Introduction

Non-Hodgkin lymphoid neoplasms (NHL) are the sixth most commonly diagnosed cancer in the United States [1]. The incidence rate of NHL has nearly doubled since the mid-1970s [2]. Established risk factors for NHL are related to either immunological function or infection such as HIV/AIDS; however, these risk factors explain only a small proportion of NHL cases [3]. Inconsistencies among studies examining associations of other risk factors with NHL have been attributed, at least in part, to the heterogeneous nature of the disease, and potential differences in associations among subpopulations including gender [3].

Cigarette smoking has multiple effects on the immune system that might influence lymphoma risk including direct carcinogenesis, immunosuppression, and increased leukocyte production [4, 5]. However, results of studies of the relationship between cigarette smoking and risk of NHL are inconsistent. A review of 22 studies in 2001 [6] found no overall association between NHL and smoking, although it called for further study of histologic subtypes of NHL, and for stratification by gender. Since then, several large case–control [7–9] and cohort studies [10–14] investigated the association of cigarette smoking with risk of NHL subtypes. In several of those studies, smoking was associated with an elevated risk of follicular lymphoma [7, 9, 14]. However, other studies found no associations [8, 11] or even reduced risk [10, 12] with follicular lymphoma. The efforts of some studies to stratify analyses of smoking by both NHL subtype and gender were constrained by sample size [9, 11, 13], while other studies did not present results stratified by subtype and gender [7, 8, 10, 12]. The limited data presented on NHL subtypes stratified by gender make it difficult to discern whether there are no differences by gender, or whether individual studies are simply underpowered. In order to further clarify this relationship, we conducted a prospective cohort analysis of NHL subtypes, gender, and cigarette smoking in a population of older U.S. men and women.

Materials and methods

Study population

Subjects in this analysis were selected from the Cancer Prevention Study II (CPS-II) Nutrition Cohort, a prospective study of cancer incidence and mortality in 184,188 men and women from the United States, described in detail elsewhere [15]. Briefly, the Nutrition Cohort is a sub-cohort of the approximately 1.2 million subjects in CPS-II, a prospective study of mortality established by the American Cancer Society in 1982 [16]. Participants in the larger study were recruited nationally and completed a four-page questionnaire at enrollment. Subjects between the ages of 50–74 years in 1992 were recruited from 21 states with population-based state cancer registries to participate in the Nutrition Cohort. Participants completed a 10-page mailed questionnaire which included information on demographic, medical, behavioral, environmental, occupational, and dietary factors. Follow-up questionnaires were sent to cohort members in 1997, 1999, 2001, 2003, 2005, and 2007 to ascertain cancer diagnoses. Responses to all surveys were received from at least 89 % of living participants after multiple mailings. The current study encompasses follow-up through the 2007 survey. All aspects of the CPS-II Nutrition Cohort study have been approved by the Emory University Institutional Review Board.

Analytic cohort

These analyses excluded subjects from the CPS-II Nutrition Cohort who were lost to follow-up (n = 6,276), reported a personal history of cancer other than non-melanoma skin cancer at baseline in 1992 (n = 22,860), reported a diagnosis of a hematopoietic cancer in the first survey interval that could not be verified (n = 67), or were missing smoking status at baseline in 1992 (n = 2,027). The final analytic cohort includes 152,958 participants including 72,752 men and 80,206 women.

Measures of cigarette smoking

In 1992, never smokers were those participants who answered “No” to the question “Have you smoked at least 100 cigarettes in your entire life?” For participants that answered “Yes” to the previous question, current and former smoking was based on the follow-up question “Do you smoke cigarettes now?” Information on the age began smoking, average number of cigarettes smoked daily, and number of years smoked was collected from both current and former smokers. Additionally, former smokers were asked the age at which they last quit smoking for the calculation of time since quitting. Smoking status was updated on each follow-up questionnaire. For each follow-up questionnaire, it was determined whether a subject was a current smoker. If not, subjects that were never and former smokers on the previous questionnaire maintained their status, and current smokers from the previous questionnaire were converted to former smokers. If the subject reported current smoking, never and former smokers were converted to current smokers, while previous current smokers maintained their status. Current smokers were assigned cigarettes per day based on the most recent questionnaire. Current smokers and former smokers that maintained the same smoking status between intervals accrued additional years of smoking or time since quitting, respectively, equivalent to the length of the interval. Current smokers that quit smoking in the interval were assumed to have quit at the midpoint (except on the 1997 survey where more detailed time since quit was available). Half of the interval was added to the years smoked, the remaining half was considered the time since quitting. Never smokers and former smokers that began smoking between surveys were assumed to have started at the midpoint and those years were added to any previously reported years smoked (for never smokers this was 0 years). Pack years were not calculated, as the above measures of smoking have previously been shown to be better predictors of smoking-related cancers in this cohort [17].

Case ascertainment

This analysis includes 1,926 cases (1,090 men and 836 women) with NHL diagnosed between the date of enrollment and 30 June 2007. The majority of cases were identified by self-report on the follow-up surveys; of these, 1,140 were verified by medical record abstraction, and 402 were verified by linkage with state cancer registries. There were 41 cases identified through the self-report of another cancer that were subsequently verified as NHL during medical record abstraction or linkage with state cancer registries. An additional 343 cases were identified as deaths through automated linkage of the entire cohort with the National Death Index and subsequently verified by linkage with the state cancer registries.

Histologic subtypes were defined using the Interlymph Pathology Working Group classifications [18], based on the 2001-revised WHO classification of tumors of hematopoietic and lymphoid tissues [19]. Using the International Classification of Disease for Oncology, Second and Third Edition (ICD-O-2 and ICD-O-3), these subtypes include: diffuse large B-cell lymphoma (DLBCL), chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), follicular lymphoma, multiple myeloma, other B-cell lymphomas, and T-cell lymphoma.

Statistical analyses

Smoking was classified in terms of smoking status (never, former, current), age began smoking (>20, 19–20, 17–18, and <=16 years), number of cigarettes smoked per day (<20, 20, and >20), and years smoked (<40, 40−<50, and 50+) in current smokers and years since quit smoking (30+, 20–<30, 10–<20, <10) in former smokers. All smoking variables were updated during follow-up to maintain a clean referent group of lifelong never smokers, update current smoking status, update the number of cigarettes per day smoked, and accumulate additional years of smoking and years since quitting. Descriptive statistics were calculated using distributions of categorical variables and means of continuous variables within strata of smoking status at baseline.

Person-years of follow-up for each participant were calculated as the amount of time from completion of the CPS-II Nutrition Cohort questionnaire in 1992/1993 to date of: (1) diagnosis of NHL; (2) diagnosis of other cancer; (3) death occurring between the last returned survey and next mailed survey; (4) return of last questionnaire (i.e., in 1997, 1999, 2001, 2003, 2005 or 2007); or (5) last questionnaire the participant was known to be cancer free, if they reported a hematopoietic cancer that could not be verified. Extended Cox regression [20] was used to compute multivariable adjusted rate ratios (RR) and 95 % confidence intervals (CI) for the association between time-dependent smoking variables and NHL, and NHL subtype incidence. In all analyses, the reference group was those who reported never smoking. Models were run in men and women separately, as well as combined. All models were stratified on age, and additionally adjusted for gender (in the analyses of men and women combined), education (<high school, high school, some college, college graduate), family history of hematopoietic cancer (no, yes), body mass index (continuous kg/m2), height (gender-specific quintiles), physical activity (continuous metabolic equivalents of energy in h/week), and alcohol use (continuous drinks/day). Variables for missing status were included for all covariates. Covariates were selected based on the literature and previous work in this cohort. Trend tests were calculated using the median values of each category for the cigarettes per day, and years smoked variables. For variables with no identifiable median in at least 1 category, trends were calculated using a continuous ordinal variable indicating a 1 unit change per category.

Effect modification by gender was assessed by computing p values for multiplicative interactions using likelihood ratio tests comparing Cox multivariable models with and without cross-product terms between each smoking and gender variable. For NHL subtypes with no significant difference by gender, models were run in men and women combined for smoking status, age at initiation, and time since quit in former smokers. All analyses were performed using Statistical Analysis Software (SAS version 9.2; SAS Institute Inc., Cary NC).

Results

Nearly 70 % of male and 45 % of female participants in this study were active cigarette smokers at some point in their lives, although only approximately nine percent in each group continued to smoke at the time of enrollment (Table 1). The smoking habits of former smokers were similar to those of current smokers in men and in women, except that current smokers had smoked on average 17–18 years longer than former smokers. Among former smokers, the average number of years since quitting smoking was 21.9 years (SD 12.31) in men and 20.1 years (SD 12.16) in women. In both men and women, current smokers were less likely to be college educated and physically active, but more likely to report drinking 2 or more drinks of alcohol per day than never and former smokers. A smaller proportion of the currently smoking women were obese when compared to the proportion of obesity in never or former smokers at baseline.

Table 1 Characteristics of the CPS-II nutrition cohort by smoking status at baseline and gender
Full size table

Smoking status was not static during follow-up. Almost 40 % of the current smokers quit smoking over the course of follow-up. On average, they had not been smoking for 7 years by the end of follow-up. Current smokers, who continued to smoke, had smoked for an additional 7 years on average during follow-up. Consequently, 40 % of these moved into a higher category of years of smoking when smoking during follow-up was considered. Only 1 % of the never smokers began smoking. Similarly, only 1 % of the former smokers reverted to smoking during follow-up. On average, the former smokers had not smoked in over 30 years by the end of follow-up. These changes in smoking status over time were similar in both men and women.

The relationship between current smoking and NHL differed by gender (p-interaction, 0.02). In men, there was no association between current smoking and risk of NHL, whereas in women, there was a 37 % higher risk of NHL for current smokers compared to never smokers (Table 2). Among current smokers, there was no evidence of a dose–response relationship of number of cigarettes smoked per day with risk of NHL in men or in women. However, there was a statistically significant dose-related association with years smoked and NHL incidence in women but not in men. The association for women culminated in a twofold higher risk for women who smoked 50 years or more compared to never smokers. Among former smokers, there were no clear dose–response relationships of years since quitting smoking with risk of NHL in men or women. Younger age at smoking initiation was associated with a higher risk of NHL in women, but not men. The risk of NHL was 11 % higher among former compared with never smokers in men and women combined.

Table 2 Multivariable adjusted rate ratios of Non-Hodgkin lymphoid neoplasms and smoking stratified by gender
Full size table

Gender differences also were observed when evaluating smoking status in relation to NHL subtypes (Fig. 1, Supplemental Table 1). In women, there was a statistically significantly higher risk of follicular lymphoma (RR 2.13, 95 % CI 1.20–3.77), and CLL/SLL (RR 1.75, 95 % CI 1.03–2.96) associated with current smoking compared to never smoking. In men, there was no association with follicular lymphoma or CLL/SLL. The interaction between smoking status and gender for follicular lymphoma was statistically significant (p-interaction, 0.02), but not for CLL/SLL (p-interaction 0.17). However, there was a significant difference by gender for the association between age at initiation and risk of CLL/SLL (p-interaction 0.02). Associations of smoking with risk of other NHL subtypes were not significantly different in men and women; thus, to increase statistical precision, men and women were combined (Table 3). In these analyses, there was a twofold higher risk of T-cell lymphoma associated with current smoking (RR 2.30, 95 % CI 1.18–4.47), and a statistically significant dose–response with time since quitting in former smokers (p-trend 0.03). Conversely, there was a lower risk of DLBCL associated with current smoking in men and women combined (RR 0.32, 95 % CI 0.15–0.69).

Fig. 1
figure 1

Plots of multivariable adjusted rate ratios comparing never smokers with former smokers and current smokers in seven different NHL subtypes stratified by gender. The NHL subtypes shown are: all B-cell lymphomas, diffuse large B-cell lymphoma (DLBCL), chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), follicular lymphoma, multiple myeloma, other B-cell lymphomas, and T-cell lymphomas

Full size image
Table 3 Multivariable adjusted rate ratios of selecteda non-Hodgkin lymphoid neoplasm subtypes and smoking in men and women combined
Full size table

Discussion

In the present study, current smoking was associated with an increased risk of NHL in women but not in men. The association between smoking and risk of NHL appeared to differ by histologic subtype of disease. Compared to never smoking, current smoking was associated with higher risk of follicular lymphoma and CLL/SLL in women but not men. In analyses of men and women combined, there were positive associations of current smoking and, among former smokers, shorter time since quitting smoking with risk of T-cell lymphoma. Conversely, there was an inverse association between current smoking and risk of DLBCL in men and women combined. There were no associations of smoking habits with risk of other NHL subtypes in men, women, or men and women combined.

A critical review [6] and four recently published prospective cohort studies that examined smoking in relation to NHL risk in both men and women combined were null [10–13]. Few studies published to date have assessed the association between smoking status and risk of NHL in men and women separately. While most show no association in men or women [7–13, 21], results of the California Teachers Study showed an association between smoking intensity and all NHL in women after excluding never smokers that were exposed to the smoke of others [14]. In CPS-II, additional adjustment for passive smoking exposure did not meaningfully alter our findings (data not shown). These results do not preclude passive smoking from having a confounding effect in other studies. In CPS-II, the majority of participants reported exposure to smoke of others and there was no difference in the distribution of passive smoking between never, former, and current smokers. In addition, the heavy smoking of this cohort would outweigh any likely smaller association with passive smoking.

Several studies found that the relationship of smoking with specific NHL subtypes might differ by gender [9, 11, 22–24]. Indeed, our study, like many other studies [7, 9, 11, 21, 24, 25], found a higher risk of follicular lymphoma among current smokers, but in our study this association was limited to women. The large Interlymph consortium of case–control studies identified a significant increased risk of follicular lymphoma associated with current smoking and number of years smoked which did not differ by gender [7]. However, two of the studies included in Interlymph [23, 26] and four additional studies [9, 11, 21, 24] found increased risks only in women. In the Iowa Women’s Health Study [21], current smoking was associated with a doubling of risk of follicular lymphoma. In the European Prospective Investigation into Cancer and Nutrition [11], more recent quitting in formerly smoking women was associated with increased risk, and in the California Teachers Study [14], women exposed to over 40 years of passive smoking had a nearly threefold higher risk of follicular lymphoma compared women with no passive smoking exposure. While three other prospective studies reported no gender differences in the association between smoking and follicular lymphoma incidence [10, 12, 25], the overall evidence for an association in women is becoming strong. No other B-cell subtype has been consistently associated with smoking in either men or women [7–12, 14, 21, 24, 25, 27–37].

At least two published studies showed statistically significantly higher risks of T-cell lymphomas associated with current smoking [9, 38], and similar but non-significant associations were observed in two other studies [12, 37]. The Interlymph study [7] identified a significant trend of increasing risk with recent quitting in former smokers and a significant trend of increasing risk with greater number of years smoked for peripheral T-cell lymphoma only. In our study, there were too few cases of specific T-cell subtypes to examine associations with smoking separately. Studies of T-cell lymphoma and smoking have been small with the largest studies including only 200 cases [7], making it difficult to rule out chance findings. However, even with limited statistical power, several have identified significant associations suggesting continued research on smoking and T-cell lymphomas is warranted.

The biologic mechanism underlying a potential association between smoking and risk of lymphoma subtypes is unclear, but intriguing possibilities exist. A higher prevalence of t(14;18) translocations has been shown in peripheral blood of heavy smokers compared with nonsmokers [39]. Consistent with our finding of a positive association between smoking and risk of follicular lymphoma in women, t(14;18) translocations are present in a very high proportion of follicular lymphomas(80–90 %) relative to other lymphoma subtypes [40]. However, two recent case–control studies of chromosomal abnormalities found evidence of an increased risk of t(14;18)-negative NHL with smoking in women [41, 42]. We can also speculate that smoking is acting on hormone levels, which could explain the gender differences observed in this study. In the past, smoking was thought to have anti-estrogenic effects in women; however, recent evidence suggests positive associations of smoking with circulating levels of estrogens and androgens [43]. Increased estrogen has been shown to reduce pre-cursor B-cell differentiation and proliferation in mice leading to an accumulation of non-cycling cells [44], possibly providing a source of cells for the development of NHL in smoking women. For commonly asymptomatic NHL such as follicular lymphoma and CLL/SLL [45], the observed gender difference in this study could be due to differential healthcare utilization in men and women leading to undiagnosed cases in smoking men, or simply chance. Heavy smoking also impacts T-cell biology including retention of CD8+ T-cells in the lungs, increased clonal expansion of multiple T-cell types and compromised T-cell responsiveness [5]—potentially affecting T-cell lymphoma development. Cigarette smoking could also indirectly lead to an elevated risk of lymphoma by increasing susceptibility to infectious risk factors such as human T-cell leukemia virus-1 and Epstein Barr virus.

The strengths of this study include its large size, prospective design, and availability of detailed lifetime smoking history. With 1,926 verified NHL cases, this is one of the largest prospective cohort studies of incident NHL and smoking to date. The large case numbers enabled us to stratify by gender and examine relationships by subtypes within gender. The long history of smoking in our current smokers allows the identification of associations that may not become apparent until after many decades of smoking. Despite the large sample size, we are still unable to finely examine dose of cigarette smoking in the lymphoma subtypes while stratifying by gender. A pooled analysis of smoking and NHL in prospective studies will likely be necessary to adequately examine the dose relationships by subtype and gender. A second limitation is that there were no measures of infectious or immune-related exposures that could be important confounders in this study, or act in the causal pathway through the weakened immune systems of smokers. However, at this time, only a few rare forms of NHL have been strongly linked to infectious agents [46, 47], making large effects of uncontrolled confounding unlikely.

In conclusion, this large cohort study supports an association with cigarette smoking and increased risk of follicular lymphoma in women. Future analyses should focus on examining the histologic subtypes of NHL in men and women. Large consortia of prospective studies with the ability to pool data on smoking status and histology are required to examine questions of duration, dose, and time since quitting which are essential to establishing a relationship. With over 20 % of adult Americans continuing to smoke in 2009 and no decrease in the previous 5 years [48], it is important to better understand the role of cigarette smoking in NHL etiology.