Fax: (312) 908-9588
Cigarette smoking and colorectal carcinoma mortality in a cohort with long-term follow-up†
Article first published online: 1 DEC 2003
Copyright © 2003 American Cancer Society
Volume 100, Issue 2, pages 288–293, 15 January 2004
How to Cite
Colangelo, L. A., Gapstur, S. M., Gann, P. H. and Dyer, A. R. (2004), Cigarette smoking and colorectal carcinoma mortality in a cohort with long-term follow-up. Cancer, 100: 288–293. doi: 10.1002/cncr.11923
See related editorial on pages 223–4, this issue.
- Issue published online: 5 JAN 2004
- Article first published online: 1 DEC 2003
- Manuscript Accepted: 13 AUG 2003
- Manuscript Received: 23 JUN 2003
- National Cancer Institute. Grant Number: PHS Grant R03 CA81617
- National Heart Lung and Blood Institute. Grant Number: PHS Grant R01 HL21010
- National Institutes of Health and Human Services
- colorectal carcinoma (CRC);
- cohort study
Evidence suggests that colorectal carcinoma (CRC) may be a tobacco-associated malignancy.
In the current study, the authors examined the association between cigarette smoking and CRC mortality in the Chicago Heart Association Detection Project in Industry study, a cohort of 39,299 men and women with an average of 26 years of follow-up. To assess whether the association was stronger in participants with a potentially long history of smoking, the authors also stratified the analysis using a baseline age ≥ 50 years versus < 50 years.
Using multivariate Cox regression analysis, there was a marginally significant trend (P = 0.06) for men and women combined between smoking and CRC mortality. In the age-stratified analysis in the older participant group, there was no apparent association for men, women, or men and women combined. In the younger participant group, there appeared to be dose-response relations for women and for men and women combined (P value for trend = 0.008 and 0.03, respectively) between smoking and CRC mortality. The relative risk for women who smoked >20 cigarettes/day compared with never smokers was 2.49 (95% confidence interval [95% CI], 0.87–7.12), and was 1.87 for men and women combined (95% CI, 1.08–3.22).
The results of the current study support an association between cigarette smoking and CRC mortality, particularly in women age < 50 years. Cancer 2004;100:288–93. © 2003 American Cancer Society.
Based on a review of epidemiologic studies examining the association between cigarette smoking and colorectal adenomas and carcinoma, Giovannucci1 proposed that colorectal carcinoma (CRC) should be added to the list of tobacco-associated malignancies. In this review, Giovannucci discussed several lines of evidence that support cigarette smoking as a risk factor for CRC after accounting for a sufficiently long lag time of 35–40 years. These lines of evidence were: 1) studies showing positive associations between long-term smoking and colorectal adenomas; 2) studies conducted during the 1950s and 1960s, which would have few men and women exceeding 40 years of smoking exposure, reporting no association between smoking and CRC; 3) studies of U.S. men with a follow-up time after 1970, by which time many smokers may have exceeded 40 years of smoking, reporting positive associations; 4) studies of U.S. women with follow-up times after 1990, the time when female smokers would achieve 40 years of smoking history, reporting positive associations; and 5) 13 studies conducted outside the U.S., 9 of which demonstrate long-term smokers to be at an elevated risk for CRC. It was emphasized that studies of men and women in the U.S. should not be expected to find positive associations unless the studies had follow-up times in the 1970s or later for men and in the 1990s or later for women.
The Chicago Heart Association Detection Project in Industry (CHA) is a large screening program on cardiovascular disease risk, with cigarette smoking information available for nearly 40,000 male and female participants. The CHA cohort was screened between 1967 and 1973 and followed for mortality through December 31, 1997. Although information regarding age at the initiation of smoking was not collected, if it is assumed that most of the male smokers began smoking at approximately age 18 years (the mean age at which U.S. men reportedly begin smoking2–4), then many male smokers in the CHA cohort who were age ≥ 50 years at baseline would have 30 years of smoking history at the time of baseline screening and 40 years after 10 years of follow-up. For women, the mean age at smoking initiation is reportedly 18 years for birth cohorts after 1940 and ranges from 20–32 years in birth cohorts prior to 1940.4 Therefore, many female smokers age > 50 years at baseline would have 30 years of smoking history at the time of baseline screening. Thus, this cohort provides an opportunity to evaluate the association between cigarette smoking and the risk of CRC mortality for both men and women who have potentially long histories of smoking.
MATERIAL AND METHODS
In the CHA, a prospective investigation of cardiovascular disease risk factors, 39,522 men and women from 84 cooperating companies and organizations in the Chicago area underwent baseline screening assessments between late 1967 and early 1973. Details regarding the recruitment of participants, methods of data collection, measurement of baseline risk factors, follow-up of the cohort, and determination of vital status have been described previously.5–7 The study received periodic institutional review board approval at Northwestern University (Chicago, IL). Because the study participants were not contacted directly in follow-up and were never identified by name in any report, this research was granted an exemption from the need to obtain informed consent from the participants after the baseline examination.
Participants excluded from the current analysis were 60 men and 76 women who did not have data recorded for height, weight, smoking, or education, and 45 men and 42 women who were missing vital status. This left 22,295 men and 17,004 women remaining in the cohort used for analysis.
Baseline risk factors included age, gender, race, education, number of cigarettes smoked per day, height, and body mass index (BMI) computed as weight in kilograms divided by height in meters squared. Height was included because previous studies8, 9 have found it to be correlated positively with the risk of CRC. Although BMI may be an adequate measure of body composition or adiposity, it may not fully capture information concerning body size independent of obesity.10
Between the time of the baseline examination until 1979, vital status was ascertained annually through direct mailings to individuals, submission of records to the Social Security Administration, mailings to employers, and direct telephone or neighborhood contact. After 1979, the National Death Index was used. Underlying cause of death was classified from death certificates and coded using the eighth revision of the International Classification of Diseases, adapted for use in the U.S. (ICD-8). Deaths from CRC were defined by ICD-8 codes 153 (colon) and 154 (rectum). For 98% of the cohort, follow-up continued until the date of death or until December 31, 1997. Approximately 2% of the patients were lost to follow-up within the first 11 years. Follow-up totaled 954,754 person-years.
Other potential risk factors selected for multivariate Cox proportional hazards analysis for cigarette smoking included age, race, education, BMI, and height. Five categories were used for cigarette smoking: never smoked, past smoker, currently smoke 1–10 cigarettes/day, currently smoke 11–20 cigarettes/day, and currently smoke > 20 cigarettes/day. Tests for trend were performed by assigning the values 1, 2,…5 to the five smoking categories and then modeling the smoking variable as a continuous variable. Analyses were conducted separately for men and women and by combining men and women with adjustment for gender. An additional analysis was performed stratifying by age < 50 years versus at age ≥ 50 years to assess whether the association between smoking and CRC mortality is stronger in the group of participants who have potentially 30 years of smoking exposure at the time of baseline screening (i.e., the group age ≥ 50 years at baseline). In this additional analysis, the 1–11 cigarettes/day and 11–20 cigarettes/day categories were combined because of the small number of deaths.
By December 31, 1997, there were 349 CRC deaths, 208 occurring among men (176 in the colon and 32 in the rectum) and 141 occurring among women (119 in the colon and 22 in the rectum). Characteristics of men and women based on smoking status are shown in Table 1. Current smokers were slightly younger than past and never smokers, had the lowest BMI, and were least likely to have at least some college education. Among male and female current smokers, 21% and 23%, respectively, were age ≥ 50 years at baseline.
|Never smoked||Past smoker||Current smoker||Never smoked||Past smoker||Current smoker|
|Age, mean (yrs)||39.6||43.1||38.7||42.1||39.4||37.9|
|Age, no. (%)|
|≤ 29 yrsa||1672 (30)||1156 (21)||2775 (50)||2113 (39)||909 (17)||2404 (44)|
|30–49 yrsb||2741 (25)||3310 (30)||4845 (44)||2697 (42)||998 (16)||2678 (42)|
|50–69 yrsc||1454 (26)||2152 (38)||2078 (37)||2705 (53)||794 (16)||1583 (31)|
|≥ 70 yrsd||44 (39)||44 (39)||24 (21)||92 (75)||11 (9)||20 (16)|
|Mean BMI (kg/m2)||26.7||27.0||26.2||24.5||24.2||23.5|
|Height, mean, cm||175.9||175.9||175.5||162.7||164.1||163.9|
|< High school||834 (19)||1173 (27)||2410 (55)||1806 (50)||421 (12)||1354 (38)|
|High school||1434 (21)||1981 (29)||3397 (50)||3654 (43)||1312 (16)||3488 (41)|
|Some college||1116 (26)||1300 (31)||1813 (43)||1194 (41)||526 (18)||1192 (41)|
|College graduate||2527 (37)||2208 (32)||2102 (31)||953 (46)||453 (22)||651 (32)|
|White||5370 (27)||6209 (31)||8537 (42)||6462 (45)||2333 (16)||5466 (38)|
|Black||315 (21)||308 (21)||854 (58)||883 (38)||326 (14)||1109 (48)|
|Other||226 (32)||145 (21)||331 (47)||262 (62)||53 (12)||110 (26)|
Table 2 shows age-adjusted and multivariate-adjusted relative risks (RRs) for smoking categories for men, women, and men and women combined. The multivariate-adjusted RRs were similar to or slightly lower than the age-adjusted RRs. Overall, there was no association with past smoking for men, women, or men and women combined. In multivariate analysis, men who smoked > 20 cigarettes/day had a 36% higher risk of CRC mortality compared with men who never smoked. For women, each of the 3 smoking categories (smoke 1–10 cigarettes/day, 11–20 cigarettes/day, and > 20 cigarettes/day) had elevated but nonsignificant RRs. When men and women were combined, there was a marginally significant dose-response relation between cigarette smoking and the risk of CRC mortality (P value for trend = 0.06).
|Smoking category||No. of deaths||Total person-years||Age-adjusted RR||95% CI||Multivariate adjusted RRa||95% CI|
|Smoke 1–10 cigs/day||10||43,226||0.75||0.38–1.47||0.75||0.38–1.48|
|Smoke 11–20 cigs/day||35||107,656||1.10||0.72–1.68||1.09||0.71–1.68|
|Smoke > 20 cigs/day||33||75,644||1.42||0.92–2.19||1.36||0.88–2.11|
|P value for trend||0.16||0.19|
|Smoke 1–10 cigs/day||17||56,206||1.26||0.74–2.14||1.23||0.72–2.09|
|Smoke 11–20 cigs/day||28||77,859||1.48||0.95–2.31||1.43||0.91–2.23|
|Smoke > 20 cigs/day||8||29,539||1.27||0.61–2.67||1.25||0.59–2.62|
|P value for trend||0.09||0.13|
|Men and women combined|
|Smoke 1–10 cigs/day||27||99,432||1.00||0.66–1.52||0.99||0.65–1.51|
|Smoke 11–20 cigs/day||63||185,515||1.25||0.92–1.71||1.22||0.89–1.66|
|Smoke > 20 cigs/day||41||105,183||1.42||0.99–2.05||1.37||0.95–1.98|
|P value for trend||0.03||0.06|
Table 3 shows the results of the analysis stratified by baseline age ≥ 50 years versus age < 50 years. In the analysis restricting the cohort to those with a baseline age < 50 years, there was a 56% higher risk of CRC mortality for men who smoked > 20 cigarettes/day compared with never smokers, but no evidence of an association was found for past smokers or for those who smoke ≤ 20 cigarettes/day. For women, there was strong evidence of a dose-response relation with an RR of 2.5 reported for women who smoke > 20 cigarettes/day. Similarly, for men and women combined the test for trend was statistically significant (P = 0.03). In the analysis restricting the cohort to those with a baseline age ≥ 50 years, there were no associations found between smoking and CRC mortality for men, women, or men and women combined.
|Smoking category||Baseline age < 50 yrs||Baseline age ≥ 50 yrs|
|No. of deaths||Total person-years||Multivariate adjusted RRa||95% CI||No. of deaths||Total person-years||Multivariate adjusted RRa||95% CI|
|Smoke 1–20 cigs/day||16||125,813||0.76||0.39–1.49||29||25,068||1.17||0.71–1.93|
|Smoke > 20 cigs/day||19||62,605||1.56||0.82–2.96||14||13,039||1.15||0.62–2.15|
|P value for trend||0.41||0.48|
|Smoke 1–20 cigs/day||21||104,566||2.40||1.17–4.91||24||29,500||1.01||0.62–1.63|
|Smoke > 20 cigs/day||5||24,336||2.49||0.87–7.12||3||5,203||0.75||0.23–2.40|
|P value for trend||0.008||0.84|
|Men and women combined|
|Smoke 1–20 cigs/day||37||230,379||1.29||0.79–2.08||53||54,568||1.07||0.76–1.52|
|Smoke > 20 cigs/day||24||86,941||1.87||1.08–3.22||17||18,242||1.04||0.61–1.76|
|P value for trend||0.03||0.73|
In this prospective study with 26 years of follow-up through 1997, there was no association detected between past cigarette smoking and CRC mortality. Compared with never smokers, the risks for those smoking > 20 cigarettes/day were elevated, but not significantly so, for men and women separately and for men and women combined. It is interesting to note that when the analysis was stratified by a baseline age ≥ 50 years versus a baseline age < 50 years, there was no association found between smoking and CRC for the older age group. Conversely, in the younger age group there was a positive association found between cigarette smoking and CRC mortality, which appeared to be stronger for women.
The findings in the current unstratified analysis are not entirely inconsistent with those of previous studies. Of the studies of men in the U.S. reviewed by Giovannucci,1 very few reported RRs that were as high as 2 and the majority were < 1.8. Among the studies of women in the U.S. reviewed by Giovannucci1 with follow-up in the 1990s, RRs approached or exceeded 2-fold for very long smoking durations only (> 40 years). The RR point estimates in the current study, although not statistically significant, are similar in magnitude to those estimated in the Cancer Prevention Study II.11 Chao et al. reported multivariate-adjusted RRs of 1.29, 1.25, 1.33, and 1.46 for the categories of < 20 cigarettes/day, 20 cigarettes/day, 21–39 cigarettes/day, and ≥ 40 cigarettes/day, respectively, for men. For women, the RRs were 1.32, 1.52, 1.26, and 1.65, respectively.
The findings in the current stratified analyses were contrary to expectation. It was expected that the association between cigarette smoking and CRC mortality would be stronger in the older age group compared with the younger age group. These expectations were based on: 1) the estimate that the induction period between smoking and elevated risk of CRC is 35–40 years in men12 and women13; and 2) the finding that U.S. studies of men and women that accounted for a time lag of 35–40 years supported an association between smoking and CRC.1 Although unexpected, the higher magnitude RRs reported in the younger age group are plausible given that this subgroup had a median of 26.5 years of follow-up through December 1997. It also is worth noting that the RRs were higher for women than for men at each level of cigarette smoking exposure, a phenomenon similar to that observed with lung carcinoma in other studies.14
The current study has some limitations that might have introduced bias into estimates of association. For example, smoking status was recorded only at baseline and information concerning age at initiation or cessation of smoking was not collected. Therefore, we could not examine smoking duration, timings of initiation and of cessation of smoking, or pack-years of smoking. Because of these limitations, in the analysis stratifying by baseline age ≥ 50 years versus < 50 years, assumptions were made that the majority of male smokers and female smokers born after 1940 began smoking at approximately age 18 years2–4 and female smokers born before 1940 began smoking at ages ranging from 20–32 years.4 If, in fact, an appreciable proportion of smokers did not begin smoking until much later than the assumed ages, this would bias the RRs toward 1. Conversely, if many smokers quit smoking after the baseline assessment, this also could reduce the RRs for the smoking categories. In addition, we did not have data regarding some important potential confounders such as dietary factors, physical activity, use of aspirin or nonsteroidal antiinflammatory agents, or CRC screening. Also, in the current study, CRC mortality rather than incidence was the endpoint. Because smoking is a strong risk factor for cardiovascular disease mortality and lung carcinoma mortality, participants who had CRC and smoked may have died of cardiovascular disease or lung carcinoma rather than CRC, resulting in an underestimation of the association between smoking and CRC. Finally, although some studies have suggested a stronger association between smoking and rectal carcinoma, the current study had too few cases of rectal carcinoma to perform an analysis.
The results of the current study provide additional evidence of an association between smoking and CRC mortality, particularly for women age < 50 years. These findings support the proposition that CRC be added to the list of tobacco-associated malignancies. Following this proposition, tobacco use prevention education programs could incorporate this finding as new information when presenting the harmful effects of tobacco use.
The authors thank Drs. Jeremiah Stamler and Philip Greenland for their scientific advice and review of this article.