Associations between Smoking, Alcohol Consumption, and Colorectal Cancer, Overall and by Tumor Microsatellite Instability Status

Cigarette smoking and alcohol consumption have been evaluated extensively as risk factors for colorectal cancer (CRC). The relationship between smoking and CRC has not been entirely clear; however, two recently published meta-analyses have reported a modest, but statistically significant, increased risk of CRC for smokers (relative risk, 1.20 for 40 years of duration) with evidence for a stronger association with rectal cancer (1, 2). Although not entirely consistent, epidemiologic studies also suggest that alcohol consumption is associated with increased CRC risk (3, 4). The 2007 World Cancer Research Fund/American Institute for Cancer Research report concluded that alcohol consumption was a risk factor for CRC with a summary relative risk of 1.06 [95% confidence interval (CI), 1.01-1.12] per 10 g of alcohol/d in a meta-analysis of nine cohort studies (5).¹¹

Microsatellite unstable (MSI-H) tumors, which account for 10% to 20% of sporadic CRC, differ from microsatellite stable (MSS) tumors by clinical and patient characteristics. MSI-H tumors are associated with location in the right colon, female sex, older age at diagnosis, and more favorable prognosis when compared with MSS tumors (6-9). Thus, it is plausible that risk factors differ between MSI-H and MSS CRC. Previous epidemiologic studies have reported that the associations between CRC and both smoking (10-13) and alcohol consumption (12, 14, 15) differ by tumor MSI status, with both factors increasing the risk of MSI-H cancers in particular.

In this analysis, we evaluated the relationships between smoking and alcohol consumption and CRC risk overall, and by tumor location and MSI status, using data from the Colorectal Cancer Family Registry (Colon CFR).

Individuals were recruited from the six centers of the Colon CFR: University of Hawaii (Honolulu, HI), Fred Hutchinson Cancer Research Center (Seattle, WA), Mayo Clinic (Rochester, MN), University of Southern California Consortium (Los Angeles, CA), Cancer Care Ontario (Toronto, ON), and the University of Melbourne (Victoria, Australia). All centers recruited cases of CRC from population-based cancer registries. Some centers recruited all incident cases of CRC whereas others oversampled cases with a family history of CRC or early age at onset. Standardized procedures were used to collect epidemiologic data, and tumor blocks from cases were obtained from the Colon CFR Jeremy Jass Memorial Pathology Bank. Detailed information about the Colon CFR can be found online¹²

Data for this analysis were abstracted from the Colon CFR central data repository in October 2007 and included population-based cases diagnosed with colon or rectal cancer from July 1998 to July 2005. We included sibships in which both a case and at least one unaffected sibling control had completed the epidemiologic risk factor questionnaire. Data from 2,248 sibships (N = 2,253 cases and 4,486 unaffected siblings) were available for the analysis of smoking, alcohol consumption, and overall risk of CRC. Approximately 50% of the cases had one unaffected sibling control, an additional 25% had two unaffected sibling controls, and the remaining cases had more than two sibling controls (range 3-10). All of the cases were interviewed within 3 y of diagnosis (86% within 2 y). In the analyses stratified by MSI status, we included 1,564 cases with microsatellite instability data and their unaffected siblings (N = 3,102). For the sex-specific analyses, we included 753 male cases (532 with MSI data) and 1,806 unaffected brothers as well as 806 female cases (553 with MSI data) and 1,752 unaffected sisters. A subset of the cases in this analysis were included in a previously published analysis of smoking, nonsteroidal anti-inflammatory drug use, and MSI-H CRC (13).

All participants provided informed consent. The study was approved by the Institutional Review Board at each of the Colon CFR sites.

Epidemiologic data were collected using a risk factor questionnaire administered to all participants at recruitment. A detailed smoking history was collected, including information on current and former smoking habits. Information was collected for cigarette, pipe, and cigar smoking; however, only cigarette smoking was included in these analyses due to the small number of regular cigar and pipe smokers (n = 154 exclusive pipe/cigar smokers; n = 450 current and 123 former cigarette smokers also reported pipe/cigar smoking). Individuals were categorized as current/former/never smokers. In addition, we evaluated dose and duration of smoking, including the number of cigarettes smoked per day, total number of years of regular smoking, and pack-years of smoking.

Participants were asked to report alcohol consumption during three time periods: ages 20 to 30, 30 to 50, and after age 50 years. We evaluated the number of drinks per week of all types of alcohol during the time period corresponding to the age of the participant at completion of the risk factor questionnaire. For example, we evaluated alcohol consumption during the ages 30 to 50 years for an individual 45 years old when the questionnaire was completed.

Data were also collected for potential confounders, including age, sex, income, education, body mass index, ethnicity, physical activity (average weekly total lifetime MET hours), and regular nonsteroidal anti-inflammatory drug use (never/former/current; ref. 16).

Tumors located in the cecum, ascending colon, hepatic flexure, transverse colon, and splenic flexure (International Classification of Diseases for Oncology, third edition codes C180, C182, C183, C184, and C185; ref. 17) were classified as right colon. Tumors located in the descending colon and sigmoid colon (International Classification of Diseases for Oncology, third edition codes C186 and C187) were classified as left colon. Rectal tumors included the rectosigmoid junction and rectum (International Classification of Diseases codes C199 and C209).

Microsatellite instability was evaluated for all cases with available tumor tissue using a panel of 10 markers (BAT25, BAT26, BAT40, MYCL, D5S346, D17S250, ACTC, D18S55, D10S197, and BAT34C4; ref. 16). Results were required for at least four markers to determine MSI status. Tumors were deemed MSI-H if instability was observed at ≥30% of markers, MSI-L if >0% and <30% of markers were unstable, and MSS if all markers were stable. Tumor MSI results were available for 69% of the cases included in this analysis.

Mutations in the mismatch repair genes MSH2, MLH1, and MSH6 were screened for in selected participants in the Colon CFR using a combined approach of denaturing high-pressure liquid chromatography/direct sequencing and multiplex ligation-dependent probe amplification. All clinic-based probands, all MSI-H or MSI-L population-based probands, and a sample of 300 MSS population-based probands were screened.

We estimated associations using multivariable conditional logistic regression with sibship as the matching factor. Potential confounders that changed the odds ratio (OR) by >10% were included in the final model. Tests for linear trend were done by treating a categorical variable as a continuous variable in the logistic regression model. We evaluated differences in the associations by MSI status and by tumor location by stratifying the matched sets on the tumor characteristics of the case. We assigned the unaffected sibling to the same MSI or tumor site category as the case and included interaction terms in the logistic regression models to estimate these stratum-specific ORs. Heterogeneity of the ORs by tumor location and MSI status was evaluated using a likelihood ratio test comparing a model with interaction terms for the stratum-specific associations with a model including main effects of the respective smoking or alcohol variable. Analyses were repeated after excluding individuals with a known mismatch repair gene mutation. Statistical analyses were done using SAS v9.1 for Windows (SAS Institute).

Selected characteristics of the study population are presented in Table 1. The age distribution was similar for CRC cases and unaffected siblings, and the majority of the study population reported non-Hispanic white ethnicity. As expected, the majority of MSI-H tumors were located in the proximal colon, whereas MSI-L and MSS tumors were more equally distributed throughout the large bowel (i.e., including the rectum; Table 1).

We observed no overall association between CRC and current smoking status, duration of smoking, number of cigarettes smoked per day, or pack-years of smoking in age-adjusted and sex-adjusted models (Table 2). No potential confounders changed the OR >10% and the more parsimonious models are presented. Adjustment for pipe and cigar smoking also did not alter the risk estimates and was not included in the final models. There were statistically significant associations for the highest categories of cigarettes per day and pack-years for rectal, but not colon cancer, with a statistically significant trend over pack-years. The relationship between smoking and CRC did not differ for males and females (for >40 pack-years of smoking: OR, 1.10; 95% CI, 0.78-1.55; P_trend = 0.24 for males versus OR, 1.33; 95% CI, 0.85-2.07; P_trend = 0.51 for females).

In the overall analysis, we observed a small, but statistically significant, increase in risk of CRC in individuals who consumed more than 12 drinks per week compared with nondrinkers (OR, 1.21; 95% CI, 1.03-1.44; P_trend = 0.02). This association was observed only for males (OR, 1.24; 95% CI, 0.94-1.65; P_trend = 0.05 for males; OR, 0.92; 95% CI, 0.62-1.35; P_trend = 0.78 for females). In males who consumed alcohol, 28% reported drinking beer, 9% reported drinking wine, 9% reported drinking hard liquor, and 54% reported drinking more than one type of alcoholic beverage. In women, the corresponding frequencies were 10%, 38%, 13%, and 39% for beer, wine, hard liquor, and more than one type of alcoholic beverage, respectively. The reported median alcohol consumption was seven drinks per week in both males and females. Alcohol consumption was statistically significantly associated with an increased risk of rectal cancer, whereas the associations for right or left colon cancer did not reach statistical significance (Table 2). This association was also restricted to males (OR, 1.60; 95% CI, 0.95-2.68; P_trend = 0.03 in males versus OR, 0.87; 95% CI, 0.41-1.84; P_trend = 0.94 in females).

In analyses stratified by tumor MSI status, we observed no association between any smoking variable and risk of MSS CRC (Table 3). However, there was an increased risk of MSI-H CRC with increasing duration of smoking (P_trend = 0.02) and a statistically nonsignificant increase for pack-years of smoking (P_trend = 0.10), although the tests for heterogeneity of ORs by MSI status were not statistically significant. We also observed a statistically significant association between >40 pack-years of smoking and MSI-L CRC (OR, 2.03; 95% CI, 1.03-4.00). Alcohol consumption was associated with MSI-L CRC in both males and females (OR, 2.02; 95% CI, 0.78-5.26; P_trend = 0.08 in males versus OR, 12.3; 95% CI, 1.26-120; P_trend = 0.006 in females).

We repeated the analyses stratified by MSI after excluding 36 MSI-H cases with a known mismatch repair gene mutation (12 MLH1, 18 MSH2, and 6 MSH6). The results did not differ substantially from those shown above (data not shown).

Overall, we did not observe a statistically significant association between smoking and CRC. Duration and pack-years of smoking were associated with an increased risk of MSI-H CRC; however, the tests for heterogeneity across categories of MSI status did not reach statistical significance. Our data also suggest that the associations between CRC and both smoking and alcohol consumption were stronger for rectal cancer than for colon cancer, although these differences were not statistically significant.

Results from two recent meta-analyses have suggested a small, but statistically significant association between CRC and smoking status, with evidence that the association is apparent following a long duration of smoking (1, 2). Although we did not observe any statistically significant associations for CRC overall, our data are compatible with the summary ORs estimated in these reports. The meta-analysis data also suggested a stronger association for rectal cancer than for colon cancer (1, 2), in general agreement with our findings.

Our data suggest that alcohol consumption is associated with a small increased risk of CRC only for males, and that the association might be stronger for rectal cancer than for colon cancer. A pooled analysis (4) and a meta-analysis (3) of prospective cohort studies have confirmed an increased risk of CRC for individuals who consume large amounts of alcohol. However, data evaluating different effects of alcohol by anatomic location in the colorectum have been mixed, and the data from prospective cohorts suggest that the association with alcohol is similar for colon and rectal cancers (3, 4). Previous studies have reported a stronger association between alcohol consumption and CRC in men than in women, and the World Cancer Research Fund/American Institute for Cancer Research report concluded that there was convincing evidence for the role of alcohol in CRC in males and probable evidence for a role in females (5).

Several studies have previously reported a stronger association between smoking and MSI-H CRC than for MSS/MSI-L CRC. This association has biological plausibility, as sporadic MSI-H tumors have been hypothesized to arise from a serrated pathway rather than through the traditional adenoma-carcinoma sequence (18), and the association between smoking seems to be stronger for serrated lesions than for adenomatous polyps (19, 20). The stronger association for MSI-H CRC was first reported by Slattery et al. (10) in a large population-based case-control study of colon cancer (n = 266 MSI-H cases). Three additional studies using either case-control (11, 13) or case-case (12) comparisons have also supported a stronger association between smoking and MSI-H CRC. In contrast, results from two other studies have not supported this hypothesis (21, 22). A more recent publication suggested that the association between smoking and CRC may be restricted to tumors with the CpG island methylator phenotype, regardless of MSI status (23). The CpG island methylator phenotype status was not available for the tumors included in this analysis, so we were unable to evaluate this hypothesis.

The association between alcohol consumption and MSI status has been evaluated previously using case-control and case-case comparisons (12, 14, 15, 22). Three of these studies suggested that alcohol intake was more strongly associated with MSI-H CRC (12, 14, 15), although the associations were modest and did not reach statistical significance in all studies. These studies did not evaluate the cases with MSI-L tumors as a separate subgroup, and our data suggest that the association between alcohol consumption and CRC risk is strongest in this subgroup. Further studies will be required to clarify this interesting relationship.

This study has several strengths, including the large number of tumors with microsatellite instability typing and the inclusion of MSI-L tumors as a separate subgroup. However, several limitations must be considered. This study uses a case-unaffected sibling design, and it is possible that this may have reduced our power to detect associations between CRC and the lifestyle factors investigated here if these factors were strongly correlated within families. Smoking and alcohol consumption were correlated within sibships in our study population (r = 0.21, P < 0.0001 for pack-years of smoking and r = 0.11, P < 0.0001 for number of drinks per week) attenuating the association of these factors with CRC. There is also the possibility that the unaffected siblings have underlying, but undetected, CRCs, which would lead to misclassification of the outcome variable. Recall bias may lead to inaccurate risk estimates if smoking and alcohol consumption were reported more accurately by cases than unaffected siblings. In addition, microsatellite instability status was not available for tumors from 31% of cases in this study; however, the overall associations between smoking, alcohol consumption, and CRC risk did not differ substantially for cases with and without tumor MSI data.

In conclusion, we observed no statistically significant associations between smoking and CRC overall; however, we did observe statistically significant associations between smoking and rectal cancer, smoking duration and MSI-H CRC, and pack-years of smoking and MSI-L CRC. Alcohol consumption was associated with a modest increased risk of CRC in males, and was also significantly associated with MSI-L CRC in both males and females. These results highlight the potential importance of incorporating tumor characteristics in studies of risk factors for CRC.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

The authors thank the CFR study coordinators and data managers who helped prepare the data set for these analyses as well as the participants in the Colon CFR who have generously donated their time for this project.