The present large cohort study of female public school professionals contributes to the current literature on the association between smoking and NHL risk in 3 important ways. First, among lifetime never smokers, we observed a statistically significantly increased risk of NHL for those exposed to high levels of passive smoking. Although results were based on small numbers, we also found that follicular lymphoma was associated with long duration, high intensity, and high level of intensity-years of lifetime passive smoking exposure from household, workplace, and social settings combined. Second, the present study demonstrated that the estimated relative risks of NHL associated with smoking were strengthened when women with household passive smoking were excluded from the reference category. Thus, previous studies that did not account for passive smoking in the reference category might have underestimated the adverse effects of smoking on NHL risk. Third, our data further strengthened the evidence regarding the increased risk of follicular lymphoma among ever smokers that was observed in previous studies (4
), and particularly in women (4
The few studies in which a dose-response relation between smoking and increased NHL risk was observed were mainly case-control studies (24
). However, the recall bias, selection bias, and survival bias inherent to case-control studies generally mitigated their findings. Results of 6 of the 7 cohort studies for which data have been published showed no association between smoking and NHL risk (1
); only the CPS-II cohort study showed an increased risk of NHL associated with smoking (30
). Our CTS data showed that lifetime cumulative exposure to passive smoking was highest among those born from the 1920s through the 1940s (40
). Interestingly, participants in the 4 US cohort studies that evaluated the dose-response effects of smoking on NHL risk were born during this period (birth year range, 1917–1946) (9
). If the referent category of never smokers in these studies indeed included a large proportion of passive smokers, this misclassification of overall smoking exposure may explain the lack of association detected in many of the cohorts.
In the CPS-II study, we observed a much stronger effect of smoking on NHL risk in women than in men (30
). Interestingly, investigators in all cohort studies that presented results separately for women reported an increased risk of follicular lymphoma among smokers (14
), whereas the findings of studies presenting results for men and women combined were mixed, showing both increased risk (8
) and decreased risk (9
) of follicular lymphoma. Results from the Third National Health and Nutrition Examination Survey showed that, between 1988 and 1991, 43.5% of US men and 32.9% of US women who were not tobacco users reported exposure to passive smoking at home or at work. Further, among adults 20 years of age or older, men had significantly higher mean serum cotinine concentrations than did women (41
). The passive smoking experience of CPS-II participants is likely therefore more representative of the national data than that in other studies because it collected smoking information in 1992 from men and women aged 50–74 years in 21 states. Thus, it is plausible that both the prevalence and intensity of passive smoking in the reference category of never smokers were higher for men than for women in CPS-II. Consequently, the higher level of misclassification of passive smoking in the reference category of nonsmokers among men would be more likely to bias the results toward the null, potentially accounting for the observed gender difference in the association between smoking and NHL risk.
Although the association between smoking and NHL risk became stronger when household passive smoking was excluded from the reference category, we note that our relative risk estimates for smoking may still have been underestimated because we were unable to exclude women who were exposed to passive smoking outside of the household, such as in the workplace or in social settings. However, based on data from our second questionnaire, we found that only 15% of women had either workplace or social exposure to passive smoking but no household exposure. Therefore, we expect that the misclassification effect of including such women in the reference category for the analyses of active smoking exposure was limited.
Potential mechanisms of importance include those associated with active smoking, such as initiation of lymphomagenesis by carcinogens (34
), in addition to altered immune function (34
), alterations in T-cell subsets, and a decrease in circulating natural killer cells (42
). Carcinogens, including benzene and many others, have been detected in passive smokers (34
). Animal studies, such as those in mice, have shown that concentrations of certain cytokines, including interleukin 4 and interleukin 10, are significantly higher in smoke-exposed animals than in nonexposed animals, a finding that could have relevance to the early events in carcinogenesis, in which an inflammatory response often precedes mild hyperplasia (44
). Moreover, it has been reported that sidestream smoke, which constitutes approximately 85% of passive smoking, might be more toxic than mainstream smoke (45
The chromosomal translocation t(14;18) has been hypothesized by some researchers to be a potential biologic mechanism for the association between follicular lymphoma and active smoking. This somatic mutation joins the immunoglobulin heavy chain gene on chromosome 14 with the bcl-2
gene on chromosome 18 and results in overexpression of bcl-2 protein, which inhibits apoptosis (46
). Both increasing age and smoking have been associated with t(14;18) translocation (47
). Our reported association between passive smoking and follicular lymphoma risk suggests that t(14;18) might also be a relevant mechanism for follicular lymphomas associated with passive smoking and warrants further investigation.
Our study had several strengths. First, our study population was optimal for studying NHL risk associated with passive smoking because 66% of all participants were never smokers. Further, the extensive evaluation of lifetime exposure to passive smoking from household, workplace, and social settings offered a detailed assessment of exposure in a way that no other large-scale cohort study has been able to do. Second, our study’s prospective design circumvented potential problems with the recall bias, selection bias, and survival bias that are common to case-control studies. Finally, the CTS utilized detailed follow-up procedures and had virtually complete ascertainment of cancer outcomes.
The main limitations of our study were the small number of smokers, particularly current smokers, and the small number of NHL cases available for subtype analyses, which resulted in a lack of precision in relative risk estimates. Another limitation was that our results could not be generalized to men, and as a group, CTS participants might not be representative of the general population of women because all CTS participants at some time in their lives have worked as public school professionals (primarily as teachers). Finally, the measures of smoking and passive smoking were derived from self-reports rather than from biologic measures. However, the reliability of respondents in recalling passive smoking history, especially for spousal and parental sources, has been demonstrated to be high (49
Our results are consistent with previous studies on the association between smoking and increased risk of follicular lymphoma (4
). Our data further support an association between passive smoking and NHL (especially follicular lymphoma) risk. Future research to clarify the association between smoking and NHL risk should take into account exposure to passive smoking.