Tobacco Use, Body Mass Index, and the Risk of Leukemia and Multiple Myeloma: A Nationwide Cohort Study in Sweden

Leukemias are a heterogeneous group of malignancies arising from cells of the hematopoietic system. Modern classification distinguishes between four main leukemia subtypes: acute lymphocytic leukemia (ALL), acute myelogenous leukemia (AML), chronic lymphocytic leukemia (CLL), and chronic myelogenous leukemia (CML). CLL belongs to B-cell lymphomas (1). The etiology of leukemia has not yet been fully elucidated. The most well established risk factor is ionizing radiation (2, 3), such as therapeutic and diagnostic radiation (4, 5). Infection with human T-cell lymphotrophic virus, type I is a well-established factor of importance for the development of adult T-cell leukemia/lymphoma (6), and exposure to benzene an established risk factor for AML (7). Tobacco smoking is today an established risk factor for AML (8).

Multiple myeloma (MM) evolves from the B-cell line of the lymphatic system producing altered plasma cells that infiltrate the bone marrow. The great majority of findings related to tobacco as a causative agent of MM have not shown any association (9–16). However, one study has reported a positive association (17).

Studies that have addressed the relation between use of smokeless tobacco products and hematopoietic malignancies are very limited and have in general not revealed a positive association (12, 18). Several reports on body mass index (BMI) and leukemias and MM suggest that there may be a weak positive association (19–22). However, there is no conclusive evidence as other studies have failed to confirm such a relationship (23, 24).

Given the conflicting results in previous studies and the increasing concern about the effect of lifestyle factors on the global burden of disease, we conducted a cohort study on Swedish construction workers with the aim of investigating the effect of tobacco use and BMI on the incidence of leukemia and MM.

Subjects. From 1969 to 1992, the Construction Industry's Organization for Working Environment, Safety and Health (25) provided outpatient medical services to all blue and white collar employees within the building industry throughout Sweden. On average, each cohort member underwent 2.6 health checkups during the period. Data from these health checkups were then put together in a computerized central registry (25).

Exposure information. From 1971 until 1975 and from 1978 until 1992, information on tobacco use, BMI, as well as other factors was acquired by a self-administered 200-item questionnaire accompanied by face-to-face interviews by a nurse. In 1975 to 1977, no exposure information on tobacco use was collected. The quality of the smoking data has been reviewed previously (26).

Definition of the cohort. The original cohort consisted of 361,280 individuals. A total number of 24,899 (7%) was excluded due to female gender (n = 17,458), cancer before entry (n = 1,229), incorrect national identification number (n = 388), missing information on BMI (n = 3,032), and inconsistencies during record linkages (n = 2,792). A total of 336,381 individuals was considered eligible for data analyses. A minor part of data investigating tobacco use only have been published previously (12). However, the follow-up was terminated in 1991 and leukemia subtypes were not investigated explicitly in that report.

Follow-up. The unique 10-digit national registration number assigned to all Swedish residents is an inimitable personal identifier and was used for follow-up by linkage to the National Causes of Death Registry, Migration Registry, and Cancer Registry. The essentially complete Swedish Cancer Registry initiated in 1958 codes malignant neoplasms according to the International Classification of Diseases.

Each worker of the complete cohort was followed from date of entry (first visit) until emigration, death, date of cancer diagnosis, or December 31, 2004, whichever occurred first, for a total of 7,475,628 person-years of observation.

Statistical analyses. Information on tobacco smoking, snuff use, and BMI was restricted to that collected at first visit. We used Cox proportional hazards regression model to estimate relative risks [incidence rate ratio (IRR)] together with corresponding 95% confidence intervals (95% CI) adjusted for attained age and BMI. Because age is an important risk factor for almost all cancer types, we used attained age (in years) as the time scale. Therefore, all estimates were implicitly adjusted for the attained age. Analyses of smoking status (current/former) were restricted to never snuff dippers and adjusted for attained age and level of BMI. Pure use of diverse smoking tobacco products and daily amount of tobacco smoking, respectively, were analyzed in the same manner. Tests for trend for amount of smoking tobacco were conducted using the median value of different strata as the score.

We used four categories for classification of BMI: ≤18.5 kg/m² (underweight), 18.6 to 24.9 kg/m² (normal weight), 25.0 to 29.9 kg/m² (overweight), and ≥30.0 kg/m² (obese). BMI analyses were adjusted for attained age, snuff use, and daily amount of tobacco smoking. Intercooled STATA Release 9.2 was used for all statistical procedures.

This study was approved by the institutional review board at Karolinska Institutet.

At baseline, the cohort members had a mean age of 34.3 years (range, 14–82 years) and were followed up for an average of 22.2 years (range, 0–33.5 years). About 56,179 workers (16.7%) were past smokers of one or several types of tobacco products (cigarette, pipe, cigar) and 137,311 (40.8%) were current smokers at entry into the cohort. Moreover, 29% of the cohort members were pure cigarette smokers, 12% were pure snuff dippers, 5% were pure pipe smokers, and 0.6% were pure cigar smokers. Mixed users consisted 23% of the study population. A total number of 4,247 (1.3%) had a BMI classified as underweight at time of enrollment, 210,081 (62.5%) were of normal weight, 105,793 (31.5%) were overweight, and 16,260 (4.8%) were obese. During follow-up, we identified 47 individuals with ALL, 224 with AML, 101 with CML, and 520 individuals with MM.

We did not detect any association between current or former smoking and the risk of ALL (Table 1). A 50% increased risk of AML was observed for current smokers (IRR, 1.50; 95% CI, 1.06–2.11). Current smokers and ex-smokers did not have an increased risk of CML or MM.

Smokers of more than 20 g tobacco per day had a 2-fold, although not statistically significant, excess risk of ALL (Table 2). No trend in risk of AML, CML, and MM was apparent for smoking intensity.

Exclusive use of cigarettes, pipe, snuff, or mixed use was not linked to a greater risk of ALL, AML, CML, or MM (Tables 2 and 3).

We did not reveal any apparent effect by having a BMI classified as overweight or obese on the risk of ALL, AML, CML, or MM compared with individuals of normal weight (Table 4).

In this cohort study, including more than 330,000 male construction workers in Sweden, we observed an increased risk of AML in current smokers and an indicative association between amount of smoking and risk of ALL. The risk of CML and MM was not associated with tobacco smoking. Snuff use was not associated with the risk of leukemia or MM, and BMI did not seem to modify the risk of any of the neoplasias under study.

The IARC have concluded that smoking is weakly associated with an increased risk of leukemia but that there is insufficient support for an increased risk of lymphoid leukemia linked to smoking (8). Results from cohort studies provide the most convincing evidence for an association between myeloid leukemia and cigarette smoking. In addition, some case-control studies have reported an increased risk of AML (22, 27–29). Nonetheless, the internal validity of the exposure data in two of these studies (28, 29) can be questioned due to usage of interviews by proxy responders.

Inconsistent with many other findings, though, two previous cohort studies did not detected any significant associations between tobacco smoking and risk of leukemias (12, 30). However, both these cohort studies were limited by not analyzing leukemia subtypes independently.

Our results support the general evidence that there is no association between MM and tobacco smoking (11, 12, 14, 15). However, one recent study has shown conflicting results (31).

The effect of snuff on the risk of AML, ALL, CML, or MM has been investigated previously in one cohort study (12) and in three case-control studies, which found that snuff dipping was unrelated to the risk of developing leukemia (18, 32) and MM (11). One case-control study has reported an increased risk of MM, although based on small numbers (18).

Obesity has been considered to increase the risk leukemias by some investigators. This includes a population-based, case-control study of adult leukemia in Canada (22), where there was a positive association for AML and CML among overweight and obese people. Further support of an increased risk of leukemia, particularly AML, with level of BMI was observed in a cohort of Iowa women (21). In contrast, no association between obesity and leukemia risk was observed in a large cohort of hospitalized patients from Sweden with a discharge diagnosis of obesity (24) as well as a 10-year follow-up of a cohort of Korean men (23). Two previous cohort studies did not reveal any association between obesity and MM (23, 24), whereas conflicting results have been published in a few other studies (19, 33, 34).

Previous studies on BMI and risk of hematopoietic malignancies had some limitations that might affect the interpretation of the results. Case-control studies could suffer from recall bias. Furthermore, BMI of cases may be affected by disease development. One cohort study (21) was restricted to older women, which limits the external validity of the study. Moreover, conflicting results could partly be explained by the use of random digit dialing when enrolling controls, causing selection bias (22). In addition, in several studies (19–22), height and weight were self-reported, which could potentially lead to misclassification bias.

Tobacco smoke is known to consist of moderately high levels of benzene. Benzene has been shown to cause chromosomal aberrations, which in turn could be of importance in the causal pathway of AML (8). The mechanisms by which obesity could play an etiologic role in leukemia development are still not fully clarified. One potential causative pathway could be through elevated levels of insulin and insulin-like growth factor-I (IGF-I) among obese individuals (35) IGF-I promotes cell proliferation and inhibits apoptosis as well as increases propagation of bone marrow progenitor cells and receptors for IGF-I, which have been shown present on explanted leukemic cells (36).

To our knowledge, this study is one of the largest of its kind, with high quality of exposure and outcome data and a thorough and extensive follow-up. The study design ensures a very high internal validity, which is not affected by issues of “'representativeness” of the study population. Because there are no strong effect modifiers that are related to social class, we believe that our study also ensures external validity.

Although this makes it a highly valuable cohort, there are also limitations that should not be disregarded. The number of incident cases in some strata is too low to get adequate statistical precision. Furthermore, the data on BMI and tobacco use were obtained only at baseline, which makes interpersonal variability over the study period possible. However, such fluctuations in exposure status are most likely to be nondifferential and would bias the results toward the null.

In conclusion, we confirmed the presence of an association, likely to be causal, between tobacco smoking and AML risk, whereas an association with ALL risk cannot be excluded. Tobacco smoking does not affect or modify the risk of CML or MM. Snuff use and BMI level do not seem to affect risk of leukemia or MM.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Assistant Professor Dominique Michaud for supporting this study and Anna Torrång, Jenny Carlsson, and Juhua Luo for technical guidance of data management and data cleaning.