Cigarette Smoking and Risk of Non-Hodgkin's Lymphoma—A Population-Based Case-Control Study

A remarkable increase in non-Hodgkin's lymphoma incidence in the latter half of the 20th century, not attributable to changes in known risk factors, sparked an intensive scientific search for unrecognized causes for this group of malignancies. Many studies have focused on the possible role of tobacco smoking, arriving, however, at conflicting results. Thus, whereas some studies have reported smoking to be associated with an increased risk of all types of non-Hodgkin's lymphoma combined (1-3), others have reported no unusual lymphoma risk among smokers (4-14). Various methodologic limitations may explain this inconsistency in results, including assessment of smoking status only at start of follow-up in cohort studies, the use of non-Hodgkin's lymphoma mortality rather than incidence as outcome, limited study sizes, and, possibly, gender-specific variation in the putative association between non-Hodgkin's lymphoma and smoking.

Furthermore, few studies include analyses stratified by non-Hodgkin's lymphoma subtypes and might thereby miss associations recognizing non-Hodgkin's lymphoma as a heterogeneous group of malignancies. Accordingly, in addition to non-Hodgkin's lymphoma subtypes varying in both histologic, molecular, and clinical characteristics as well as in prognosis (15, 16), there is mounting evidence that similar heterogeneity applies to their risk factors (17-21). With respect to smoking, several epidemiologic studies have suggested a positive association with non-Hodgkin's lymphoma subtypes, including B-cell high-grade non-Hodgkin's lymphoma (1, 2). Most consistent, however, associations have been found for follicular non-Hodgkin's lymphoma (22-27). The latter association is biologically particularly interesting because smoking may induce the chromosomal translocation t(14;18) (refs. 28, 29), which is found in 70% to 95% of follicular lymphomas (16).

These observations, consistent with the suspicion of etiologic heterogeneity, emphasize that etiologic studies must take histologic subtypes into consideration. We therefore tested the hypothesis that cigarette smoking increases the risk of certain lymphoma subtypes in a large Danish-Swedish population-based study of 3,055 cases of non-Hodgkin's lymphoma and 3,187 controls.

Between October 1999 and August 2002, a nationwide population-based case-control study of risk factors for malignant lymphomas was carried out in all of Denmark and Sweden. The Scandinavian Lymphoma Etiology study base encompassed the entire population, ages 18 to 74 years old, living in Denmark from June 1, 2000 to August 30, 2002, and in Sweden from October 1, 1999 to April 15, 2002. In Denmark, the study also included participants in a regional pilot phase beginning on November 1, 1999. Subjects with insufficient knowledge of the Danish or Swedish language, with a history of organ transplantation, HIV infection, or prior hematopoietic malignancy, were excluded from the study.

Cases were all patients with a first, newly diagnosed non-Hodgkin's lymphoma (ICD-10: C82-C85, C88.0, C91.3-5, C91.7), including chronic lymphocytic leukemia/small lymphocytic lymphoma (ICD-10: C91.1), according to the WHO classification (16). Newly diagnosed patients were identified through a rapid case ascertainment system organized for this study, consisting of a network of contact physicians from all departments where malignant lymphomas are diagnosed and treated (in total, 39 departments in Denmark and 118 in Sweden). Continuous collaboration with the six regional cancer registries in Sweden and the Danish National Pathology Registry (with an estimated coverage close to 100 percent; ref. 30) ensured complete reporting through the network. Controls were randomly sampled from the entire Danish and Swedish populations using continuously updated computerized population registers. Thus, a subset of controls was sampled every 6 months during the study period, frequency matched by gender and age (in 10-year intervals) within each country on the expected distribution of cases of non-Hodgkin's lymphoma.

All study participants completed (in person) a telephone interview including questions on height and weight, education, occupation, smoking habits, medical history, and other possible risk factors for malignant lymphoma. We were unable to blind the interviewers to case or control status, but interviewers were unaware of the specific hypotheses under study and instructed to treat cases and controls strictly the same. Participants were asked whether they had ever smoked tobacco daily for at least 1 year, and, if so, they were inquired about the type of tobacco they had been using (cigarettes/other tobacco products/both cigarettes and other tobacco products), and about current or former smoking habits. Concerning cigarettes, questions were asked about age at initiation, average daily consumption, and age at possible cessation.

The classification of the tumor material has previously been described in detail (30). Briefly, in Denmark, review of tumor material from cases was done within the Danish Lymphoma Group Registry (LYFO; ref. 31) and involved evaluation by expert hematopathologists. In Sweden, cases were histopathologically evaluated by six specially appointed senior hematopathologists/cytologists. All cases in both countries were classified according to the current WHO classification of hematopoietic and lymphoid tumors (16).

We evaluated the association between cigarette smoking (whether combined with other smoked tobacco products or not) and risk of non-Hodgkin's lymphoma by multiple logistic regression analysis. Cigarette smoking was assessed as never/ever, and ever cigarette smokers were further divided into former and current smokers. Ever smoking was defined as daily cigarette smoking for a minimum of 1 year and former smokers as those who had stopped smoking cigarettes at least 1 year before the interview. To assess a possible dose-response relationship between smoking and lymphoma risk, the data were analyzed by comparing age at initiation (≤14, 15-19, 20+ years), smoking intensity (1-9, 10-19, 20+ cigarettes per day), smoking duration (≤9, 10-29, 30+ years), and estimated lifetime exposure (≤149, 150-299, 300+ cigarette-years) in current smokers versus never smokers, and further, by comparing never smokers with former smokers, classified by time since cessation (1-9, 10-19, 20+ years). Two percent of the study participants (n = 119) were excluded from the analyses because their cigarette smoking exposure could not be uniquely characterized (due to a temporary one-time phrasing of an interview question, which made the distinction of cigarette smokers from smokers of other types of tobacco impossible).

Odds ratios (OR) were first calculated with adjustment for the matching variables (age in 10-year intervals, country, and gender) and then in multivariate analyses, with further adjustment for educational level (≤9, 10-12, 13+ years). The latter additional adjustment only changed a few estimates marginally, and therefore only multivariate estimates are shown. Trend tests were based on continuous measures of exposure as originally reported. Ninety-five percent confidence intervals (95% CI) were based on Wald tests. Effect modification by country, age group, or gender was tested by introducing an interaction term in the multivariate regression model. All significance tests were two sided.

Within the study population and time period, 3,055 patients (81% of all identified incident non-Hodgkin's lymphoma patients) and 3,187 controls (71% of potential controls) gave informed consent to participate in the study. Among eligible patients, early death (n = 279, 7%) was the main reason for nonparticipation, whereas unwillingness was the most common reason among controls (n = 718, 23%). Table 1 shows the characteristics of all 6,242 participants. Fifty-three percent of the controls and 54% of the non-Hodgkin's lymphoma patients reported ever having smoked cigarettes. Of these, 43% of the controls and 41% of the patients were current smokers 1 year before the interview.

Risk of all types of non-Hodgkin's lymphoma combined was not associated with cigarette smoking in multivariate analyses (Table 2). We found no evidence of heterogeneity between women and men nor between Danish and Swedish participants (data not shown). The risk of non-Hodgkin's lymphoma did not seem to vary between current and former smokers, by age at initiation of smoking, number of cigarettes smoked per day, duration of smoking, cumulative lifetime exposure to cigarette smoking, or by time since cessation among former smokers. Similarly, in subtype-specific analyses, cigarette smoking was not associated with increased risks of chronic lymphocytic leukemia, diffuse large B-cell lymphoma, or follicular lymphoma (Table 2). An exception from this pattern was a statistically significantly increased risk of T-cell lymphomas among current smokers as compared with never smokers.

We further analyzed the risk of non-Hodgkin's lymphoma subtypes, stratified by gender. In these analyses, currently smoking women were at a statistically significantly increased risk of follicular lymphoma (OR, 1.41; 95% CI, 1.04-1.92) compared with women who had never smoked (Table 3), whereas no such increased risk was observed among currently smoking men. This gender variation was statistically significant (P = 0.03, data not shown) using an interaction term instead of stratification. We found no evidence of any dose-response relationship between smoking and risk of follicular lymphoma (Table 3). Further stratification by age did not reveal any significant variation in risk estimates among the women (results not shown).

In T-cell lymphoma, we found a statistically increased risk of T-cell lymphoma in ever smokers (OR, 1.38; 95% CI, 1.01-1.88) and in current smokers (OR, 1.55; 95% CI, 1.07-2.25), when men and women were analyzed together (Table 2). However, gender stratification revealed smoking history in men as a risk factor for T-cell lymphoma, independent of the smoking status (ever smoker: OR, 1.67; 95% CI, 1.11-2.51; former smoker: OR, 1.63; 95% CI, 1.03-2.57; current smoker: OR, 1.73; 95% CI, 1.05-2.84). The risk estimates in women were not suggestive of a similar association (ever smoker: OR, 1.04; 95% CI, 0.64-1.70; former smoker: OR, 0.75; 95% CI, 0.39-1.45; current smoker: OR, 1.34; 95% CI, 0.77-2.32). However, the gender-specific estimates were not statistically significantly different using an interaction term (P = 0.35, data not shown). We found no evidence of any dose-response relationship between smoking and risk of T-cell lymphoma (Tables 2 and 3).

Based on a uniquely large material of 3,055 non-Hodgkin's lymphoma patients and 3,187 population controls, we found no association between cigarette smoking and the overall risk of non-Hodgkin's lymphoma. This observation is consistent with the majority of previous studies (4-14, 32-39), although not all as some studies have suggested an association between cigarette smoking and non-Hodgkin's lymphoma, especially in heavy smokers (1-3, 23, 40, 41). In our study, cigarette smoking was also not associated with increased risk of any major subtype of B-cell lymphomas, when men and women were analyzed together.

Stratified analyses indicated that smoking might be associated with an increased risk of follicular lymphoma in women and of T-cell lymphomas in men. For T-cell lymphoma, the only other indication of an association with smoking stems from a small scaled Italian study encompassing only eight cases of T-cell lymphoma, and reporting an OR of 25.84 (95% CI, 1.95-342.17) in heavy smokers (42).

An increased risk of follicular lymphoma associated with smoking has previously been suggested. Stagnaro et al. (23) reported ORs of 2.1 (95% CI, 1.4-3.2) and 1.8 (95% CI, 1.1-3.0) for follicular lymphoma among ever smokers of blond and mixed tobacco, respectively, reanalyzing an Italian case-control study from 2001 (27) including 1,450 cases of non-Hodgkin's lymphoma and 1,779 controls. A French case-control study including 180 cases of non-Hodgkin's lymphoma and 360 matched controls observed a more than 3-fold increased risk for follicular lymphoma in current smokers (22), and an American cohort study with 674 cases of non-Hodgkin's lymphoma found a relative risk of 1.9 (95% CI, 1.2-2.9) for follicular lymphoma among former smokers (24). Consistent with our findings, the association between smoking and follicular lymphoma in women has also been reported previously. Parker et al. (26) reported a relative risk of 2.3 (95% CI, 1.0-5.0) for follicular lymphoma in currently smoking women in a cohort study including 200 women with non-Hodgkin's lymphoma. Similarly, in the aforementioned original case-control study (27), Stagnaro et al. reported an increased risk of follicular lymphoma among smokers (OR, 1.8; 95% CI, 1.3-2.7), with women, however, apparently being at particularly high risk (OR, 2.3; 95% CI, 1.4-3.8) as compared with smoking men (OR, 1.3; 95% CI, 0.69-2.3). Further, Morton et al. (25) in an American case-control study with 601 cases of non-Hodgkin's lymphoma observed an increased risk of follicular lymphoma in women with a cumulative lifetime exposure of 34 pack-years or greater (OR, 1.8; 95% CI, 1.1-3.2).

Although the gender variation is not easily explained, other observations support a potential role for smoking in lymphoma development. Specifically, cigarette smoking has been associated with several changes in immune function, such as alterations in T-cell subsets (43) and a decrease in the proportion of circulating natural killer cells (44). Although incompletely understood, immune dysfunction is the most well-established risk factor for malignant lymphoma (45, 46). With respect to follicular lymphoma in particular, the chromosomal translocation t(14;18) can furthermore be detected in the peripheral blood lymphocytes of apparently healthy smokers, with increasing frequency in heavy smokers (28, 29). This particular translocation, which up-regulates the antiapoptotic BCL-2 gene and leads to a prolonged survival and increased chance of clonal expansion of B cells, is present in the malignant cells in the vast majority (70-95%) of follicular lymphomas (16). We did not have access to information on t(14;18) status of the follicular lymphomas included in the present study, but note with interest that one study of male lymphoma patients indicated that an association between cigarette smoking and follicular lymphoma may be more pronounced for t(14;18)-positive non-Hodgkin's lymphoma (OR, 1.7; 95% CI, 0.9-3.3) than for t(14;18)-negative non-Hodgkin's lymphoma (OR, 1.1; 95% CI, 0.7-1.7), albeit the difference was not formally statistically significant (47).

Inherited traits may interact with environmental exposure and influence the individual susceptibility to malignant disease. For cigarette smoking, polymorphisms in the glutathione S-transferase (GST) gene, which encodes for polycyclic aromatic hydrocarbons involved in the metabolism of tobacco carcinogens, have been shown to affect the risk of lung cancer (48, 49). A potential etiologic role of the polymorphism has also been suggested for non-Hodgkin's lymphoma in some studies (50-53). Furthermore, it is noteworthy that GST polymorphisms have been identified as a prognostic factor in pediatric non-Hodgkin's lymphoma (54) and as a risk factor for chronic lymphocytic leukemia (55). The latter association is interesting as several epidemiologic studies have shown that patients with chronic lymphocytic leukemia have an increased risk of smoking-related malignancies such as cancers of the lung, urinary bladder, and larynx (56-58). However, there were no positive associations between smoking and risk of chronic lymphocytic leukemia in our analyses, which is consistent with previous studies (11, 23, 24). We cannot, however, exclude that the observed associations for follicular lymphomas and T-cell lymphomas reflect an even stronger association in genetically predisposed individuals, as discussed for several smoking-related malignancies (59-61).

Although positive findings for follicular lymphoma and T-cell lymphoma were consistent with a priori hypotheses based on the literature and, moreover, were observed in analyses restricted to either Danish or Swedish participants, the possible causal implications should be interpreted cautiously. Specifically, for neither lymphoma subgroup evidence could be established of a dose-response relationship between intensity of cigarette smoking and risk. Despite the relatively large size of our initial subtype groups, the stratification by gender rendered the studied case numbers low. Even more given the large number of comparisons made in our study, it cannot be ruled out that the observed associations with follicular lymphoma and T-cell lymphoma are chance findings.

The strengths of our study lie in its uniquely large size, the population-based design, the rapid case ascertainment of incident cases of non-Hodgkin's lymphoma, and the uniform classification of non-Hodgkin's lymphoma subtypes. One limitation of our study remains, however, that exposure data were based on self-reported smoking habits, with the probability of exposure misclassification. This applies in particular to the assessment of past exposure history and the distinction between current and former smoking status and of lifetime dose. It is likely that this bias would affect cases and controls equally and would lead to an attenuation of any true differences between cases and controls, and thus to underestimation of true effects. Importantly, however, two reviews of studies comparing self-reported data of tobacco smoking to biochemical markers hereof showed a high degree of reliability and accuracy in the classification of ever/never and former/current smokers (62, 63). Misclassification of former/current smoking cases could also have resulted from the fact that a number of non-Hodgkin's lymphoma patients were interviewed more than 1 year after diagnosis. However, supplementary analyses excluding these cases did not change the observed estimates (data not shown).

Another possible bias in our case-control study design could result from differential recollection of smoking data in patients and controls. However, the hypothesis of an association between cigarette smoking and risk of non-Hodgkin's lymphoma or one of its subtypes is not well known, and we therefore consider it unlikely that patients would have underreported their smoking habits.

In conclusion, we found little evidence of an association between cigarette smoking and risk of non-Hodgkin's lymphoma overall. Although our results were compatible with an increased risk of follicular lymphoma in female smokers and of T-cell lymphoma in male smokers, we were unable to show a dose-response relationship, leaving limited support for a causal association.

We thank project coordinators Charlotte Appel (Statens Serum Institut) and Leila Nyrén (Karolinska Institutet), as well as Kirsten Ehlers at LYFO. We also thank cytologist Edneia Tani (Karolinska Institutet) and pathologists Måns Åkerman (Lund University Hospital), Åke Öst (Medilab, Stockholm), and Christer Sundström (Akademiska Hospital, Uppsala) for extensive review of tumor material. We are deeply indebted to all of the contact doctors and nurses in Denmark and Sweden who participated in our rapid case ascertainment system.