Smoking Exposure as a Risk Factor for Prevalent and Recurrent Colorectal Adenomas¹

Colorectal adenomatous polyps are considered precursor lesions for colorectal cancer (1, 2, 3). Greater understanding of the factors associated with the formation of these precursor lesions and with their progression to colorectal cancer may provide opportunities for early detection and prevention. Characteristics of prevalent adenomas, specifically size, number, histology and location in the colon, have been found to affect the risk of colon cancer (4, 5, 6, 7, 8, 9).

Across several study designs, 21 of 22 studies cited showed positive associations between smoking and occurrence of adenomatous polyps (10), with smoking described by ever- versus never-smoking (11, 12, 13), by smoking status (current, former, and never-smokers; Refs. 11, 12, 13) or by pack-years (14, 15, 16, 17, 18, 19). Smoking has also been studied in relation to characteristics of prevalent adenomas and to adenoma recurrence. Two studies showed that smoking was related to adenoma prevalence but not to specific histology (16, 20). Zahm et al. (16) found a positive relationship between smoking and multiplicity and size but not histology. These results were mirrored in Monnet et al. (20), who looked specifically at smoking status and histology. Boutron et al. (21) found that smoking was independently related to large adenomas in a case-control study.

Lee et al. (22), using a case-control study design, found a significant dose-response relationship in males between pack-years of smoking and the location of the adenoma with ORs³ ranging from 1.3 to 3.5 across levels of smoking exposure for proximal adenomas and OR from 1.3 to 2.0 across exposure levels for distal adenomas. The association between smoking and large adenomas has been studied by several authors (2, 18, 22, 23, 24, 25, 26). The results of these studies show that shorter term cigarette smoking exposure is more strongly related to smaller adenomas, suggesting that smoking exposure may have an effect on the earlier development of adenomas. There has been some evidence linking smoking directly with adenoma recurrence (10, 27, 28), primarily after ≥20 years of smoking.

To evaluate the relationship among smoking, prevalent adenoma characteristics, and adenoma recurrence, we conducted this investigation using data from a completed clinical trial of colorectal adenoma recurrence. Two sets of analyses were performed. The first was a cross-section analysis of the association of several measures of smoking exposure to prevalent adenoma characteristics was completed. Adenoma characteristics evaluated included adenoma location, histology, size, and number. The second was a prospective analysis of the association of smoking exposure and adenoma recurrence.

The analysis was based on participants randomized into a trial of WBF versus placebo fiber at the Arizona Cancer Center. The details of this Phase III trial design have already been previously described (29) and the primary end point results have been published (30). Briefly, this trial investigated the effects of WBF supplementation on the recurrence of adenomatous polyps among individuals with recently resected colorectal adenomas. Men and women, ages 40–80 years with one or more colorectal adenoma(s) ≥3 mm removed at colonoscopy within 3 months before study entry and without a personal history of inflammatory bowel disease or hereditary colon cancer syndromes, were recruited from the greater Phoenix metropolitan area. Risk factor data were collected at baseline. A total of 1304 subjects of the original 1429 randomized (91%) underwent at least one subsequent colonoscopy, marking successful completion of the study protocol. WBF supplementation had no effect on adenoma recurrence. The University of Arizona Human Subjects Committee and Phoenix Metropolitan area hospital review committees approved this study.

Colonoscopy and polyp pathology reports on baseline and recurrent adenomas were reviewed by trained staff according to a standardized protocol and information on each polyp and procedure was extracted. The review abstracted location, histology according to the community pathologist, and size. The diagnosis from the community pathologist was used in this analysis, except when the specific adenoma classification was listed as incipient or unspecified. In those cases (∼30% of all adenomas), the more specific diagnosis of the study pathologist (A. K. B.) was used. This method provided the most complete and consistent data on histology available. The dependent variables in this study included distal location (rectum, rectosigmoid, sigmoid colon, descending colon, and splenic flexure) versus proximal adenoma location (transverse colon, hepatic flexure, ascending colon, and cecum). Individuals with baseline adenomas in both the distal and proximal colon were excluded from the analysis of location (n = 1161) but included in the analyses of the other adenoma characteristics (n = 1429). Tubular adenomas, defined as adenomas with <25% villous tissue were compared with tubulovillous and villous that contained 26–75% and >75% villous tissue, respectively. Adenoma size was determined for the pathology report. When this information was missing, the size was determined from the operative report. Adenoma size was divided into three categories: ≤0.5 cm; 0.6 to <1.0 cm; and ≥ 1 cm. For individuals with more than one adenoma, the size of the largest adenoma found was used. Adenoma number was divided into three categories: 1 adenoma; 2 adenomas; and ≥3 adenomas. Although the average number of adenomas/person was 1.82 (SD = 1.3), 57% of the randomized population had only one adenoma.

Smoking exposure was assessed at baseline using a self-administered questionnaire. Ever-smoked status was defined as having smoked at least 100 cigarettes. Anyone reporting smoking in the last 30 days was classified as a current smoker. Current and former smokers were asked to report the age they started smoking, years of smoking, and the average number of cigarettes smoked/day. Measures of smoking exposure considered in this study included smoking status, cigarettes/day, years of smoking, and pack-years. Intervals of cigarettes/day were defined as <20, 20–29, ≥30. These intervals were selected because they represented the way these data were reported in units of packs (20 cigarettes) and half-packs. Intervals of years of smoking (<15, 15 to <25, 25 to <35, and ≥35 years) were used in previous articles of smoking and adenomas (31). Three intervals of pack-years of smoking (<10, 10 to <30, and ≥30 pack-years) were selected that would distinguish low, medium, and high exposures. Pack-years were calculated as cigarettes/day/20 × years of smoking. Never-smokers were included in all smoking measures as 0, designating no exposure, and used as the reference group. Each smoking variable was also evaluated by tertiles and as a continuous variable. Additional risk factors considered in these analyses included age (continuous), gender, BMI (continuous), family history of colorectal cancer, history of previous polyps, and alcohol consumption (g/day; continuous).

Differences in the distribution of baseline characteristics between the 1429 randomized participants and the 1304 subjects with complete follow-up were evaluated using χ² and t tests. Unconditional logistic regression was used to generate risk estimates for the association of smoking with location, histology, and recurrence. Unordered polytomous logistic regression was used to generate association estimates for smoking exposure and adenoma size and number. All adjusted models accounted for age (continuous), gender, BMI (continuous), previous history of polyps, and other adenoma characteristics. Other potential confounders such as family history of colorectal cancer and alcohol consumption were evaluated but were not included in the final models because they did not contribute significantly to outcome.

The other adenoma characteristics were included in each model because of observed associations among the characteristics. The adjusted model for adenoma recurrence also included the number of colonoscopies during the study follow-up period and calcium intake (mg/day) as described by Martinez et al. (32).

The distribution of several the risk factor and demographic characteristics have been described in previous publications (30). In summary, there were no significant differences between the complete randomized sample and the subjects who complete follow-up. Approximately 52% were former smokers and 14% were current smokers. Over 50% of the women were never-smokers compared with 62% of men who were categorized as former smokers. Women generally smoked fewer cigarettes/day than men (27 versus 55%, respectively, P < 0.001), but proportionally, both genders smoked for as many years (data not shown). With respect to baseline adenoma characteristics, 54% had only distal adenomas, 70% had adenomas with <25% villous tissue, and 28% had at least one adenoma ≥1 cm (data not shown). Overall, subjects had an average of 1.8 adenomas at baseline. A greater proportion of men had multiple adenomas than women (47 versus 35%, respectively, P < 0.001).

Table 1 summarizes the adjusted ORs for distal versus proximal adenomas and tubular versus TV/V adenomas by the smoking exposure variables. Compared with never-smokers, only former smokers have a significantly elevated prevalence of distal versus proximal adenomas (OR, 1.47; 95% CI, 1.09–1.98). The two lowest levels of cigarettes/day (<20 and 20–29) were significantly elevated (OR, 1.49; 95% CI, 1.02–2.17 and OR, 1.48; 95% CI, 1.02–2.16, respectively). Both the 15 to <25 and ≥35 intervals of years of smoking showed significant increases in prevalence (OR, 1.95; 95% CI, 1.23–3.09 and OR, 1.54; 95% CI, 1.08–2.21, respectively). Intervals of pack-years show similar increases in prevalence. None of the trends were significant. Current smokers had significantly elevated prevalence of tubular versus TV/V histology (OR, 2.17; 95% CI, 1.39–3.39). Cigarettes/day, years of smoking, and pack-years showed significantly increased prevalence estimates, ∼70% for upper categories, compared with never smokers. In a gender stratified analysis, men appeared to drive this relationship, although the differences in the risk estimates were not significant.

Table 2 presents the adjusted ORs for adenoma number and size by smoking exposure. Compared with never-smokers, former and current smokers had significantly elevated prevalence of two adenomas versus one (OR, 1.42; 95% CI, 1.02–1.96 and OR, 2.31; 95% CI, 1.47–3.63, respectively). One and one-half packs/day and ≥35 years of smoking represented significant increases in prevalence (OR, 2.21; 95% CI, 1.51–3.23 and OR, 1.85; 95% CI, 1.28–2.68, respectively). The estimates for ≥3 versus one adenoma with smoking ≥35 years was significantly increased (OR, 1.74; 95% CI, 1.15–2.65; data not shown).

Smoking >30 cigarettes/day and smoking for between 25 and <35 years was significantly associated with an increased risk of large adenomas (≥1.0 cm) versus small adenomas (≤0.5 cm) compared with never-smokers, with OR, 1.94; 95% CI, 1.24–3.02 and OR, 1.94; 95% CI, 1.17–3.22, respectively. Thirty or more pack-years were significantly associated with larger versus small adenomas (OR, 1.73; 95% CI, 1.16–2.59). Similar prevalence estimates were seen with adenomas 0.6 to <1.0 cm (data not shown). Women smokers appeared to drive these findings, although the risk estimates between genders were not significantly different.

Table 3 presents the unadjusted and adjusted ORs for smoking with respect to risk of adenoma recurrence. The risk estimate for smoking ≥35 years was significantly increased (OR, 1.42; 95% CI, 1.01–1.98).

The goal of this study was to investigate the association between exposure to cigarette smoking and adenomatous polyps, specifically their location, size, histology, multiplicity, and recurrence.

Exposure to smoking was associated with significantly increased prevalence of distal versus proximal adenomas. In addition, cigarettes/day and years of smoking showed a consistent elevation in prevalence. These results agree with those published by Lee et al. (22) in a case-control study that suggested that the effect of smoking exposure was greater for distal adenomas (OR, 3.5; 95% CI, 1.7–7.0 in subject with >40 pack-years exposure) than for proximal adenomas when compared with controls without adenomas. Intervals of cigarettes/day and years of smoking had a significant positive association with tubular versus villous adenomas, peaking at >30 cigarettes/day, with a significant trend overall. These results contrast with a study by Monnet et al. (20), who found no relationship between smoking and adenoma histology in a case-control study.

The prevalence of multiple adenomas (two and more than three adenomas) versus singular adenomas was also significantly associated with current smoking status, cigarettes/day, and years of smoking. The results are also supported by Lee et al. (22) in a case-control study. The risk of adenomas >0.5 versus ≤0.5 cm was increased in former smokers and significantly increased in subjects smoking ≥30 cigarettes/day and between 25 and 35 years of smoking. These results reflect conclusions drawn by several investigators (10, 16, 21) who found, across several study designs, strong evidence that smoking was associated with larger adenomas. There is some suggestion, although inconclusive, that gender may influence the effect of smoking on adenoma characteristics. However, larger studies that include more women with a history of smoking are necessary to clarify true gender effects from differences in smoking patterns.

The prevalence of adenoma recurrence was elevated only in the longest duration of exposure (≥35 years). This supports the findings of Nagata et al. (27), Giovannucci (10), and Terry and Neugut (28) who found associations between newly diagnosed adenoma recurrence and 20, 30+, and ≥40 years of smoking, respectively. Jacobson et al. (18) found that a substantial pack-year exposure (between 30 and 40 pack-years) was associated with a 2-fold increased risk with increasing pack-year exposure. In a recent article by Terry et al. (31), which used data pooled from four large case-control studies of adenomatous polyps, smoking exposure was found to be related to newly diagnosed advanced and nonadvanced adenomas. This association did not differentially increase the risk of advanced versus nonadvanced adenomas. In contrast, Baron et al. (33) reported that only extremely high doses of smoking (>60 cigarettes/day) suggested an increased risk of recurrence (relative risk = 1.55).

The major limitation of this study was that the evaluation of adenoma characteristics was done in a cross-sectional manner, using subjects who had adenomas at baseline. This limits the interpretation of these results because they represent prevalence of one characteristic versus another and not comparisons to adenoma-free subjects. Our conclusions can only be applied to individuals who have already formed an adenoma and cannot explore temporal or causative associations. In addition, the analysis for adenoma recurrence did not account for changes in smoking habits during the course of follow-up or for continued exposures. Assuming that a small percentage of the smokers quit during the course of the trial (personal communication), there would be a minimal effect on the risk estimates for smoking status. Smoking during the trial adds only 2–3 years of smoking at most to relatively large exposures measured at baseline.

The major strength of this study is that it used information from a well-characterized population with detailed data on characteristics of baseline adenomas and multiple factors associated with adenomatous polyps, including a history of smoking. Smoking exposure was found to be associated with distal versus proximal, tubular versus any villous, large versus small, and multiple versus single adenomas. Smoking was marginally associated with adenoma recurrence and only among participants who smoked for more than three decades. Given these results in summary, smoking appears to be related to some adenoma characteristics that reflect lower risk of progression: distal and tubular, while also being associated with characteristics considered to increase risk of progression; and large size and multiplicity. These results support suggestions that the impact of smoking on adenoma development, in addition to being early in the carcinongenesis pathway (34), may involve a genetic subtype of adenomatous polyps (35, 36). Alternatively, smoking may increase the prevalence of tubular adenomas that may grow slowly into large adenomas, explaining the long induction period between smoking initiation and colorectal cancer development. Additional investigations of adenoma subtypes that include both detailed exposure histories and comprehensive genetic profiles, with sufficient sample sizes to evaluate the differences between genders, may additionally explain the complicated relationship between smoking exposure, adenomas, and colorectal cancer risk.

We thank Janine Einspahr, Cheryl Kramer, and Nancy Hart for their expert assistance and the staff and physicians from the Phoenix study sites for their valuable contributions.