Abstract
The effect of duration of cigarette smoking cessation on colorectal cancer risk by molecular subtypes remains unclear. Using duplication-method Cox proportional-hazards regression analyses, we examined associations between duration of smoking cessation and colorectal cancer risk according to status of CpG island methylator phenotype (CIMP), microsatellite instability, v-raf murine sarcoma viral oncogene homolog B1 (BRAF) mutation, or DNA methyltransferase-3B (DNMT3B) expression. Follow-up of 134,204 individuals in 2 US nationwide prospective cohorts (Nurses' Health Study (1980–2008) and Health Professionals Follow-up Study (1986–2008)) resulted in 1,260 incident rectal and colon cancers with available molecular data. Compared with current smoking, 10–19, 20–39, and ≥40 years of smoking cessation were associated with a lower risk of CIMP-high colorectal cancer, with multivariate hazard ratios (95% confidence intervals) of 0.53 (0.29, 0.95), 0.52 (0.32, 0.85), and 0.50 (0.27, 0.94), respectively (Ptrend = 0.001), but not with the risk of CIMP-low/CIMP-negative cancer (Ptrend = 0.25) (Pheterogeneity = 0.02, between CIMP-high and CIMP-low/CIMP-negative cancer risks). Differential associations between smoking cessation and cancer risks by microsatellite instability (Pheterogeneity = 0.02), DNMT3B expression (Pheterogeneity = 0.03), and BRAF (Pheterogeneity = 0.10) status appeared to be driven by the associations of CIMP-high cancer with microsatellite instability–high, DNMT3B-positive, and BRAF-mutated cancers. These molecular pathological epidemiology data suggest a protective effect of smoking cessation on a DNA methylation–related carcinogenesis pathway leading to CIMP-high colorectal cancer.
Keywords: carcinogen, carcinoma, hypermethylation, epigenomics, molecular epidemiology, public health, tobacco, translational epidemiology
Smoking is a risk factor for several cancers, including colorectal cancer, and remains a global health problem (1, 2). Although the carcinogenic effect of smoking is not refutable, the effect of duration of smoking cessation on colorectal cancer risk remains unclear. Beyond a simple comparison of former versus current smokers, some epidemiologic studies suggest a modest association between duration of smoking cessation and risk reduction in overall colorectal cancer incidence compared with continued smoking (3, 4), whereas other studies did not confirm this association (5–7).
Colorectal cancers are a heterogeneous group of neoplasms displaying a complex mixture of epigenetic and genetic alterations (8). Molecular classification of colorectal cancer has become crucial for epidemiologic research and clinical decision making (8–11). The CpG island methylator phenotype (CIMP) is a form of epigenomic instability characterized by widespread promoter CpG island hypermethylation (12–16), and microsatellite instability (MSI) represents a distinct form of genomic instability (8, 17). A high degree of CIMP in colorectal cancer (CIMP-high) is associated with v-raf murine sarcoma viral oncogene homolog B1 (BRAF) oncogene mutation as well as a high degree of MSI (through epigenetic silencing of MLH1) (13–15, 18–20). Experimental and observational evidence suggests that DNA methyltransferase-3B (DNMT3B) expression could contribute to CIMP in colorectal cancer (21–25). Epidemiologic studies suggest that cigarette smoking is associated with higher risks for specific molecular subtypes of colorectal cancer—namely, CIMP-high (26–28), MSI-high (26, 28–34), and BRAF-mutated (26–28, 35) cancers. However, to our knowledge, no previous study has prospectively examined duration of smoking cessation and colorectal cancer incidence by tumor epigenetic subtyping. Experimental evidence suggests that cigarette smoking could affect epigenetic status and induce hypermethylation in CpG islands (36–38). Therefore, we hypothesized that duration of smoking cessation might be associated specifically with a decreased risk of CIMP-high colorectal cancer.
We conducted a molecular pathological epidemiology (MPE) (10, 11) study to prospectively examine the relation between duration of smoking cessation and colorectal cancer risk by epigenetics-related tumor classifications, including status of CIMP, MSI, BRAF mutation, and DNMT3B expression. Studies have shown that these tumor molecular features are interrelated (13–15, 18–28, 34, 35). For this purpose, we used tumor specimens of 1,260 incident colorectal cancer cases from 2 US nationwide prospective cohort studies with more than 134,000 participants.
MATERIALS AND METHODS
Study population
Details on our study population are described in the Web Appendix (available at http://aje.oxfordjournals.org/). Briefly, we used the Nurses' Health Study and the Health Professionals Follow-up Study (39, 40). Questionnaires were sent to participants every 2 years to update information on smoking status and other lifestyle factors. A total of 88,397 women and 45,807 men were eligible for inclusion in the analysis. Informed consent was obtained from all participants. This study was approved by the Human Subjects Committees at Harvard School of Public Health and Brigham and Women's Hospital.
Assessment of smoking status
Details on the method used to obtain information on smoking have been reported previously (41, 42). Current smoking status and the number of cigarettes smoked per day were reported by participants on questionnaires updated every 2 years, beginning in 1980 for women and in 1986 for men. In addition, at the cohort baseline questionnaires, we collected information on age when smoking was started, age when smoking was stopped (for former smokers), and pack-years smoked before age 30 years. Thus, we could calculate the duration of smoking cessation and cumulative pack-years smoked (cumulative average of packs per day × the number of years during which smoking occurred).
Assessment of incident colorectal cancer
Details on the assessment of incident colorectal cancer are described in the Web Appendix. Briefly, we obtained the information from biennial questionnaires, medical records, and the National Death Index (43). On the basis of the colorectal continuum model, we used both colon and rectal cancers as outcomes (43, 44). We retrieved formalin-fixed paraffin-embedded colorectal cancer tissue blocks from hospitals throughout the United States at which participants with colorectal cancer had undergone surgical resection (45).
Assessment of tumor characteristics
Detailed methods of the assessment of tumor characteristics are described in the Web Appendix. We conducted DNA extraction, Pyrosequencing of BRAF (codon 600) (46), MSI analysis (20), and methylation analysis for 8 CpG islands (18, 20, 47), using validated bisulfite DNA treatment and real-time polymerase chain reaction (MethyLight assay) (48). We performed immunohistochemistry for DNMT3B (22).
Statistical methods
We used Cox proportional-hazards model to estimate hazard ratios, with adjustment for multiple potential confounders. For each 2-year interval, we used the most up-to-date questionnaire data for all covariates before the next follow-up cycle. We treated all variables as time-dependent variables to take into account changes over time (39). Follow-up ended at diagnosis of colorectal cancer, death from other causes, or June 30, 2008, whichever came first. To reduce within-individual variation and to better estimate long-term influence, we used cumulative average for relevant variables, which was the mean of all available data up to before each biennial follow-up cycle (39). Covariates included body mass index (weight (kg)/height (m)2; <25 vs. 25–30 vs. ≥30); history of colorectal cancer in any first-degree relative (yes vs. no); regular use of aspirin (2 or more tablets per week or at least 2 times per week vs. less); physical activity level (quintiles of mean metabolic equivalent task hours per week); alcohol consumption (0 gram per day or quartiles of grams per day); total caloric intake (quintiles of calories per day) and red meat intake (quintiles of servings per day). Models were stratified with calendar year of the questionnaire cycle, age in month, and sex (only in combined cohorts). We observed no evidence for a violation of the proportional hazard assumption on the basis of the interaction terms between smoking status and follow-up time (P > 0.1 for all the combination of smoking variables and colorectal cancer outcomes). The linear trend test was conducted by using the median value of each category. We examined the possibly nonlinear relation between years of smoking cessation and colorectal cancer risk by molecular subtypes nonparametrically using restricted cubic splines (49). To compare differential associations of smoking with colorectal cancer risk by molecular subtypes, we conducted duplication-method Cox proportional hazards model (50). This methodology permits the estimation of separate regression coefficients for smoking status stratified by the type of outcome. Using a likelihood ratio test, we examined whether smoking conferred differential risk by molecular subtype (e.g., CIMP-low/negative vs. CIMP-high). All P values were two-sided. All statistical analyses were performed using SAS version 9.2 (SAS Institute, Inc., Cary, North Carolina). No attempt was made to adjust for multiple testing because of difficulty in determining the number of independent hypotheses tested (i.e., the smoking indicators were related and the tumor biomarkers were related). Nonetheless, statistical significance was evaluated cautiously considering the exploratory nature of the analyses and the number of biomarkers analyzed.
RESULTS
Table 1 shows the age-adjusted baseline characteristics of the study population in the Nurses’ Health Study and the Health Professionals Follow-up Study. The rate of restart of smoking was 1.5%–1.2% in 1980s and decreased in recent years (0.7%–0.5% in 2000s). We identified 1,260 incident colorectal cancers with available pathological specimens suitable for molecular analysis, during follow-up of 134,204 individuals (3,101,031 person-years). There were 205 (18% of 1,170) CIMP-high tumors, 188 (16% of 1,200) MSI-high tumors, 178 (15% of 1,218) BRAF-mutated tumors, and 108 (15% of 728; DNMT3B data were limited to those included in tissue microarray) DNMT3B-positive tumors. The relations between tumor molecular features, tumor location, and sex are shown in Web Table 1.
Table 1.
Age-adjusted Characteristics of Participants During Follow-upa According to Smoking Status in the Nurses' Health Study (1980–2008) and the Health Professionals Follow-up Study (1986–2008)
| Variable | Women (Nurses' Health Study) |
Men (Health Professionals Follow-up Study) |
||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Former Smoker |
Former Smoker |
|||||||||||||||
| Never Smoker (n = 38,576) |
Cessation for <10 Years (n = 14,289) |
Cessation for ≥10 Years (n = 9,940) |
Current Smoker (n = 25,592) |
Never Smoker (n = 21,366) |
Cessation for <10 Years (n = 13,880) |
Cessation for ≥10 Years (n = 5,934) |
Current Smoker (n = 4,627) |
|||||||||
| % | Mean (SD) | % | Mean (SD) | % | Mean (SD) | % | Mean (SD) | % | Mean (SD) | % | Mean (SD) | % | Mean (SD) | % | Mean (SD) | |
| Total person-years | 511,458 | 325,952 | 124,626 | 191,635 | 188,401 | 156,652 | 36,951 | 30,770 | ||||||||
| Ageb | 68.9 (8.7) | 70.0 (7.9) | 68.7 (8.3) | 68.1 (8.0) | 61.1 (11.0) | 65.1 (10.4) | 59.8 (10.4) | 59.8 (9.8) | ||||||||
| Body mass indexc | ||||||||||||||||
| <25 | 70 | 70 | 74 | 79 | 69 | 66 | 64 | 71 | ||||||||
| 25–29.9 | 21 | 21 | 19 | 16 | 27 | 29 | 30 | 25 | ||||||||
| ≥30 | 9 | 9 | 7 | 5 | 5 | 5 | 6 | 4 | ||||||||
| Family history of colorectal cancer in any first-degree relative | 13 | 13 | 12 | 11 | 12 | 12 | 11 | 11 | ||||||||
| Regular use of aspirin | 40 | 42 | 43 | 42 | 45 | 49 | 49 | 45 | ||||||||
| Postmenopausal hormone use (ever) | 63 | 68 | 63 | 56 | N/A | N/A | N/A | N/A | ||||||||
| Physical activity, MET-hours/weekd | 15.8 (17.5) | 17.3 (19.4) | 15.6 (18.6) | 13.5 (17.6) | 31.3 (29.4) | 30.2 (28.1) | 25.2 (25.2) | 23.0 (24.5) | ||||||||
| Alcohol consumption, g/day | 3.8 (7.0) | 7.1 (9.3) | 7.8 (10.4) | 9.1 (12.4) | 7.9 (11.1) | 13.1 (14.6) | 14.7 (16.3) | 16.8 (18.8) | ||||||||
| Total calories, kcal/day | 1,697 (449) | 1,672 (430) | 1,638 (439) | 1,637 (463) | 1,985 (554) | 1,966 (549) | 1,970 (571) | 2,012 (589) | ||||||||
| Red meat intake, servings/day | 1.1 (0.6) | 1.0 (0.6) | 1.1 (0.6) | 1.2 (0.7) | 1.1 (0.8) | 1.1 (0.8) | 1.3 (0.9) | 1.5 (0.9) | ||||||||
| Cumulative pack-years | N/A | 13.1 (13.5) | 29.6 (20.8) | 40.3 (21.8) | N/A | 19.5 (15.6) | 32.0 (22.5) | 39.7 (24.6) | ||||||||
| Pack-years smoked before age 30b | N/A | 6.8 (6.1) | 6.3 (5.0) | 6.7 (4.4) | N/A | 10.6 (6.9) | 9.9 (6.7) | 10.2 (6.6) | ||||||||
| <20 years of ageb at start of smoking, % | N/A | 59 | 56 | 58 | N/A | 54 | 50 | 51 | ||||||||
Abbreviations: MET, metabolic equivalent task; N/A, not applicable; SD, standard deviation.
a Updated information of smoking status from biennial questionnaires was averaged, using person-years in each category of smoking status up to censoring (including death from other causes) or immediately before personal colorectal cancer diagnosis if it occurred. Values were standardized to the age distribution of the study population.
b Not age-adjusted.
c Weight (kg)/height (m)2.
d MET calculated according to the frequency of a range of physical activities in 1986 for both women and men.
Web Table 2 shows cohort (sex)-specific results for smoking cessation and incident colorectal cancer risk by molecular subtypes. We conducted tests of heterogeneity using the Q statistic and observed no significant heterogeneity between the 2 cohorts (Pheterogeneity ≥ 0.05) for the associations of smoking cessation with any of the specific cancer subtypes. For further analyses, we utilized the combined cohorts to increase statistical power.
In the combined cohorts, compared with current smoker, duration of smoking cessation was not significantly associated with the risk of colorectal cancer overall (Table 2). Although smoking cessation appeared to be more protective for proximal colon cancer than for distal colorectal cancer, the difference was not statistically significant (Pheterogeneity = 0.28) (Table 2). Web Table 3 shows the risk for proximal colon cancer and distal colorectal cancer by molecular subtypes; the statistical power was limited in these subsite-specific analyses.
Table 2.
Duration of Smoking Cessation and Incident Colorectal Cancer Risk by Molecular Subtypesa in the Nurses' Health Study (1980–2008) and the Health Professionals Follow-up Study (1986–2008)
| Current Smoker (n = 439,508 person-years) |
Cessation for 1–4 Years (n = 161,905 person-years) |
Cessation for 5–9 Years (n = 155,720 person-years) |
Cessation for 10–19 Years (n = 312,757 person-years) |
Cessation for 20–39 Years (n = 511,426 person-years) |
Cessation for ≥40 Years (n = 126,688 person-years) |
Ptrendb | Pheterogeneityc | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | |||
| Cancers | ||||||||||||||
| All colorectal cancer | ||||||||||||||
| No. | 139 | 60 | 86 | 129 | 242 | 105 | ||||||||
| Age-adjusted | 1.00 | Referent | 0.98 | 0.72, 1.33 | 1.31 | 1.00, 1.71 | 0.96 | 0.75, 1.22 | 0.92 | 0.74, 1.13 | 1.02 | 0.78, 1.33 | 0.19 | |
| Multivariated | 1.00 | Referent | 0.99 | 0.73, 1.34 | 1.30 | 0.99, 1.71 | 0.96 | 0.75, 1.23 | 0.92 | 0.74, 1.14 | 1.05 | 0.80, 1.37 | 0.29 | |
| Proximal colon cancer | ||||||||||||||
| No. | 63 | 32 | 42 | 57 | 109 | 51 | ||||||||
| Age-adjusted | 1.00 | Referent | 1.16 | 0.76, 1.78 | 1.32 | 0.89, 1.96 | 0.86 | 0.60, 1.24 | 0.81 | 0.59, 1.12 | 0.82 | 0.55, 1.21 | 0.01 | |
| Multivariated | 1.00 | Referent | 1.17 | 0.76, 1.80 | 1.32 | 0.89, 1.96 | 0.86 | 0.60, 1.24 | 0.81 | 0.59, 1.12 | 0.84 | 0.57, 1.24 | 0.02 | 0.28 |
| Distal colorectal cancer | ||||||||||||||
| No. | 75 | 28 | 43 | 72 | 129 | 52 | ||||||||
| Age-adjusted | 1.00 | Referent | 0.82 | 0.53, 1.27 | 1.26 | 0.86, 1.84 | 1.03 | 0.74, 1.43 | 0.90 | 0.67, 1.20 | 0.95 | 0.65, 1.38 | 0.28 | |
| Multivariated | 1.00 | Referent | 0.83 | 0.54, 1.29 | 1.26 | 0.86, 1.84 | 1.03 | 0.74, 1.43 | 0.90 | 0.67, 1.21 | 0.96 | 0.66, 1.41 | 0.34 | |
| CIMP Status | ||||||||||||||
| CIMP-low/negative | ||||||||||||||
| No. | 103 | 42 | 66 | 105 | 194 | 72 | ||||||||
| Age-adjusted | 1.00 | Referent | 0.91 | 0.63, 1.31 | 1.37 | 1.00, 1.87 | 1.07 | 0.81, 1.41 | 0.97 | 0.76, 1.24 | 0.93 | 0.68, 1.28 | 0.17 | |
| Multivariated | 1.00 | Referent | 0.92 | 0.64, 1.32 | 1.37 | 1.00, 1.88 | 1.07 | 0.81, 1.42 | 0.98 | 0.77, 1.26 | 0.95 | 0.69, 1.32 | 0.25 | 0.02 |
| CIMP-high | ||||||||||||||
| No. | 31 | 15 | 15 | 18 | 37 | 16 | ||||||||
| Age-adjusted | 1.00 | Referent | 1.09 | 0.58, 2.02 | 0.89 | 0.48, 1.66 | 0.52 | 0.29, 0.93 | 0.52 | 0.32, 0.84 | 0.48 | 0.26, 0.90 | 0.001 | |
| Multivariated | 1.00 | Referent | 1.12 | 0.60, 2.08 | 0.89 | 0.48, 1.67 | 0.53 | 0.29, 0.95 | 0.52 | 0.32, 0.85 | 0.50 | 0.27, 0.94 | 0.001 | |
| MSI Status | ||||||||||||||
| MSS | ||||||||||||||
| No. | 108 | 40 | 68 | 101 | 201 | 86 | ||||||||
| Age-adjusted | 1.00 | Referent | 0.83 | 0.57, 1.19 | 1.34 | 0.98, 1.82 | 0.97 | 0.74, 1.28 | 0.93 | 0.73, 1.19 | 0.96 | 0.71, 1.30 | 0.26 | |
| Multivariated | 1.00 | Referent | 0.83 | 0.58, 1.20 | 1.34 | 0.98, 1.82 | 0.97 | 0.73, 1.28 | 0.94 | 0.74, 1.20 | 0.98 | 0.72, 1.33 | 0.36 | 0.02 |
| MSI-high | ||||||||||||||
| No. | 27 | 16 | 14 | 20 | 30 | 17 | ||||||||
| Age-adjusted | 1.00 | Referent | 1.27 | 0.68, 2.37 | 0.97 | 0.51, 1.86 | 0.66 | 0.37, 1.19 | 0.50 | 0.29, 0.84 | 0.60 | 0.31, 1.13 | 0.001 | |
| Multivariated | 1.00 | Referent | 1.29 | 0.69, 2.40 | 0.96 | 0.50, 1.84 | 0.67 | 0.37, 1.20 | 0.50 | 0.29, 0.85 | 0.62 | 0.33, 1.17 | 0.002 | |
| BRAF Mutation Status | ||||||||||||||
| BRAF-wildtype | ||||||||||||||
| No. | 114 | 42 | 70 | 105 | 207 | 89 | ||||||||
| Age-adjusted | 1.00 | Referent | 0.81 | 0.57, 1.16 | 1.28 | 0.95, 1.73 | 0.93 | 0.71, 1.21 | 0.88 | 0.69, 1.11 | 0.89 | 0.66, 1.19 | 0.12 | |
| Multivariated | 1.00 | Referent | 0.82 | 0.57, 1.17 | 1.28 | 0.95, 1.73 | 0.93 | 0.71, 1.21 | 0.88 | 0.70, 1.12 | 0.91 | 0.67, 1.22 | 0.18 | 0.10 |
| BRAF-mutated | ||||||||||||||
| No. | 22 | 14 | 13 | 19 | 30 | 13 | ||||||||
| Age-adjusted | 1.00 | Referent | 1.47 | 0.75, 2.89 | 1.19 | 0.60, 2.37 | 0.87 | 0.47, 1.63 | 0.73 | 0.42, 1.28 | 0.76 | 0.37, 1.56 | 0.02 | |
| Multivariated | 1.00 | Referent | 1.48 | 0.75, 2.91 | 1.17 | 0.59, 2.34 | 0.88 | 0.47, 1.64 | 0.73 | 0.41, 1.28 | 0.77 | 0.38, 1.59 | 0.02 | |
| DNMT3B Expression Status | ||||||||||||||
| DNMT3B-negative | ||||||||||||||
| No. | 73 | 35 | 38 | 72 | 123 | 37 | ||||||||
| Age-adjusted | 1.00 | Referent | 1.10 | 0.73, 1.65 | 1.15 | 0.77, 1.70 | 1.17 | 0.84, 1.63 | 1.02 | 0.76, 1.37 | 0.96 | 0.63, 1.47 | 0.40 | |
| Multivariated | 1.00 | Referent | 1.11 | 0.74, 1.66 | 1.15 | 0.77, 1.71 | 1.19 | 0.85, 1.65 | 1.04 | 0.77, 1.41 | 1.01 | 0.66, 1.54 | 0.61 | 0.03 |
| DNMT3B-positive | ||||||||||||||
| No. | 17 | 5 | 8 | 5 | 16 | 5 | ||||||||
| Age-adjusted | 1.00 | Referent | 0.76 | 0.28, 2.07 | 0.99 | 0.42, 2.32 | 0.32 | 0.12, 0.87 | 0.50 | 0.25, 1.01 | 0.43 | 0.15, 1.23 | 0.01 | |
| Multivariated | 1.00 | Referent | 0.78 | 0.28, 2.12 | 1.00 | 0.43, 2.34 | 0.33 | 0.12, 0.90 | 0.52 | 0.26, 1.05 | 0.44 | 0.15, 1.25 | 0.01 | |
Abbreviations: CI, confidence interval; CIMP, CpG island methylator phenotype; DNMT3B, DNA methyltransferase 3B; HR, hazard ratio; MSI, microsatellite instability; MSS, microsatellite stable.
a All models were stratified by calendar year of the questionnaire cycle, age, and sex.
b Based on the linear trend test across the median values in each category. To test whether the duration of smoking cessation reduced the cancer risk compared with current smoking, trend tests and heterogeneity tests were performed on current and past smokers, excluding never smokers.
c Tests for heterogeneity (for a multivariate HR linear trend) showed significance of differential association of cessation with colorectal cancer risk by molecular subtypes (i.e., CIMP-low/negative vs. CIMP-high; MSS vs. MSI-high; BRAF-wildtype vs. BRAF-mutated; DNMT3B-negative vs. DNMT3B-positive).
d Models were adjusted for body mass index, family history of colorectal cancer in any first-degree relative, regular use of aspirin, physical activity level, alcohol consumption, total caloric intake, and red meat intake.
Duration of smoking cessation and colorectal cancer risk by molecular subtypes
Compared with current smokers, duration of smoking cessation was associated with a significantly reduced risk of CIMP-high colorectal cancer (Ptrend = 0.001). Compared with current smokers, multivariate hazard ratios for smoking cessation of 10–19, 20–39, and ≥40 years were 0.53 (95% confidence interval (CI): 0.29, 0.95), 0.52 (95% CI: 0.32, 0.85), and 0.50 (95% CI: 0.27, 0.94), respectively (Table 2). Approximately 50% lower risk of CIMP-high cancer among former smokers with long-term cessation (compared with current smokers) was similar to the risk of CIMP-high cancer among never smokers compared with current smokers (hazard ratio (HR) = 0.47; 95% CI: 0.31, 0.73; for never smokers compared with current smokers; HR = 2.08; 95% CI: 1.35, 3.20; for current smokers compared with never smokers). In contrast, smoking cessation was not significantly associated with CIMP-low/negative cancer risk (Ptrend = 0.25), and the association of smoking cessation with the cancer risk significantly differed by CIMP status (Pheterogeneity = 0.02).
Longer duration of smoking cessation was associated with a decrease in MSI-high cancer risk (Ptrend = 0.002), but was not significantly associated with microsatellite-stable cancer risk (Ptrend = 0.36; Pheterogeneity = 0.02) (Table 2). Longer duration of smoking cessation was associated with a decreased risk for DNMT3B-positive cancer (Ptrend = 0.01), but not with DNTM3B-negative cancer risk (Ptrend = 0.61; Pheterogeneity = 0.03) (Table 2). The association of smoking cessation with cancer risk did not significantly differ by BRAF mutation status (Pheterogeneity = 0.10).
Smoothing spline plots (Web Figure 1) show dose-response relation between the duration of smoking cessation and a decrease in the risk of CIMP-high, MSI-high, or DNMT3B-positive cancers. Web Table 4 shows the risk estimates for duration of smoking cessation compared with never smokers.
Smoking cessation and risk of combined molecular subtypes
Because CIMP-high is associated with MSI-high and DNMT3B-positive status in colorectal cancer (13–15, 18–20), we examined combined molecular features, to assess which molecular subtype risk was reduced by smoking cessation independent of other molecular features. This combined analysis was conducted using the molecular features which were significantly associated with smoking cessation in Table 2, and could confound each other. Compared with current smokers, the risk reduction associated with smoking cessation was apparent for CIMP-high cancers regardless of MSI status (Ptrend ≤ 0.02), and CIMP-high cancers regardless of DNMT3B status (Ptrend ≤ 0.02) (Table 3). In analysis using combined BRAF and CIMP status, the relation between smoking cessation and CIMP-high cancer risk was apparent irrespective of BRAF mutation status (data not shown). The findings suggest that risk reduction associated with smoking cessation might be present primarily on CIMP-high cancer.
Table 3.
Duration of Smoking Cessation and Colorectal Cancer Risk by Combined Molecular Subtypesa in the Nurses' Health Study (1980–2008) and the Health Professionals Follow-up Study (1986–2008)
| Current Smoker |
Cessation for 1–4 Years |
Cessation for 5–9 Years |
Cessation for ≥10 Years |
Ptrendb | |||||
|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | ||
| CIMP/MSI Subtyping | |||||||||
| CIMP-low/negative | |||||||||
| MSS | |||||||||
| No. | 94 | 37 | 60 | 346 | |||||
| Age-adjusted | 1.00 | Referent | 0.88 | 0.60, 1.29 | 1.37 | 0.99, 1.90 | 1.02 | 0.80, 1.29 | 0.66 |
| Multivariatec | 1.00 | Referent | 0.88 | 0.60, 1.29 | 1.37 | 0.99, 1.90 | 1.03 | 0.81, 1.30 | 0.81 |
| MSI-high | |||||||||
| No. | 6 | 3 | 2 | 13 | |||||
| Age-adjusted | 1.00 | Referent | 1.13 | 0.28, 4.59 | 0.71 | 0.14, 3.56 | 0.57 | 0.21, 1.56 | 0.08 |
| Multivariatec | 1.00 | Referent | 1.15 | 0.28, 4.68 | 0.70 | 0.14, 3.52 | 0.58 | 0.21, 1.58 | 0.08 |
| CIMP-high | |||||||||
| MSS | |||||||||
| No. | 11 | 2 | 3 | 19 | |||||
| Age-adjusted | 1.00 | Referent | 0.45 | 0.10, 2.05 | 0.52 | 0.14, 1.88 | 0.37 | 0.17, 0.80 | 0.02 |
| Multivariatec | 1.00 | Referent | 0.47 | 0.10, 2.14 | 0.54 | 0.15, 1.94 | 0.38 | 0.18, 0.83 | 0.02 |
| MSI-high | |||||||||
| No. | 20 | 13 | 12 | 50 | |||||
| Age-adjusted | 1.00 | Referent | 1.40 | 0.69, 2.82 | 1.08 | 0.52, 2.22 | 0.56 | 0.33, 0.96 | 0.002 |
| Multivariatec | 1.00 | Referent | 1.43 | 0.71, 2.88 | 1.07 | 0.52, 2.20 | 0.57 | 0.33, 0.97 | 0.003 |
| CIMP/DNMT3B Subtyping | |||||||||
| CIMP-low/negative | |||||||||
| DNMT3B-negative | |||||||||
| No. | 56 | 27 | 32 | 206 | |||||
| Age-adjusted | 1.00 | Referent | 1.10 | 0.69, 1.75 | 1.27 | 0.82, 1.97 | 1.25 | 0.92, 1.69 | 0.28 |
| Multivariatec | 1.00 | Referent | 1.10 | 0.69, 1.75 | 1.28 | 0.82, 1.98 | 1.27 | 0.93, 1.73 | 0.20 |
| DNMT3B-positive | |||||||||
| No. | 10 | 2 | 6 | 19 | |||||
| Age-adjusted | 1.00 | Referent | 0.52 | 0.11, 2.41 | 1.33 | 0.48, 3.70 | 0.55 | 0.25, 1.22 | 0.06 |
| Multivariatec | 1.00 | Referent | 0.53 | 0.11, 2.44 | 1.32 | 0.47, 3.69 | 0.56 | 0.25, 1.25 | 0.07 |
| CIMP-high | |||||||||
| DNMT3B-negative | |||||||||
| No. | 14 | 6 | 3 | 21 | |||||
| Age-adjusted | 1.00 | Referent | 0.95 | 0.36, 2.49 | 0.40 | 0.12, 1.41 | 0.41 | 0.20, 0.82 | 0.02 |
| Multivariatec | 1.00 | Referent | 0.98 | 0.37, 2.57 | 0.41 | 0.12, 1.42 | 0.42 | 0.21, 0.85 | 0.02 |
| DNMT3B-positive | |||||||||
| No. | 7 | 3 | 2 | 7 | |||||
| Age-adjusted | 1.00 | Referent | 1.07 | 0.27, 4.17 | 0.55 | 0.11, 2.67 | 0.28 | 0.10, 0.81 | 0.01 |
| Multivariatec | 1.00 | Referent | 1.12 | 0.29, 4.38 | 0.56 | 0.12, 2.73 | 0.29 | 0.10, 0.85 | 0.01 |
Abbreviations: CI, confidence interval; CIMP, CpG island methylator phenotype; DNMT3B, DNA methyltransferase 3B; HR, hazard ratio; MSI, microsatellite instability; MSS, microsatellite stable.
a All models were stratified by calendar year of the questionnaire cycle, age, and sex.
b Based on the linear trend test by using the median value of each category. To test whether the duration of smoking cessation reduced the cancer risk compared with current smoking, trend test and heterogeneity tests were performed on current and past smokers, excluding never smokers.
c Models were adjusted for body mass index, family history of colorectal cancer in any first-degree relative, regular use of aspirin, physical activity level, alcohol consumption, total caloric intake, and red meat intake.
Smoking cessation and tumor molecular subtypes in strata of cumulative pack-years smoked
We examined the association of smoking cessation with the risk for specific cancer subtypes in strata of cumulative pack-years smoked, in an attempt to control for confounding by cumulative pack-years. Among current/former smokers with 20 or more pack-years, longer duration of cessation was associated with significantly lower risk for CIMP-high cancer (Ptrend = 0.02), and DNMT3B-positive cancer (Ptrend = 0.04) (Web Table 5). The association of smoking cessation with colorectal cancer risk differed significantly by CIMP status (Pheterogeneity = 0.02) and DNMT3B expression status (Pheterogeneity = 0.03). Statistical power was limited in the stratum of <20 pack-years.
Other smoking variables and colorectal cancer risk by molecular subtypes
We examined the association between other smoking indicators (including cumulative pack-years, pack-years smoked before age 30, and age at start of smoking) and colorectal cancer risk by molecular subtypes separately women and men (Web Tables 6 and 7), and among the combined cohorts (Tables 4 and 5). The category of never smokers was used as the referent group because we attempted to see whether smoking increased the risk of specific cancer subtype. Compared with never smokers, smoking of 40 or more pack-years was associated with higher risks of CIMP-high cancer (multivariate HR = 2.12; 95% CI: 1.48, 3.03; Ptrend < 0.0001), MSI-high cancer (multivariate HR = 2.27; 95% CI: 1.56, 3.31; Ptrend < 0.0001), and BRAF-mutated cancer (multivariate HR = 2.00; 95% CI: 1.37, 2.92; Ptrend = 0.0001) (Table 4). In contrast, cumulative pack-years were not significantly associated with the risk of CIMP-low/negative cancer, microsatellite-stable cancer, or BRAF-wildtype cancer (Ptrend ≥ 0.10). The association of cumulative pack-years with the cancer risk differed by CIMP status (Pheterogeneity = 0.001), MSI status (Pheterogeneity = 0.0003), and BRAF mutation status (Pheterogeneity = 0.01). The relation between cumulative pack-years and cancer risk did not significantly differ by DNMT3B status (Pheterogeneity = 0.83).
Table 4.
Smoking Status, Cumulative Pack-years of Smoking, and Incident Colorectal Cancer Risk by Molecular Subtypesa in the Nurses' Health Study (1980–2008) and the Health Professionals Follow-up Study (1986–2008)
| Smoking Status |
Cumulative Pack-years of Smoking |
|||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Never (n = 1,383,154 person-years) |
Former (n = 1,278,369 person-years) |
Current (n = 439,508 person-years) |
Ptrendb | Pheterogeneityc | 1–19 (n = 844,894 person-years) |
20–39 (n = 511,272 person-years) |
≥40 (n = 338,416 person-years) |
Ptrendb | Pheterogeneityc | |||||||
| HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | |||||
| All colorectal cancer | ||||||||||||||||
| No. | 490 | 631 | 139 | 300 | 226 | 216 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.23 | 1.09, 1.38 | 1.23 | 1.02, 1.49 | 0.001 | 1.09 | 0.94, 1.26 | 1.22 | 1.04, 1.43 | 1.35 | 1.15, 1.59 | <0.0001 | ||
| Multivariated | 1.00 | Referent | 1.18 | 1.05, 1.34 | 1.17 | 0.96, 1.43 | 0.02 | 1.06 | 0.91, 1.23 | 1.17 | 0.99, 1.38 | 1.28 | 1.08, 1.51 | 0.002 | ||
| CIMP status | 0.04 | 0.001 | ||||||||||||||
| CIMP-low/negative | ||||||||||||||||
| No. | 377 | 485 | 103 | 244 | 178 | 148 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.21 | 1.06, 1.39 | 1.17 | 0.94, 1.47 | 0.02 | 1.15 | 0.98, 1.35 | 1.20 | 1.00, 1.44 | 1.20 | 0.99, 1.46 | 0.04 | ||
| Multivariated | 1.00 | Referent | 1.17 | 1.02, 1.35 | 1.12 | 0.89, 1.41 | 0.07 | 1.12 | 0.95, 1.32 | 1.16 | 0.97, 1.39 | 1.14 | 0.94, 1.39 | 0.15 | ||
| CIMP-high | ||||||||||||||||
| No. | 71 | 103 | 31 | 34 | 36 | 56 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.34 | 0.99, 1.81 | 2.19 | 1.43, 3.37 | 0.001 | 0.87 | 0.58, 1.31 | 1.37 | 0.91, 2.05 | 2.23 | 1.57, 3.18 | <0.0001 | ||
| Multivariated | 1.00 | Referent | 1.30 | 0.95, 1.76 | 2.08 | 1.35, 3.20 | 0.002 | 0.86 | 0.57, 1.29 | 1.31 | 0.87, 1.96 | 2.12 | 1.48, 3.03 | <0.0001 | ||
| MSI status | 0.03 | 0.0003 | ||||||||||||||
| MSS | ||||||||||||||||
| No. | 400 | 504 | 108 | 254 | 175 | 159 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.17 | 1.02, 1.34 | 1.19 | 0.96, 1.48 | 0.02 | 1.12 | 0.95, 1.31 | 1.10 | 0.92, 1.32 | 1.21 | 1.00, 1.45 | 0.06 | ||
| Multivariated | 1.00 | Referent | 1.13 | 0.99, 1.30 | 1.14 | 0.91, 1.42 | 0.09 | 1.09 | 0.93, 1.28 | 1.06 | 0.89, 1.28 | 1.15 | 0.95, 1.39 | 0.21 | ||
| MSI-high | ||||||||||||||||
| No. | 63 | 98 | 27 | 34 | 37 | 50 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.46 | 1.06, 2.01 | 2.16 | 1.36, 3.41 | 0.001 | 0.98 | 0.65, 1.49 | 1.60 | 1.06, 2.41 | 2.36 | 1.62, 3.44 | <0.0001 | ||
| Multivariated | 1.00 | Referent | 1.42 | 1.03, 1.95 | 2.05 | 1.29, 3.26 | 0.002 | 0.96 | 0.63, 1.47 | 1.52 | 1.01, 2.30 | 2.27 | 1.56, 3.31 | <0.0001 | ||
| BRAF mutation status | 0.63 | 0.01 | ||||||||||||||
| BRAF-wildtype | ||||||||||||||||
| No. | 404 | 522 | 114 | 261 | 187 | 164 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.20 | 1.05, 1.36 | 1.28 | 1.03, 1.58 | 0.003 | 1.14 | 0.97, 1.33 | 1.16 | 0.97, 1.38 | 1.24 | 1.03, 1.49 | 0.02 | ||
| Multivariated | 1.00 | Referent | 1.16 | 1.01, 1.32 | 1.22 | 0.98, 1.52 | 0.02 | 1.11 | 0.95, 1.30 | 1.11 | 0.93, 1.33 | 1.18 | 0.98, 1.43 | 0.10 | ||
| BRAF-mutated | ||||||||||||||||
| No. | 67 | 89 | 22 | 31 | 28 | 48 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.28 | 0.93, 1.76 | 1.43 | 0.87, 2.33 | 0.08 | 0.83 | 0.54, 1.27 | 1.19 | 0.76, 1.85 | 2.08 | 1.43, 3.03 | <0.0001 | ||
| Multivariated | 1.00 | Referent | 1.24 | 0.90, 1.71 | 1.38 | 0.84, 2.25 | 0.13 | 0.81 | 0.53, 1.25 | 1.15 | 0.73, 1.79 | 2.00 | 1.37, 2.92 | 0.0001 | ||
| DNMT3B expression status | 0.38 | 0.83 | ||||||||||||||
| DNMT3B-negative | ||||||||||||||||
| No. | 238 | 309 | 73 | 160 | 104 | 103 | ||||||||||
| Age-adjusted | 1.00 | Referent | 1.26 | 1.06, 1.49 | 1.16 | 0.89, 1.51 | 0.05 | 1.21 | 0.99, 1.48 | 1.11 | 0.88, 1.40 | 1.29 | 1.02, 1.63 | 0.07 | ||
| Multivariated | 1.00 | Referent | 1.22 | 1.02, 1.45 | 1.10 | 0.84, 1.45 | 0.13 | 1.19 | 0.97, 1.46 | 1.08 | 0.85, 1.36 | 1.22 | 0.96, 1.55 | 0.19 | ||
| DNMT3B-positive | ||||||||||||||||
| No. | 52 | 39 | 17 | 15 | 16 | 23 | ||||||||||
| Age-adjusted | 1.00 | Referent | 0.69 | 0.46, 1.05 | 1.31 | 0.75, 2.28 | 0.95 | 0.51 | 0.28, 0.90 | 0.79 | 0.45, 1.39 | 1.26 | 0.77, 2.08 | 0.28 | ||
| Multivariated | 1.00 | Referent | 0.67 | 0.44, 1.02 | 1.22 | 0.69, 2.13 | 0.76 | 0.50 | 0.28, 0.89 | 0.75 | 0.42, 1.32 | 1.18 | 0.71, 1.95 | 0.42 | ||
Abbreviations: CI, confidence interval; CIMP, CpG island methylator phenotype; DNMT3B, DNA methyltransferase 3B; HR, hazard ratio; MSI, microsatellite instability; MSS, microsatellite stable.
a All models were stratified by calendar year of the questionnaire cycle, age, and sex.
b Based on the linear trend test by using the median value of each category.
c Tests for heterogeneity (for a multivariate HR linear trend) of the associations of smoking with one molecular subtype versus the other molecular subtype (i.e., CIMP-low/negative vs. CIMP-high; MSS vs. MSI-high; BRAF-wildtype vs. BRAF-mutated; DNMT3B-negative vs. DNMT3B-positive).
d Models were adjusted for body mass index, family history of colorectal cancer in any first-degree relative, regular use of aspirin, physical activity level, alcohol consumption, total caloric intake, and red meat intake.
Table 5.
Pack-years of Smoking Before Age 30 Years, Age at Start of Smoking, and Incident Colorectal Cancer Risk by Molecular Subtypesa in the Nurses' Health Study (1980–2008) and the Health Professionals Follow-up Study (1986–2008)
| Never Smoker (n = 1,383,154 person-years) |
Pack-years Smoked Before Age 30 Years |
Age at Start of Smoking, years |
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1–9 (n = 1,085,062 person-years) |
≥10 (n = 560,470 person-years) |
Ptrendb | Pheterogeneityc | ≥20 (n = 744,382 person-years) |
<20 (n = 976,780 person-years) |
Ptrendb | Pheterogeneityc | |||||||
| HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | |||||
| All colorectal cancer | ||||||||||||||
| No. | 490 | 414 | 300 | 347 | 402 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.09 | 0.96, 1.25 | 1.31 | 1.14, 1.52 | 0.0002 | 1.16 | 1.01, 1.33 | 1.21 | 1.06, 1.38 | 0.004 | ||
| Multivariated | 1.00 | Referent | 1.05 | 0.92, 1.20 | 1.25 | 1.08, 1.45 | 0.003 | 1.12 | 0.97, 1.29 | 1.16 | 1.01, 1.33 | 0.03 | ||
| CIMP status | 0.09 | 0.17 | ||||||||||||
| CIMP-low/negative | ||||||||||||||
| No. | 377 | 322 | 222 | 274 | 301 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.16 | 1.00, 1.35 | 1.13 | 0.95, 1.34 | 0.09 | 1.19 | 1.01, 1.39 | 1.16 | 0.99, 1.35 | 0.05 | ||
| Multivariated | 1.00 | Referent | 1.12 | 0.96, 1.31 | 1.07 | 0.90, 1.28 | 0.30 | 1.15 | 0.98, 1.35 | 1.11 | 0.95, 1.30 | 0.18 | ||
| CIMP-high | ||||||||||||||
| No. | 71 | 74 | 50 | 57 | 71 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.33 | 0.96, 1.85 | 1.62 | 1.11, 2.35 | 0.01 | 1.27 | 0.89, 1.80 | 1.50 | 1.08, 2.09 | 0.02 | ||
| Multivariated | 1.00 | Referent | 1.29 | 0.93, 1.79 | 1.54 | 1.06, 2.25 | 0.02 | 1.23 | 0.87, 1.75 | 1.44 | 1.03, 2.01 | 0.03 | ||
| MSI status | 0.05 | 0.21 | ||||||||||||
| MSS | ||||||||||||||
| No. | 400 | 323 | 240 | 275 | 318 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.10 | 0.95, 1.28 | 1.13 | 0.96, 1.33 | 0.10 | 1.11 | 0.95, 1.29 | 1.15 | 0.99, 1.33 | 0.07 | ||
| Multivariated | 1.00 | Referent | 1.07 | 0.92, 1.24 | 1.07 | 0.91, 1.27 | 0.32 | 1.08 | 0.92, 1.26 | 1.10 | 0.94, 1.28 | 0.23 | ||
| MSI-high | ||||||||||||||
| No. | 63 | 75 | 45 | 63 | 60 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.58 | 1.12, 2.21 | 1.61 | 1.09, 2.40 | 0.01 | 1.62 | 1.14, 2.30 | 1.45 | 1.01, 2.06 | 0.03 | ||
| Multivariated | 1.00 | Referent | 1.53 | 1.09, 2.15 | 1.55 | 1.04, 2.31 | 0.01 | 1.57 | 1.11, 2.24 | 1.39 | 0.97, 1.99 | 0.06 | ||
| BRAF mutation status | 0.39 | 0.73 | ||||||||||||
| BRAF-wildtype | ||||||||||||||
| No. | 404 | 334 | 253 | 289 | 328 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.14 | 0.99, 1.32 | 1.16 | 0.99, 1.37 | 0.03 | 1.15 | 0.99, 1.34 | 1.17 | 1.01, 1.36 | 0.03 | ||
| Multivariated | 1.00 | Referent | 1.11 | 0.95, 1.28 | 1.11 | 0.94, 1.31 | 0.15 | 1.12 | 0.96, 1.31 | 1.12 | 0.97, 1.31 | 0.12 | ||
| BRAF-mutated | ||||||||||||||
| No. | 67 | 68 | 37 | 53 | 56 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.25 | 0.89, 1.75 | 1.35 | 0.89, 2.04 | 0.08 | 1.30 | 0.90, 1.86 | 1.24 | 0.87, 1.78 | 0.21 | ||
| Multivariated | 1.00 | Referent | 1.21 | 0.86, 1.71 | 1.29 | 0.85, 1.96 | 0.13 | 1.26 | 0.88, 1.82 | 1.20 | 0.83, 1.72 | 0.30 | ||
| DNMT3B expression status | 0.41 | 0.07 | ||||||||||||
| DNMT3B-negative | ||||||||||||||
| No. | 238 | 222 | 131 | 179 | 191 | |||||||||
| Age-adjusted | 1.00 | Referent | 1.23 | 1.02, 1.48 | 1.09 | 0.87, 1.35 | 0.25 | 1.21 | 1.00, 1.48 | 1.17 | 0.97, 1.42 | 0.08 | ||
| Multivariated | 1.00 | Referent | 1.20 | 0.99, 1.45 | 1.04 | 0.83, 1.30 | 0.54 | 1.18 | 0.97, 1.45 | 1.13 | 0.92, 1.37 | 0.21 | ||
| DNMT3B-positive | ||||||||||||||
| No. | 52 | 27 | 24 | 27 | 28 | |||||||||
| Age-adjusted | 1.00 | Referent | 0.67 | 0.42, 1.08 | 0.93 | 0.56, 1.53 | 0.72 | 0.82 | 0.52, 1.32 | 0.76 | 0.48, 1.21 | 0.23 | ||
| Multivariated | 1.00 | Referent | 0.65 | 0.41, 1.04 | 0.87 | 0.53, 1.45 | 0.53 | 0.80 | 0.50, 1.28 | 0.72 | 0.45, 1.15 | 0.15 | ||
Abbreviations: CI, confidence interval; CIMP, CpG island methylator phenotype; DNMT3B, DNA methyltransferase 3B; HR, hazard ratio; MSI, microsatellite instability; MSS, microsatellite stable.
a All models were stratified by calendar year of the questionnaire cycle, age, and sex.
b Based on the linear trend test by using the median value of each category.
c Tests for heterogeneity (for a multivariate HR linear trend) of the associations of smoking with one molecular subtype versus the other molecular subtype (i.e., CIMP-low/negative vs. CIMP-high; MSS vs. MSI-high; BRAF-wildtype vs. BRAF-mutated; DNMT3B-negative vs. DNMT3B-positive).
d Models were adjusted for body mass index, family history of colorectal cancer in any first-degree relative, regular use of aspirin, physical activity level, alcohol consumption, total caloric intake, and red meat intake.
Because CIMP-high is associated with both MSI-high and BRAF mutation in colorectal cancer (13–15, 18–20), we examined the relation between cumulative pack-years and cancer risk by combined molecular subtyping (Table 6). Combined molecular analysis was conducted using the molecular features which were significantly associated with cumulative pack-years in Table 4, and could confound each other. In CIMP/MSI subtyping, compared with never smokers, 40 or more pack-years smoked were associated with a higher risk for CIMP-high/MSI-high cancer (multivariate HR = 2.75; 95% CI: 1.78, 4.26; Ptrend < 0.0001), but not with the other 3 CIMP/MSI subtypes (Ptrend ≥ 0.15).
Table 6.
Cumulative Pack-years of Smoking and Incident Colorectal Cancer Risk by Combined Molecular Subtypesa in the Nurses' Health Study (1980–2008) and the Health Professionals Follow-up Study (1986–2008)
| Never Smoker |
Cumulative Pack-years of Smoking |
Ptrendb | |||||||
|---|---|---|---|---|---|---|---|---|---|
| 1–19 |
20–39 |
≥40 |
|||||||
| HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | ||
| CIMP/MSI Subtyping | |||||||||
| CIMP-low/negative | |||||||||
| MSS | |||||||||
| No. | 340 | 227 | 161 | 137 | |||||
| Age-adjusted | 1.00 | Referent | 1.18 | 0.99, 1.40 | 1.20 | 0.99, 1.45 | 1.22 | 1.00, 1.49 | 0.05 |
| Multivariatec | 1.00 | Referent | 1.15 | 0.97, 1.37 | 1.16 | 0.96, 1.41 | 1.16 | 0.95, 1.43 | 0.15 |
| MSI-high | |||||||||
| No. | 19 | 10 | 8 | 6 | |||||
| Age-adjusted | 1.00 | Referent | 0.98 | 0.45, 2.11 | 1.14 | 0.49, 2.62 | 1.14 | 0.45, 2.88 | 0.72 |
| Multivariatec | 1.00 | Referent | 0.96 | 0.44, 2.07 | 1.08 | 0.47, 2.48 | 1.08 | 0.43, 2.75 | 0.82 |
| CIMP-high | |||||||||
| MSS | |||||||||
| No. | 29 | 11 | 7 | 14 | |||||
| Age-adjusted | 1.00 | Referent | 0.69 | 0.34, 1.39 | 0.60 | 0.26, 1.38 | 1.40 | 0.74, 2.67 | 0.42 |
| Multivariatec | 1.00 | Referent | 0.68 | 0.34, 1.37 | 0.58 | 0.25, 1.33 | 1.30 | 0.68, 2.48 | 0.56 |
| MSI-high | |||||||||
| No. | 41 | 22 | 28 | 42 | |||||
| Age-adjusted | 1.00 | Referent | 0.98 | 0.58, 1.65 | 1.90 | 1.17, 3.08 | 2.85 | 1.85, 4.40 | <0.0001 |
| Multivariatec | 1.00 | Referent | 0.97 | 0.57, 1.63 | 1.82 | 1.12, 2.96 | 2.75 | 1.78, 4.26 | <0.0001 |
| CIMP/BRAF Subtyping | |||||||||
| CIMP-low/negative | |||||||||
| BRAF-wildtype | |||||||||
| No. | 341 | 232 | 164 | 136 | |||||
| Age-adjusted | 1.00 | Referent | 1.20 | 1.02, 1.42 | 1.21 | 1.00, 1.46 | 1.23 | 1.00, 1.50 | 0.04 |
| Multivariatec | 1.00 | Referent | 1.17 | 0.99, 1.39 | 1.17 | 0.96, 1.41 | 1.17 | 0.95, 1.43 | 0.15 |
| BRAF-mutated | |||||||||
| No. | 22 | 8 | 6 | 10 | |||||
| Age-adjusted | 1.00 | Referent | 0.65 | 0.29, 1.48 | 0.81 | 0.33, 2.02 | 1.30 | 0.60, 2.81 | 0.49 |
| Multivariatec | 1.00 | Referent | 0.65 | 0.29, 1.47 | 0.81 | 0.32, 2.00 | 1.27 | 0.59, 2.75 | 0.53 |
| CIMP-high | |||||||||
| BRAF-wildtype | |||||||||
| No. | 28 | 12 | 13 | 20 | |||||
| Age-adjusted | 1.00 | Referent | 0.81 | 0.41, 1.60 | 1.14 | 0.59, 2.22 | 1.92 | 1.07, 3.42 | 0.02 |
| Multivariatec | 1.00 | Referent | 0.80 | 0.41, 1.59 | 1.10 | 0.56, 2.14 | 1.83 | 1.02, 3.27 | 0.03 |
| BRAF-mutated | |||||||||
| No. | 43 | 21 | 22 | 36 | |||||
| Age-adjusted | 1.00 | Referent | 0.87 | 0.52, 1.47 | 1.45 | 0.86, 2.43 | 2.44 | 1.56, 3.81 | <0.0001 |
| Multivariatec | 1.00 | Referent | 0.86 | 0.51, 1.45 | 1.39 | 0.82, 2.33 | 2.32 | 1.48, 3.63 | <0.0001 |
| MSI/BRAF Subtyping | |||||||||
| MSS | |||||||||
| BRAF-wildtype | |||||||||
| No. | 360 | 239 | 165 | 142 | |||||
| Age-adjusted | 1.00 | Referent | 1.16 | 0.99, 1.37 | 1.14 | 0.95, 1.38 | 1.20 | 0.99, 1.46 | 0.08 |
| Multivariatec | 1.00 | Referent | 1.13 | 0.96, 1.34 | 1.10 | 0.91, 1.33 | 1.14 | 0.93, 1.40 | 0.24 |
| BRAF-mutated | |||||||||
| No. | 36 | 14 | 9 | 17 | |||||
| Age-adjusted | 1.00 | Referent | 0.70 | 0.37, 1.30 | 0.70 | 0.34, 1.46 | 1.39 | 0.77, 2.50 | 0.35 |
| Multivariatec | 1.00 | Referent | 0.69 | 0.37, 1.28 | 0.68 | 0.33, 1.42 | 1.33 | 0.74, 2.40 | 0.42 |
| MSI-high | |||||||||
| BRAF-wildtype | |||||||||
| No. | 32 | 18 | 18 | 19 | |||||
| Age-adjusted | 1.00 | Referent | 1.04 | 0.58, 1.86 | 1.44 | 0.80, 2.57 | 1.86 | 1.04, 3.30 | 0.02 |
| Multivariatec | 1.00 | Referent | 1.03 | 0.57, 1.83 | 1.36 | 0.76, 2.44 | 1.79 | 1.01, 3.19 | 0.03 |
| BRAF-mutated | |||||||||
| No. | 30 | 16 | 19 | 31 | |||||
| Age-adjusted | 1.00 | Referent | 0.96 | 0.52, 1.76 | 1.82 | 1.02, 3.25 | 2.94 | 1.77, 4.87 | <0.0001 |
| Multivariatec | 1.00 | Referent | 0.94 | 0.51, 1.72 | 1.74 | 0.97, 3.11 | 2.81 | 1.69, 4.68 | <0.0001 |
Abbreviations: CI, confidence interval; CIMP, CpG island methylator phenotype; HR, hazard ratio; MSI, microsatellite instability; MSS, microsatellite stable.
a All models were stratified by calendar year of the questionnaire cycle, age, and sex.
b Based on the linear trend test by using the median value of each category.
c Models were adjusted for body mass index, family history of colorectal cancer in parent or sibling, regular use of aspirin, physical activity level, alcohol consumption, total caloric intake, and red meat intake.
In CIMP/BRAF subtyping, cumulative pack-years was significantly associated with a higher risk for CIMP-high cancer regardless of BRAF status (Ptrend ≤ 0.03), but not with CIMP-low/negative cancer risk (Ptrend ≥ 0.15). In MSI/BRAF subtyping, cumulative pack-years smoked was significantly associated with a higher risk for MSI-high cancer regardless of BRAF status (Ptrend ≤ 0.03), but not with microsatellite-stable cancer risk (Ptrend ≥ 0.24).
DISCUSSION
We conducted this unique analysis to prospectively examine the relation between duration of smoking cessation and colorectal cancer risk by molecularly-defined subtypes. We utilized 2 US nationwide prospective cohort studies with available lifestyle information, including smoking status at multiple time points during follow-up, as well as tumor molecular data. We showed that, compared with current smokers, duration of smoking cessation was associated with a decreased risk of CIMP-high colorectal cancer (but not with the risk of CIMP-low/negative cancer). There might be a plateau of the effect of cessation duration beyond 10 years, as risk estimates were similar beyond 10 years of cessation (multivariate HRs of 0.50–0.53, compared with current smoking). Our data suggest that smoking cessation might be effective in preventing specific molecular subtypes of colorectal cancer. Our data also underscore the importance of cessation in as early as possible, because, after 10 years of cessation, the CIMP-high cancer risk appeared to be almost similar to never smokers.
We observed a significant trend of risk reduction for proximal colon cancer but not for distal colorectal cancer; this anatomical difference in cancer risk might be due to higher prevalence of CIMP-high in proximal colon cancers (43, 44). Considering the “colorectal continuum” hypothesis (43, 44), the effect of smoking and its cessation might continuously change along the bowel subsites. Additional studies are necessary to examine the effect of smoking on carcinogenesis in detailed colorectal subsites.
Molecular features of colorectal cancer such as CIMP-high, MSI-high, BRAF mutations and DNMT3B expression are known to be interrelated (13–15, 18–25). Smoking cessation was associated with lower risks of MSI-high and DNMT3B-positive colorectal cancers, and these associations appeared to be driven by CIMP-high cancers enriched in these molecular subtypes. The well-documented association between smoking and BRAF-mutated cancer (26–28, 35) might be due to enrichment of the CIMP-high subtype in the BRAF-mutated cancers. Therefore, our current analysis emphasizes the importance of considering influence of multiple molecular features on epidemiologic associations (so-called “molecular confounding” (51)).
The relation between smoking and a specific cancer epigenotype is plausible. Cigarette smoke contains over 4,000 toxic chemicals, many of which can induce DNA damage (52). Evidence suggests that cigarette smoking and nicotine can induce DNA methylation (36–38, 53, 54). Changes in DNA methylation could be observed within 9 months after cigarette smoke condensate was applied to human epithelial cells (37). Additional studies are needed to elucidate the exact mechanisms of effects of smoking on epigenetic alterations.
Our present study represents MPE research (10, 11, 55). MPE is based on the unique tumor principle (51, 56) and etiologic heterogeneity according to molecular subtypes (e.g., CIMP-high vs. non-CIMP-high). Thus, MPE differs from conventional molecular epidemiology which typically deals with “colon cancer” as a single entity (57–59). MPE analysis can not only refine risk estimates for specific subtypes of cancer, but also provide evidence for causality and insights into pathogenic mechanisms (10, 11, 51, 60–65). We previously discussed how MPE research can provide evidence for causality in depth (10, 11). For example, although traditional epidemiology research has linked smoking to colorectal cancer, effect size for overall colorectal cancer risk by smoking has been modest (hazard ratio of about 1.2–1.3). In contrast, MPE research can find a consistent link between smoking and CIMP-high colorectal cancers with an accurate and substantial effect estimate for the CIMP-high subtype (hazard ratio of almost 2). This consistent link can provide further evidence for causality. The MPE approach enabled us to find a possible preventive effect of smoking cessation on the development of the specific epigenotype (i.e., CIMP-high) of colorectal cancer.
Analyses of etiologic factors and molecular variation are important in epidemiologic research (66–68). One case-cohort study reported that duration of smoking cessation at study baseline was not associated with v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) oncogene mutation status, compared with never smokers (7, 69). To our knowledge, no previous study has prospectively examined the relation between duration of smoking cessation and colorectal cancer risk by tumor epigenetic features. Previous studies (26–31, 33, 34, 70–72) have shown positive associations between smoking and either MSI-high, CIMP-high, or BRAF-mutated cancer subtype. The case-control study by Samowitz et al. (27) attempted to subtype cancers using combined molecular subtypes, and reported that CIMP-high and BRAF-mutated cancer subtypes might be attributable to smoking. Caveats of that study (27) include the case-control design, and the use of methylation-specific polymerase chain reaction and the classic CIMP panel (12), which might not be as specific as the newer Weisenberger CIMP panel (18). The issue of tumor misclassification could be even more important when combined molecular subtyping is attempted. By using our large prospective cohort studies of men and women, and a validated MethyLight CIMP assay (20, 48), we were able to demonstrate that smoking was associated specifically with CIMP-high cancer risk and that the association between smoking and BRAF-mutated colorectal cancer appeared to be mediated by the well-known association between BRAF mutation and CIMP-high (18, 20, 27). Our data on smoking cessation also support the hypothesis that CIMP-high is the molecular subtype caused by smoking.
Our findings could have clinical implications in terms of personalized screening and prevention. With the emergence of assays that detect markers of DNA methylation in stool, specific screening tests might become available that could be targeted to smokers, as a particularly high-risk group for CIMP-high cancer. In addition, for other specific high-risk groups (e.g., older women) who are known to have greater susceptibility for CIMP-high cancer, smoking abstinence or cessation could prove to be a high-priority prevention strategy. Research on CIMP has been progressing (14, 16, 73–83), and besides smoking cessation, there might be effective prevention strategy for this unique cancer pathway.
There are several key strengths in our study. Firstly, the prospective design minimizes recall bias. Secondly, because we prospectively collected updated information on smoking every 2 years, we could assess the risk reduction by duration of smoking cessation as well as multiple smoking-related variables more precisely. Thirdly, we collected updated data on the known and many suspected risk factors for colorectal cancer from health professionals, who tend to report with high accuracy on medication use, allowing us to effectively control for potentially confounding variables. Finally, our tumor molecular analysis data enabled us to conduct integrative MPE research, which resulted in unique evidence for the association of duration of smoking cessation with a specific epigenotype of colorectal cancer.
Limitations of our study include the possibility of residual confounding including birth cohort effect, informative censoring and, in particular, a confounding effect of pack-years on duration of cessation. To address the issue of pack-years smoked, we performed analysis stratified by cumulative pack-years. We could not obtain tumor paraffin blocks from all of the colorectal cancer cases. However, baseline features of participants without tumor analysis data did not differ materially from those with tumor analysis data. Our cohort represents a selected population, consisting of all health professionals, to maintain high compliance of questionnaire returns. Most of the participants are Caucasians. Therefore, the association of smoking cessation with cellular epigenetic instability in other occupational and other ethnic groups remains to be investigated. Results of sex-specific analysis need to be interpreted cautiously because of our limited statistical power in each sex stratum.
In summary, this prospective study suggests that smoking cessation could reduce the risk of the specific epigenotype, CIMP-high colorectal cancer. Our results provide not only insight into the colorectal carcinogenic mechanisms, but also yield further scientific support to the recommendation of smoking avoidance and cessation for the promotion of public health.
Supplementary Material
ACKNOWLEDGMENTS
Author affiliations: Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston, Massachusetts (Reiko Nishihara, Teppei Morikawa, Aya Kuchiba, Paul Lochhead, Mai Yamauchi, Xiaoyun Liao, Yu Imamura, Katsuhiko Nosho, Kaori Shima, Zhi Rong Qian, Charles S. Fuchs, Shuji Ogino); Department of Social and Behavioral Sciences, Harvard School of Public Health, Boston, Massachusetts (Reiko Nishihara, Ichiro Kawachi); Department of Nutrition, Harvard School of Public Health, Boston, Massachusetts (Reiko Nishihara, Aya Kuchiba, Edward Giovannucci); Division of Health Sciences, Osaka University Graduate School of Medicine, Osaka, Japan (Reiko Nishihara); Channing Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts (Charles S. Fuchs, Andrew T. Chan, Edward Giovannucci); Division of Gastroenterology, Massachusetts General Hospital, Boston, Massachusetts (Andrew T. Chan); Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts (Edward Giovannucci, Shuji Ogino); and Department of Pathology, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts (Shuji Ogino).
This work was supported by US National Institutes of Health grants (P01 CA87969; P01 CA55075; 1UM1 CA167552; P50 CA127003 to Charles S. Fuchs; R01 CA137178 to Andrew T. Chan; and R01 CA151993 to Shuji Ogino); and by grants from the Bennett Family Fund and the Entertainment Industry Foundation through National Colorectal Cancer Research Alliance. Andrew T. Chan is a Damon Runyon Clinical Investigator. Paul Lochhead is a Scottish Government Clinical Academic Fellow and was supported by a Harvard University Frank Knox Memorial Fellowship.
We thank the Nurses' Health Study and the Health Professionals Follow-up Study cohort participants who have agreed to provide us with information through questionnaires and biological specimens; hospitals and pathology departments throughout the United States for generously providing us with tissue specimens.
Reiko Nishihara, Teppei Morikawa, Aya Kuchiba, and Paul Lochhead contributed equally, and Charles S. Fuchs, Andrew T. Chan, Edward Giovannucci, and Shuji Ogino contributed equally.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the US National Institutes of Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Conflict of interest: none declared.
REFERENCES
- 1.Botteri E, Iodice S, Bagnardi V, et al. Smoking and colorectal cancer: a meta-analysis. JAMA. 2008;300(23):2765–2778. doi: 10.1001/jama.2008.839. [DOI] [PubMed] [Google Scholar]
- 2.Liang PS, Chen TY, Giovannucci E. Cigarette smoking and colorectal cancer incidence and mortality: systematic review and meta-analysis. Int J Cancer. 2009;124(10):2406–2415. doi: 10.1002/ijc.24191. [DOI] [PubMed] [Google Scholar]
- 3.Verla-Tebit E, Lilla C, Hoffmeister M, et al. Cigarette smoking and colorectal cancer risk in Germany: a population-based case-control study. Int J Cancer. 2006;119(3):630–635. doi: 10.1002/ijc.21875. [DOI] [PubMed] [Google Scholar]
- 4.Leufkens AM, Van Duijnhoven FJ, Siersema PD, et al. Cigarette smoking and colorectal cancer risk in the European prospective investigation into cancer and nutrition study. Clin Gastroenterol Hepatol. 2011;9(2):137–144. doi: 10.1016/j.cgh.2010.10.012. [DOI] [PubMed] [Google Scholar]
- 5.Slattery ML, Potter JD, Friedman GD, et al. Tobacco use and colon cancer. Int J Cancer. 1997;70(3):259–264. doi: 10.1002/(sici)1097-0215(19970127)70:3<259::aid-ijc2>3.0.co;2-w. [DOI] [PubMed] [Google Scholar]
- 6.Newcomb PA, Storer BE, Marcus PM. Cigarette smoking in relation to risk of large bowel cancer in women. Cancer Res. 1995;55(21):4906–4909. [PubMed] [Google Scholar]
- 7.Weijenberg MP, Aardening PW, de Kok TM, et al. Cigarette smoking and KRAS oncogene mutations in sporadic colorectal cancer: results from the Netherlands Cohort Study. Mutat Res. 2008;652(1):54–64. doi: 10.1016/j.mrgentox.2007.12.008. [DOI] [PubMed] [Google Scholar]
- 8.Ogino S, Goel A. Molecular classification and correlates in colorectal cancer. J Mol Diagn. 2008;10(1):13–27. doi: 10.2353/jmoldx.2008.070082. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lao VV, Grady WM. Epigenetics and colorectal cancer. Nat Rev Gastroenterol Hepatol. 2011;8(12):686–700. doi: 10.1038/nrgastro.2011.173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ogino S, Stampfer M. Lifestyle factors and microsatellite instability in colorectal cancer: the evolving field of molecular pathological epidemiology. J Natl Cancer Inst. 2010;102(6):365–367. doi: 10.1093/jnci/djq031. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ogino S, Chan AT, Fuchs CS, et al. Molecular pathological epidemiology of colorectal neoplasia: an emerging transdisciplinary and interdisciplinary field. Gut. 2011;60(3):397–411. doi: 10.1136/gut.2010.217182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Toyota M, Ahuja N, Ohe-Toyota M, et al. CpG island methylator phenotype in colorectal cancer. Proc Natl Acad Sci U S A. 1999;96(15):8681–8686. doi: 10.1073/pnas.96.15.8681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Samowitz WS, Albertsen H, Herrick J, et al. Evaluation of a large, population-based sample supports a CpG island methylator phenotype in colon cancer. Gastroenterology. 2005;129(3):837–845. doi: 10.1053/j.gastro.2005.06.020. [DOI] [PubMed] [Google Scholar]
- 14.Curtin K, Slattery ML, Samowitz WS. CpG island methylation in colorectal cancer: past, present and future. Patholog Res Int. 2011;2011:902674. doi: 10.4061/2011/902674. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Teodoridis JM, Hardie C, Brown R. CpG island methylator phenotype (CIMP) in cancer: causes and implications. Cancer Lett. 2008;268(2):177–186. doi: 10.1016/j.canlet.2008.03.022. [DOI] [PubMed] [Google Scholar]
- 16.Hughes LA, Khalid-de Bakker CA, Smits KM, et al. The CpG island methylator phenotype in colorectal cancer: progress and problems. Biochim Biophys Acta. 2012;1825(1):77–85. doi: 10.1016/j.bbcan.2011.10.005. [DOI] [PubMed] [Google Scholar]
- 17.The Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature. 2012;487(7407):330–337. doi: 10.1038/nature11252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Weisenberger DJ, Siegmund KD, Campan M, et al. CpG island methylator phenotype underlies sporadic microsatellite instability and is tightly associated with BRAF mutation in colorectal cancer. Nat Genet. 2006;38(7):787–793. doi: 10.1038/ng1834. [DOI] [PubMed] [Google Scholar]
- 19.Barault L, Charon-Barra C, Jooste V, et al. Hypermethylator phenotype in sporadic colon cancer: study on a population-based series of 582 cases. Cancer Res. 2008;68(20):8541–8546. doi: 10.1158/0008-5472.CAN-08-1171. [DOI] [PubMed] [Google Scholar]
- 20.Nosho K, Irahara N, Shima K, et al. Comprehensive biostatistical analysis of CpG island methylator phenotype in colorectal cancer using a large population-based sample. PLoS One. 2008;3(11):e3698. doi: 10.1371/journal.pone.0003698. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Linhart HG, Lin H, Yamada Y, et al. Dnmt3b promotes tumorigenesis in vivo by gene-specific de novo methylation and transcriptional silencing. Genes Dev. 2007;21(23):3110–3122. doi: 10.1101/gad.1594007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Nosho K, Shima K, Irahara N, et al. DNMT3B expression might contribute to CpG island methylator phenotype in colorectal cancer. Clin Cancer Res. 2009;15(11):3663–3671. doi: 10.1158/1078-0432.CCR-08-2383. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ibrahim AE, Arends MJ, Silva AL, et al. Sequential DNA methylation changes are associated with DNMT3B overexpression in colorectal neoplastic progression. Gut. 2011;60(4):499–508. doi: 10.1136/gut.2010.223602. [DOI] [PubMed] [Google Scholar]
- 24.Steine EJ, Ehrich M, Bell GW, et al. Genes methylated by DNA methyltransferase 3b are similar in mouse intestine and human colon cancer. J Clin Invest. 2011;121(5):1748–1752. doi: 10.1172/JCI43169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Palakurthy RK, Wajapeyee N, Santra MK, et al. Epigenetic silencing of the RASSF1A tumor suppressor gene through HOXB3-mediated induction of DNMT3B expression. Mol Cell. 2009;36(2):219–230. doi: 10.1016/j.molcel.2009.10.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Curtin K, Samowitz WS, Wolff RK, et al. Somatic alterations, metabolizing genes and smoking in rectal cancer. Int J Cancer. 2009;125(1):158–164. doi: 10.1002/ijc.24338. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Samowitz WS, Albertsen H, Sweeney C, et al. Association of smoking, CpG island methylator phenotype, and V600E BRAF mutations in colon cancer. J Natl Cancer Inst. 2006;98(23):1731–1738. doi: 10.1093/jnci/djj468. [DOI] [PubMed] [Google Scholar]
- 28.Limsui D, Vierkant RA, Tillmans LS, et al. Cigarette smoking and colorectal cancer risk by molecularly defined subtypes. J Natl Cancer Inst. 2010;102(14):1012–1022. doi: 10.1093/jnci/djq201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Poynter JN, Haile RW, Siegmund KD, et al. Associations between smoking, alcohol consumption, and colorectal cancer, overall and by tumor microsatellite instability status. Cancer Epidemiol Biomarkers Prev. 2009;18(10):2745–2750. doi: 10.1158/1055-9965.EPI-09-0517. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Lindor NM, Yang P, Evans I, et al. Alpha-1-antitrypsin deficiency and smoking as risk factors for mismatch repair deficient colorectal cancer: a study from the colon cancer family registry. Mol Genet Metab. 2010;99(2):157–159. doi: 10.1016/j.ymgme.2009.09.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Slattery ML, Curtin K, Anderson K, et al. Associations between cigarette smoking, lifestyle factors, and microsatellite instability in colon tumors. J Natl Cancer Inst. 2000;92(22):1831–1836. doi: 10.1093/jnci/92.22.1831. [DOI] [PubMed] [Google Scholar]
- 32.Phipps AI, Baron J, Newcomb PA. Prediagnostic smoking history, alcohol consumption, and colorectal cancer survival: The Seattle Colon Cancer Family Registry. Cancer. 2011;117(21):4948–4957. doi: 10.1002/cncr.26114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Eaton AM, Sandler R, Carethers JM, et al. 5,10-Methylenetetrahydrofolate reductase 677 and 1298 polymorphisms, folate intake, and microsatellite instability in colon cancer. Cancer Epidemiol Biomarkers Prev. 2005;14(8):2023–2029. doi: 10.1158/1055-9965.EPI-05-0131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Chia VM, Newcomb PA, Bigler J, et al. Risk of microsatellite-unstable colorectal cancer is associated jointly with smoking and nonsteroidal anti-inflammatory drug use. Cancer Res. 2006;66(13):6877–6883. doi: 10.1158/0008-5472.CAN-06-1535. [DOI] [PubMed] [Google Scholar]
- 35.Rozek LS, Herron CM, Greenson JK, et al. Smoking, gender, and ethnicity predict somatic BRAF mutations in colorectal cancer. Cancer Epidemiol Biomarkers Prev. 2010;19(3):838–843. doi: 10.1158/1055-9965.EPI-09-1112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Wang T, Chen M, Liu L, et al. Nicotine induced CpG methylation of Pax6 binding motif in StAR promoter reduces the gene expression and cortisol production. Toxicol Appl Pharmacol. 2011;257(3):328–337. doi: 10.1016/j.taap.2011.09.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Liu F, Killian JK, Yang M, et al. Epigenomic alterations and gene expression profiles in respiratory epithelia exposed to cigarette smoke condensate. Oncogene. 2010;29(25):3650–3664. doi: 10.1038/onc.2010.129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Du H, Sun J, Chen Z, et al. Cigarette smoke-induced failure of apoptosis resulting in enhanced neoplastic transformation in human bronchial epithelial cells. J Toxicol Environ Health A. 2012;75(12):707–720. doi: 10.1080/15287394.2012.690088. [DOI] [PubMed] [Google Scholar]
- 39.Chan AT, Ogino S, Fuchs CS. Aspirin and the risk of colorectal cancer in relation to the expression of COX-2. N Engl J Med. 2007;356(21):2131–2142. doi: 10.1056/NEJMoa067208. [DOI] [PubMed] [Google Scholar]
- 40.Liao X, Lochhead P, Nishihara R, et al. Aspirin use, tumor PIK3CA mutation status, and colorectal cancer survival. N Engl J Med. 2012;367(17):1596–1606. doi: 10.1056/NEJMoa1207756. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Giovannucci E, Colditz GA, Stampfer MJ, et al. A prospective study of cigarette smoking and risk of colorectal adenoma and colorectal cancer in U.S. women. J Natl Cancer Inst. 1994;86(3):192–199. doi: 10.1093/jnci/86.3.192. [DOI] [PubMed] [Google Scholar]
- 42.Giovannucci E, Rimm EB, Stampfer MJ, et al. A prospective study of cigarette smoking and risk of colorectal adenoma and colorectal cancer in U.S. men. J Natl Cancer Inst. 1994;86(3):183–191. doi: 10.1093/jnci/86.3.183. [DOI] [PubMed] [Google Scholar]
- 43.Yamauchi M, Morikawa T, Kuchiba A, et al. Assessment of colorectal cancer molecular features along bowel subsites challenges the conception of distinct dichotomy of proximal versus distal colorectum. Gut. 2012;61(6):847–854. doi: 10.1136/gutjnl-2011-300865. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Yamauchi M, Lochhead P, Morikawa T, et al. Colorectal cancer: a tale of two sides or a continuum? Gut. 2012;61(6):794–797. doi: 10.1136/gutjnl-2012-302014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Morikawa T, Kuchiba A, Yamauchi M, et al. Association of CTNNB1 (beta-catenin) alterations, body mass index, and physical activity with survival in patients with colorectal cancer. JAMA. 2011;305(16):1685–1694. doi: 10.1001/jama.2011.513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Ogino S, Kawasaki T, Kirkner GJ, et al. CpG island methylator phenotype-low (CIMP-low) in colorectal cancer: possible associations with male sex and KRAS mutations. J Mol Diagn. 2006;8(5):582–588. doi: 10.2353/jmoldx.2006.060082. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Ogino S, Nosho K, Kirkner GJ, et al. CpG island methylator phenotype, microsatellite instability, BRAF mutation and clinical outcome in colon cancer. Gut. 2009;58(1):90–96. doi: 10.1136/gut.2008.155473. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Ogino S, Kawasaki T, Brahmandam M, et al. Precision and performance characteristics of bisulfite conversion and real-time PCR (MethyLight) for quantitative DNA methylation analysis. J Mol Diagn. 2006;8(2):209–217. doi: 10.2353/jmoldx.2006.050135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med. 1989;8(5):551–561. doi: 10.1002/sim.4780080504. [DOI] [PubMed] [Google Scholar]
- 50.Lunn M, McNeil D. Applying Cox regression to competing risks. Biometrics. 1995;51(2):524–532. [PubMed] [Google Scholar]
- 51.Ogino S, Giovannucci E. Commentary: lifestyle factors and colorectal cancer microsatellite instability—molecular pathological epidemiology science, based on unique tumour principle. Int J Epidemiol. 2012;41(4):1072–1074. doi: 10.1093/ije/dys076. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Canales L, Chen J, Kelty E, et al. Developmental cigarette smoke exposure: liver proteome profile alterations in low birth weight pups. Toxicology. 2012;300(1-2):1–11. doi: 10.1016/j.tox.2012.04.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Breitling LP, Yang R, Korn B, et al. Tobacco-smoking-related differential DNA methylation: 27K discovery and replication. Am J Hum Genet. 2011;88(4):450–457. doi: 10.1016/j.ajhg.2011.03.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Wan ES, Qiu W, Baccarelli A, et al. Cigarette smoking behaviors and time since quitting are associated with differential DNA methylation across the human genome. Hum Mol Genet. 2012;21(13):3073–3082. doi: 10.1093/hmg/dds135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Ogino S, Galon J, Fuchs CS, et al. Cancer immunology—analysis of host and tumor factors for personalized medicine. Nat Rev Clin Oncol. 2011;8(12):711–719. doi: 10.1038/nrclinonc.2011.122. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Ogino S, Fuchs CS, Giovannucci E. How many molecular subtypes? Implications of the unique tumor principle in personalized medicine. Expert Rev Mol Diagn. 2012;12(6):621–628. doi: 10.1586/erm.12.46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Ogino S, King EE, Beck AH, et al. Interdisciplinary education to integrate pathology and epidemiology: towards molecular and population-level health science. Am J Epidemiol. 2012;176(8):659–667. doi: 10.1093/aje/kws226. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Kuller LH. Invited commentary: the 21st century epidemiologist—a need for different training? Am J Epidemiol. 2012;176(8):668–671. doi: 10.1093/aje/kws227. [DOI] [PubMed] [Google Scholar]
- 59.Ogino S, Beck AH, King EE, et al. Ogino et al. Respond to “The 21st Century Epidemiologist”. Am J Epidemiol. 2012;176(8):672–674. doi: 10.1093/aje/kws229. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Hughes LA, Simons CC, van den Brandt PA, et al. Body size, physical activity and risk of colorectal cancer with or without the CpG Island Methylator Phenotype (CIMP) PLoS One. 2011;6(4):e18571. doi: 10.1371/journal.pone.0018571. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Limsui D, Vierkant RA, Tillmans LS, et al. Postmenopausal hormone therapy and colorectal cancer risk by molecularly defined subtypes among older women. Gut. 2012;61(9):1299–1305. doi: 10.1136/gutjnl-2011-300719. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Limburg PJ, Limsui D, Vierkant RA, et al. Postmenopausal hormone therapy and colorectal cancer risk in relation to somatic KRAS mutation status among older women. Cancer Epidemiol Biomarkers Prev. 2012;21(4):681–684. doi: 10.1158/1055-9965.EPI-11-1168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Gay LJ, Mitrou PN, Keen J, et al. Dietary, lifestyle and clinico-pathological factors associated with APC mutations and promoter methylation in colorectal cancers from the EPIC-Norfolk Study. J Pathol. 2012;228(3):405–415. doi: 10.1002/path.4085. [DOI] [PubMed] [Google Scholar]
- 64.Schernhammer ES, Giovannucci E, Kawasaki T, et al. Dietary folate, alcohol and B vitamins in relation to LINE-1 hypomethylation in colon cancer. Gut. 2010;59(6):794–799. doi: 10.1136/gut.2009.183707. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Kuchiba A, Morikawa T, Yamauchi M, et al. Body mass index and risk of colorectal cancer according to fatty acid synthase expression in the Nurses’ Health Study. J Natl Cancer Inst. 2012;104(5):415–420. doi: 10.1093/jnci/djr542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Yamaji T, Iwasaki M, Sasazuki S, et al. Association between plasma 25-hydroxyvitamin D and colorectal adenoma according to dietary calcium intake and vitamin D receptor polymorphism. Am J Epidemiol. 2012;175(3):236–244. doi: 10.1093/aje/kwr295. [DOI] [PubMed] [Google Scholar]
- 67.Lubbe SJ, Di Bernardo MC, Broderick P, et al. Comprehensive evaluation of the impact of 14 genetic variants on colorectal cancer phenotype and risk. Am J Epidemiol. 2012;175(1):1–10. doi: 10.1093/aje/kwr285. [DOI] [PubMed] [Google Scholar]
- 68.Shin A, Hong CW, Sohn DK, et al. Associations of cigarette smoking and alcohol consumption with advanced or multiple colorectal adenoma risks: a colonoscopy-based case-control study in Korea. Am J Epidemiol. 2011;174(5):552–562. doi: 10.1093/aje/kwr098. [DOI] [PubMed] [Google Scholar]
- 69.Chia WK, Ali R, Toh HC. Aspirin as adjuvant therapy for colorectal cancer-reinterpreting paradigms. Nat Rev Clin Oncol. 2012;9(10):561–570. doi: 10.1038/nrclinonc.2012.137. [DOI] [PubMed] [Google Scholar]
- 70.Diergaarde B, Vrieling A, van Kraats AA, et al. Cigarette smoking and genetic alterations in sporadic colon carcinomas. Carcinogenesis. 2003;24(3):565–571. doi: 10.1093/carcin/24.3.565. [DOI] [PubMed] [Google Scholar]
- 71.Satia JA, Keku T, Galanko JA, et al. Diet, lifestyle, and genomic instability in the North Carolina Colon Cancer Study. Cancer Epidemiol Biomarkers Prev. 2005;14(2):429–436. doi: 10.1158/1055-9965.EPI-04-0486. [DOI] [PubMed] [Google Scholar]
- 72.Gay LJ, Arends MJ, Mitrou PN, et al. MLH1 promoter methylation, diet, and lifestyle factors in mismatch repair deficient colorectal cancer patients from EPIC-Norfolk. Nutr Cancer. 2011;63(7):1000–1010. doi: 10.1080/01635581.2011.596987. [DOI] [PubMed] [Google Scholar]
- 73.Ang PW, Loh M, Liem N, et al. Comprehensive profiling of DNA methylation in colorectal cancer reveals subgroups with distinct clinicopathological and molecular features. BMC Cancer. 2010;10:227. doi: 10.1186/1471-2407-10-227. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 74.Dahlin AM, Palmqvist R, Henriksson ML, et al. The role of the CpG island methylator phenotype in colorectal cancer prognosis depends on microsatellite instability screening status. Clin Cancer Res. 2010;16(6):1845–1855. doi: 10.1158/1078-0432.CCR-09-2594. [DOI] [PubMed] [Google Scholar]
- 75.Jover R, Nguyen TP, Perez-Carbonell L, et al. 5-Fluorouracil adjuvant chemotherapy does not increase survival in patients with CpG island methylator phenotype colorectal cancer. Gastroenterology. 2011;140(4):1174–1181. doi: 10.1053/j.gastro.2010.12.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 76.Hinoue T, Weisenberger DJ, Lange CP, et al. Genome-scale analysis of aberrant DNA methylation in colorectal cancer. Genome Res. 2012;22(2):271–282. doi: 10.1101/gr.117523.110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 77.Wong JJ, Hawkins NJ, Ward RL, et al. Methylation of the 3p22 region encompassing MLH1 is representative of the CpG island methylator phenotype in colorectal cancer. Mod Pathol. 2011;24(3):396–411. doi: 10.1038/modpathol.2010.212. [DOI] [PubMed] [Google Scholar]
- 78.Zlobec I, Bihl M, Foerster A, et al. Comprehensive analysis of CpG Island Methylator Phenotype (CIMP)-high, -low, and -negative colorectal cancers based on protein marker expression and molecular features. J Pathol. 2011;225(3):336–343. doi: 10.1002/path.2879. [DOI] [PubMed] [Google Scholar]
- 79.Yamamoto E, Suzuki H, Yamano HO, et al. Molecular dissection of premalignant colorectal lesions reveals early onset of the CpG Island Methylator Phenotype. Am J Pathol. 2012;181(5):1847–1861. doi: 10.1016/j.ajpath.2012.08.007. [DOI] [PubMed] [Google Scholar]
- 80.Yagi K, Takahashi H, Akagi K, et al. Intermediate methylation epigenotype and its correlation to KRAS mutation in conventional colorectal adenoma. Am J Pathol. 2012;180(2):616–625. doi: 10.1016/j.ajpath.2011.10.010. [DOI] [PubMed] [Google Scholar]
- 81.Sproul D, Kitchen RR, Nestor CE, et al. Tissue of origin determines cancer-associated CpG island promoter hypermethylation patterns. Genome Biol. 2012;13(10):R84. doi: 10.1186/gb-2012-13-10-r84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 82.Xia D, Wang D, Kim SH, et al. Prostaglandin E(2) promotes intestinal tumor growth via DNA methylation. Nat Med. 2012;18(2):224–226. doi: 10.1038/nm.2608. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 83.Wu C, Bekaii-Saab T. CpG island methylation, microsatellite instability, and BRAF mutations and their clinical application in the treatment of colon cancer. Chemother Res Pract. 2012;2012:359041. doi: 10.1155/2012/359041. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.