0792b 1.16 (1.01-1.33) .0297b 1.16 (1.00-1.35) .0452b
 Age, 1-y increase 1.00 (1.00-1.00) .1847 1.01 (1.01-1.02) < .0001 1.02 (1.01-1.02) < .0001
 Stage, III vs II 2.78 (2.57-3.00) < .0001 2.28 (2.13-2.44) < .0001 2.35 (2.18-2.53) < .0001
 Treatment, experimental vs control 0.86 (0.81-0.91) < .0001 .88 (0.83-0.92) < .0001 0.91 (0.86-0.97) .0017
Women, N = 11,555
 BMI category, overall .4597 .2117 .1070
  Underweight vs normal 1.09 (0.98-1.21) .1255b 1.11 (1.01-1.23) .0362b 1.12 (1.00-1.25) .0455b
  Overweight vs normal 1.04 (0.96-1.12) .3776b 1.03 (0.96-1.11) .4506b 1.05 (0.97-1.14) .1970b
  Obese vs normal 1.01 (0.93-1.11) .7637b 1.04 (0.96-1.13) .3268b 1.09 (1.00-1.20) .0553b
   Class 1 vs normal 1.00 (0.90-1.11) .9828b 1.03 (0.94-1.14) .5091b 1.10 (0.99-1.23) .0655b
   Class 2-3 vs normal 1.04 (0.91-1.19) .5827b 1.06 (0.93-1.21) .3548b 1.07 (0.93-1.24) .3258b
 Age, 1-y increase 1.00 (0.99-1.00) .0191 1.00 (1.00-1.01) .0051 1.01 (1.01-1.01) < .0001
 Stage, III vs II 2.76 (2.52-3.01) < .0001 2.45 (2.26-2.65) < .0001 2.55 (2.34-2.79) < .0001
 Treatment, experimental vs control 0.89 (0.83-0.95) .0004 0.89 (0.84-0.95) .0003 0.91 (0.85-0.97) .0064
  • Abbreviations: BMI, body mass index; CI, confidence interval; HR, hazard ratio.
  • a Likelihood-ratio P value was calculated after stratifying by study treatment group.
  • b Wald chi-square P value.

The BMI category was significantly prognostic in men (Table 4, Fig. 1A,C) for TTR (P = .0009), DFS (P < .0001), and OS (P < .0001) but not in women (all P > .10; OS, Pinteraction = .0129) (Table 4, Fig. 1B,D). Obese and underweight men had significantly poorer clinical outcomes versus normal-weight men (Table 4). A greater impact of class 2 and 3 obesity versus class 1 obesity on DFS and OS rates was observed among men (Table 4). The significant interaction between BMI and clinical outcome variables was primarily because underweight men, but not women, had inferior TTR (HR, 1.22; 95% CI, 1.05-1.42; P = .0078), DFS (HR, 1.31; 95% CI, 1.15-1.50; P < .0001), and OS (HR, 1.39; 95% CI, 1.21-1.60; P < .0001; Pinteraction = .0340) compared with normal-weight men.

We explored whether a curvilinear or quadratic relation could describe the observed results for BMI. By using restricted cubic splines, we observed that the continuous BMI (using 4 knots) displayed a significant curvilinear relation with OS both overall (P < .025) (Fig. 2A) and according to patient sex (Pinteraction = 0.05) (Fig. 2B). Underweight men had worse OS compared with underweight women, and obesity played a comparatively lesser role in predicting poor OS (Fig. 2B). The relation between continuous BMI and DFS and TTR was more quadratic in nature, in which only 3 knots were significant using restricted cubic splines. Given this finding, we modeled BMI as a quadratic variable in multivariate Cox models and observed that the quadratic BMI was associated significantly with DFS (P < .0001) and TTR (P = .0008) after adjusting for age, stage, treatment, and sex.

Details are in the caption following the image

Body mass index (BMI) is modeled using restricted cubic splines (A) in all patients and (B) according to patient sex for overall survival (OS). Hazard ratios (HRs) for OS are illustrated for BMI values in kg/m2 (range, 15-40 kg/m2 relative to no effect). Plots used 4 knots16, and all were statistically significant (P < .025). Instead of setting the y-axis in a log scale, transformation to an HR scale was performed to enhance interpretability.

We assessed whether BMI was predictive of benefit from 5-FU–based adjuvant chemotherapy among patients with stage II and III disease. We examined 8 adjuvant studies in which a treatment benefit was observed. Among those 8 trials, 6 evaluated 5-FU versus observation, and 2 evaluated 5-FU plus oxaliplatin versus 5-FU. There was no statistically significant interaction for BMI and treatment for TTR, DFS, or OS in either univariate or multivariable analysis. The multivariate interaction models revealed a continued treatment benefit across all BMI categories after adjusting for age, stage, and sex. Furthermore, there was no evidence for a differential treatment effect by adjuvant chemotherapy regimen.

DISCUSSION

In this study, we used BMI measured by trained personnel at the time patients entered an adjuvant study. We observed that obese patients, but not overweight patients, had significantly poorer survival compared with normal-weight patients after adjustment for covariates, and this effect was most evident among patients who had severe (class 2 and 3) obesity. An important finding of our study is that an adverse impact of obesity on colon cancer outcomes was limited to men. Severely obese men had a 16% increase in mortality relative to normal-weight patients. Another important finding is that underweight patients had increased cancer recurrence and inferior outcomes. A statistically significant interaction was observed between underweight status and patient sex, whereby underweight men had a 39% increase in all-cause mortality compared with normal-weight men or women. In previous studies, the inferior outcome among underweight patients with cancer was attributed to noncancer–related deaths.17, 18 However, we observed a shorter TTR and DFS for underweight patients, suggesting that the impact on prognosis is cancer-related.

Inconsistent data exist for the impact of patient sex on outcome in obese patients with colon cancer. In patients with stage II and III colon cancers, the association between obesity and mortality was reported to be stronger among women than among men.9 In contrast, no differences according to sex were observed in another study that also examined data from colon cancer adjuvant trials.12 Our finding of a stronger association of obesity with adverse outcome among men versus women is consistent with the reported higher rate of incident colon cancers among obese men versus women.4, 6, 9, 20 Mechanisms underlying this observation may be related to body fat distribution, because BMI is related more closely to abdominal or central adiposity in men.21, 22 Abdominal adiposity is associated with hyperinsulinemia, insulin resistance, and the insulin-like growth factor-I axis as potential mediators of increased CRC risk and mortality.23-25 The attenuated impact of obesity on colon cancer outcomes observed among women versus men in our study may be because of an effect modification by estrogen. Estrogen levels correlate with BMI in postmenopausal women, because their major source is conversion from androgens in adipose tissue.26 The association of obesity with CRC risk is reduced after menopause,27 and hormone-replacement therapy is consistently associated with reduced colon cancer mortality.28-30

Excess mortality among underweight patients with colon cancer has been attributed to noncancer causes, and mainly to chronic respiratory conditions.18 In our study, differences in performance status based on BMI were not clinically meaningful, and strict eligibility criteria for the adjuvant studies excluded patients who had significant comorbidities. Because BMI was recorded at adjuvant trial enrollment, it will be important to distinguish between patients who are underweight but have stable weight over time versus those who experienced significant cancer-related weight loss before trial enrollment. Significant cancer-related weight loss may identify a poor prognostic subgroup, because cancer cachexia is associated with inferior outcomes.31-33 Loss of adipose tissue accounts for the majority of the cancer-related weight loss, yet the preferential loss of skeletal muscle adversely impacts mortality.34-36 Although it was not included in the ACCENT database, cigarette smoking is associated with a lower BMI, and current smokers who are underweight or obese have high mortality rates,37 especially among men.38, 39

Evidence suggests that the time frame during which BMI is determined can influence its association with clinical outcome. A prospective cohort study found that self-reported BMI prediagnosis (mean, 7 years before CRC diagnosis) was associated independently with a statistically significant increase in the risk of all-cause and cancer-specific mortality, whereas postdiagnosis BMI (mean interval, 1.5 years after diagnosis) was not.13 Prediagnosis BMI measurements have also demonstrated a higher risk of all-cause mortality in obese women versus normal-weight women.17, 40

We also determined whether BMI was predictive of a clinical benefit in patients who received 5-FU–based adjuvant chemotherapy versus observation or no 5-FU. Our data indicate that adjuvant chemotherapy is beneficial for patients in high-risk and low-risk BMI categories, and outcomes were similar among those who received older versus more modern adjuvant regimens as well as among North American and European patients.

Strengths of our study include BMI measurements performed by trained staff and the rigorous collection of data on recurrence and survival within clinical trials over an extended follow-up period. Limitations include the retrospective study design and lack of data on smoking, diet, physical activity, menopausal status, or receipt of hormone replacement therapy, which may have independent associations with outcomes and may inform the interpretation of sex-related differences. It is noteworthy that neither lifestyle factors nor demographic factors had an impact on the association of prediagnosis BMI with survival among patients who had colon cancer in a large cohort study.13 Although the large sample size within the ACCENT database indicates that modest absolute differences in clinical outcomes may be statistically significant, our results must be interpreted in that context.

In summary, obese and underweight BMI are associated with increased mortality in colon cancer survivors, especially among men. In underweight patients, shorter TTR and reduced DFS rates suggest increased tumor aggressiveness. Together, these data suggest that interventions to modify patient BMI after a colon cancer diagnosis have the potential to improve patient outcomes.

FUNDING SOURCES

This work was supported by a National Cancer Institute Senior Scientist Award (K05CA-142885 to Dr. Sinicrope) and by a North Central Cooperative Treatment Group Biospecimen Resource National Institutes of Health grant (CA-114740).

CONFLICT OF INTEREST DISCLOSURE

The authors made no disclosures.