Tobacco and alcohol consumption and risk of lymphoma: Results of apopulation-based case-control study in Germany
Article first published online: 3 AUG 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 2, pages 422–430, 15 January 2006
How to Cite
Nieters, A., Deeg, E. and Becker, N. (2006), Tobacco and alcohol consumption and risk of lymphoma: Results of apopulation-based case-control study in Germany. Int. J. Cancer, 118: 422–430. doi: 10.1002/ijc.21306
- Issue published online: 9 NOV 2005
- Article first published online: 3 AUG 2005
- Manuscript Accepted: 5 APR 2005
- Manuscript Received: 4 OCT 2004
- The Federal Office for Radiation Protection. Grant Numbers: StSch4261, StSch4420
- German Research Foundation. Grant Number: 4850/161/03
Changing trends in lifestyle exposures are suggested to be contributing factors to the increasing incidence rates for lymphoma. We investigated the relationship between smoking and alcohol consumption and the risk of lymphoma among adult participants of a population-based case-control study recently conducted in Germany. In 710 case-control pairs, an increased risk of lymphoma was associated with a long duration of smoking (p for trend = 0.01 for men) and smoking of > 20 cigarettes per day(OR = 2.7; 95% CI = 1.4–5.2 for women). Elevated odds ratios were seen for most lymphoma subentities, albeit mostly without reaching statistical significance. A strong association was evident between smoking and multiple myeloma (OR = 2.4, 95% CI = 0.98–5.74 for men; OR = 2.9, 95% CI = 1.1–7.4 for women) and Hodgkin's lymphoma among men (OR = 3.6; 95% CI = 1.7–7.5). Alcohol consumption 10 years prior to the date of interview appeared to decrease the risk of lymphoma. Odds ratios for men who reported alcohol consumption were 53% lower (95% CI = 0.31–0.71) compared to men who drank very little or no alcohol. The same tendency was evident for women, although the association was less pronounced. The inverse relationship was also seen for low amounts of alcohol and did not appear to be restricted to specific types of beverages. Although biologic rationale for a protective effect of alcohol consumption may be given, a more in-depth analysis involving genetic markers is indicated to clarify if ethanol, other components in alcoholic beverages, or factors associated with moderate drinking reduce lymphoma risk among adults. In conclusion, this investigation suggests a positive association between tobacco smoking and lymphoma risk and finds decreased odds ratios among consumers of alcohol. © 2005 Wiley-Liss, Inc.
Tobacco and alcohol consumption are established risk factors for a broad range of cancer sites. Tobacco consumption was shown to be the most important single risk factor for cancer, accounting for an estimated 20% of all cancer deaths worldwide.1 Alcohol consumption is a moderate risk factor accounting for about 3% of all cancers.2
Lymphomas belong to the cancer sites with the strongest increase in incidence over the past decades, with an annual percent change of about 2–3% in Germany3 and a similar trend in other countries.4 Recent data indicate that the increase appears to level off in Germany.5 Since only a few other risk factors, mainly related to immunologic conditions or viruses, could be established, the etiology of these cancers remains largely obscure. Historically, neither alcohol6 nor tobacco (see, e.g., IARC7) was considered as established risk factors for lymphomas. However, some lymphoma subtypes such as Hodgkin's lymphoma and follicular lymphoma have repeatedly been associated with tobacco consumption.8, 9, 10, 11, 12, 13, 14 Furthermore, some recent studies reported a smoking-related increased risk of all non-Hodgkin's lymphoma (NHL) subtypes combined.13, 14, 15 For alcohol, in contrast, several publications reported surprisingly a decreased lymphoma risk especially correlated with specific drinking habits (wine consumption).16, 17, 18, 19, 20
We used the data of a recent population-based case-control study conducted in Germany on the etiology of lymphoma to analyze the potential impact of tobacco and alcohol consumption on the risk of lymphoma and main lymphoma subentities classified according to the WHO classification.
Material and methods
The study was carried out from 1999 to 2002 in 6 regions of Germany (Ludwigshafen/Upper Palatinate, Heidelberg/Rhine-Neckar County, Würzburg/Lower Frankonia, Hamburg, Bielefeld/Guetersloh, Munich) as a population-based case-control study among 18- to 80-year-old adults, 1:1-matched for gender, age (± 1 year of birth) and study region. Details on study design have been published elsewhere.21, 22 Briefly, the cases were recruited from hospitals and office-based physicians involved in diagnosis and treatment of lymphoma in the study regions, interviewed by trained interviewers and asked for a 20 ml blood sample. Eligible cases were asked for participation as soon as possible after their diagnosis. The median time from diagnosis to interview was 27 days (mean, 38 days). Participation rate was 87.4%. Diagnoses were collected in the form of copies of official pathology or physicians' reports, which were for about 46% of the cases verified by reference pathologists. Additionally, a 10% random sample of all included cases was reevaluated within a reassessment system in the context of the European collaboration, Epilymph. All cases were classified according to the WHO classification system.23 The controls were drawn randomly from population registers of the study regions, which have an almost 100% coverage due to compulsory registration by law. The controls were individually matched for gender, age and study region. Participation rate was 44.3%. The study comprises 710 case-control pairs (390 males, 320 females), including 92 cases of follicular lymphoma (FL), 158 cases of diffuse large B-cell lymphoma (DLBCL), 104 cases of B-cell chronic lymphatic leukemia (CLL), 76 cases of multiple myeloma, 29 cases of mucosa associated lymphoid tissue (MALT) lymphoma, 554 cases of B-NHL, 35 cases of T-NHL and 115 cases of Hodgkin's lymphoma (HL). One case was diagnosed as HL and NHL.22
Information on demographic factors and lifestyle characteristics including lifelong smoking habits and alcohol consumption 5–10 years prior to the date of interview were obtained by in-person interviews. Individuals were classified as smokers if they answered positively to the following question: “Do you smoke regularly or have you been smoking on a regular basis at least 1 cigarette, cigar, pipe, … per day for at least 6 months?” A more detailed smoking history was obtained from all smokers, including questions of current and past smoking habits, age of smoking initiation (ever smokers) and quitting (former smokers), intensity of smoking in phases and duration of smoking. Current smokers were defined as study participants who reported smoking 2 years prior to the date of interview. Smoking of small cigars, regular cigars and pipes was converted to the equivalent of 2 (for small cigars) or 3 (for regular cigars, pipes) cigarettes.24
Individuals were queried about their average beer, wine, spirits and aperitif consumption 5–10 years prior to the date of interview. For each beverage type separately, they could answer in bottles or glasses of common size per day, week, or month. Nondrinkers were defined as men with less than 2 g and women with less than 0.5 g of total alcohol consumption per day. Data analysis was carried out for all lymphoma combined, the major subgroups HL, B-NHL and T-NHL and the more frequent subentities (DLBCL, FL, MALT lymphoma and multiple myeloma). More detailed analyses of alcohol consumption and smoking intensity and duration by subtype are left for subsequent pooled analyses in the context of the Epilymph study, a multicenter European study in which the German project is embedded.
All lymphoma subtypes combined were analyzed using matched conditional logistic regression, taking the matching variables gender, age and study region into account. Due to sparse data, subentities were analyzed by unconditional logistic regression using the matching variables for adjustment. As age variable, we used age at diagnosis of the case and age of the control at the time of diagnosis of the respective matched case. Relative risk was estimated by odds ratios (ORs) and associated 95% confidence intervals (95% CIs) using the SAS procedure PHREG for conditional and LOGISTIC for unconditional logistic regression (SAS version 8; SAS Institute, Cary, NC). The 2-sided Cochran-Armitage test for trend was used.
Smoking was only considered up to 2 years prior to the diagnosis/reference date. Pack-years were calculated by multiplying smoking duration and intensity. According to the above definition, nonsmokers did not smoke at all or smoked not more than 0.075 pack-years and were taken as reference.
We defined nondrinkers differently between both genders. Nondrinking men were defined as consumers of less than 2 g of alcohol per day, whereas nondrinking women were defined as consumers of less than 0.5 g of alcohol per day. This corresponds to less than one bottle of beer or less than one glass of wine per week (men) and month (women), respectively. The daily intake of alcohol of a particular beverage type in grams was calculated by multiplying the frequency of consumption by its respective ethanol content [1 bottle of beer = 400 ml, 1 ml beer = 0.035 g ethanol; 1 glass of wine = 200 ml, 1 ml wine = 0.1 g ethanol (just for study region 3: 1 ml “most” or wine = 0.07875 g ethanol); 1 glass of spirit = 35 ml, 1 ml spirits = 0.35 g ethanol; 1 glass of aperitif = 50 ml, 1 ml aperitif = 0.16 g ethanol; according to the German food code and nutrition database (BLS) II version 3, BgVV 1999, Berlin, Germany]. The overall daily alcohol intake was calculated by summing the ethanol amounts (in grams) of each beverage type. Nondrinkers, as defined above, were taken as reference category.
The ORs were adjusted for educational level in 2 categories (low = <10 years of school and none or regular vocational training; medium/high = ≥10 years of school and further education and/or training, low educational level taken as reference) and for a binary variable drinker for the analysis of smoking variables, or smoking in pack-years (as continuous variable) for the analysis of alcohol variables, respectively. For the evaluation of specific types of beverages, we additionally adjusted for total ethanol in 4 categories. The unconditional logistic regression models were further adjusted for age and study region in 6 categories. For some subentity-specific analyses, small study regions were collapsed and therefore fewer categories considered for adjustment.
In Table I, the age distribution of the study population, the level of education, as well as smoking and drinking patterns in relation to gender are presented. The educational level differed between cases and controls especially for women (64% of the cases were of low educational level compared to 54% of the controls). Among men, 45% reported a low educational level compared to 39% of the controls. More men (69%) than women (42%) reported ever smoking. For men, ever smoking was a bit more prevalent in cases (71%) than in controls (66%), whereas among woman slightly more controls (43%) than cases (41%) smoked. Current smoking 2 years prior to the date of diagnosis/reference date was reported among men by 32% of the cases and by 25% of the controls. Also among females, more cases (23%) than controls (20%) were current smokers. Differences between men and women were seen for variables of smoking intensity: men tended to smoke more cigarettes per day, were characterized by a longer duration of smoking and tended to start smoking earlier.
|Men, Cases/controls, 390 (100%)/390 (100%)||Women, Cases/controls, 320 (100%)/320 (100%)|
|Age at diagnosis1|
|18–24||23 (5.9%)/25 (6.4%)||16 (5.0%)/16 (5.0%)|
|25–34||33 (8.5%)/33 (8.5%)||26 (8.1%)/27 (8.4%)|
|35–44||50 (12.8%)/48 (12.3%)||35 (10.9%)/34 (10.6%)|
|45–54||53 (13.6%)/55 (14.1%)||50 (15.6%)/51 (15.9%)|
|55–64||97 (24.9%)/93 (23.8%)||69 (21.6%)/64 (20.0%)|
|65–74||101 (25.9%)/100 (25.6%)||87 (27.2%)/91 (28.4%)|
|≥75||33 (8.5%)/36 (9.2%)||37 (11.6%)/37 (11.6%)|
|Low||176 (45.1%)/151 (38.7%)||204 (63.8%)/172 (53.8%)|
|Medium or high||210 (53.8%)/238 (61.0%)||115 (35.9%)/148 (46.3%)|
|Ever smokers||277 (71.0%)/259 (66.4%)||132 (41.3%)/137 (42.8%)|
|Current smokers||125 (32.1%)/99 (25.4%)||72 (22.5%)/63 (19.7%)|
|Former smokers||152 (39.0%)/160 (41.0%)||60 (18.8%)/74 (23.1%)|
|Smoking characteristics among ever smokers|
|Amount of smoking (cigarettes/day)||18.3/18.5||14.5/11.3|
|Duration of smoking (years)||25.7/22.9||22.3/18.5|
|Amount of smoking (pack-years)||25.4/22.3||17.3/11.2|
|Onset of smoking (age)||18.5/19.0||21.0/22.1|
|Drinking (5–10 years prior to the date of interview)|
|Alcohol drinkers||287 (73.6%)/329 (84.4%)||233 (72.8%)/255 (79.7%)|
|Drinking characteristics among drinkers|
|Alcohol intake (g ethanol/day)||22.2/21.7||8.5/8.4|
|Proportion of beer (%)||44.0/41.6||20.7/17.6|
|Proportion of wine (%)||41.4/45.5||63.9/66.3|
|Proportion of spirits (%)||13.2/11.4||10.6/11.1|
|Proportion of aperitifs (%)||1.4/1.5||4.8/5.0|
Among men, 74% of the cases and 84% of the controls consumed alcohol (defined as consumption of 2 g or more alcohol per day) 5–10 years prior to the date of interview. For women, the respective numbers were 73% and 80%. Men consumed on average 2.6-fold more alcohol than women. Wine and beer contributed to the total amount of alcohol to similar proportions among men, whereas among women, wine consumption was more prevalent.
In Table II, the association of smoking variables with the risk of lymphoma is described separately for men and women. Compared to nonsmokers, ever smoking was correlated with a nonsignificant 29% increased risk of lymphoma among men, current smoking with a statistically significant 51% increased risk. For women, ever smoking did not increase lymphoma risk. The second (10 to < 20 cigarettes/day) and third (20 to < 30 cigarettes/day) category of smoking intensity correlated among men with nonsignificantly increased ORs, which did not extend to the highest category. For women, the highest category of smoking intensity was associated with a statistically significant 2.7-fold risk of lymphoma. In comparison to nonsmoking men, long-term smokers (≥ 30 years of smoking) had a 57% higher risk of lymphoma. For women, the association of lymphoma risk with smoking duration was less evident. For men and women, more than 15 pack-years were associated with an increased risk of all lymphoma subtypes combined (p for trend = 0.01 for men and 0.12 for women). After quitting, risk remained high for 15 years and disappeared thereafter. For men, a tendency of higher lymphoma risk with decreasing age at initiation of smoking was evident, which, however, lacks statistical significance.
|Smoking characteristics1||All lymphoma subtypes combined|
|Men (n = 780)||Women (n = 640)|
|Cases/controls||OR3||95% CI||Cases/controls||OR3||95% CI|
|Average intensity of smoking, cigarettes per day|
|> 0 to < 10||55/61||1.08||0.69–1.71||39/55||0.76||0.47–1.25|
|10 to < 20||94/83||1.39||0.92–2.10||49/61||0.91||0.57–1.46|
|20 to < 30||85/71||1.45||0.94–2.25||43/20||2.66||1.37–5.18|
|Duration of smoking, years|
|> 0 to < 10||37/54||0.79||0.47–1.31||26/43||0.60||0.33–1.10|
|10 to < 20||60/61||1.37||0.84–2.23||35/34||1.03||0.57–1.87|
|20 to < 30||74/64||1.47||0.94–2.31||34/32||1.17||0.67–2.07|
|> 0.075 to < 15||99/119||1.00||0.68–1.48||69/100||0.73||0.48–1.12|
|> 15 to < 30||75/62||1.53||0.97–2.41||41/26||1.63||0.93–2.85|
|Age at onset of smoking|
|< 15 years||36/24||1.64||0.90–3.01||73/70||1.15||0.73–1.81|
|15 to < 20||149/144||1.25||0.87–1.80||73/70||1.15||0.73–1.81|
|20 to < 25||54/60||1.13||0.70–1.83||28/23||1.34||0.74–2.43|
|Years since smoking cessation|
|0 to < 15 years||189/147||1.51||1.06–2.15||109/95||1.27||0.86–1.88|
|15 to < 30 years||51/70||0.92||0.56–1.50||15/31||0.51||0.25–1.04|
In Table III, the effect of smoking on the main subentities of lymphoma is shown. The elevated risk among men was especially seen for HL. Ever smokers had a significant 2.8-fold risk of HL, which was largely due to an increased OR for current smokers. A strong effect of current smoking was evident for multiple myeloma in both genders. Elevated odds ratios were also seen for follicular and MALT lymphoma, as well as CLL and T-NHL, however, restricted to men. Diffuse large B-cell lymphoma showed an inverse relationship with smoking among men.
|Cases/controls||OR3||95% CI||Cases/controls||OR3||95% CI|
|Diffuse large B-cell lymphoma|
HL among men presented a particularly clear association with smoking duration. The respective ORs were 3.6 (95% CI = 1.4–9.3) for 10–20 years of smoking, 7.0 (95% CI = 2.6–19.1) for 20–30 years of smoking and 5.9 (95% CI = 1.7–20.8) for the last category of smoking duration. Only among women were smoking of ≥ 20 cigarettes per day and > 15 pack-years associated with a statistically significant increased risk of follicular lymphoma. Multiple myelomas were significantly more frequent among men who reported more than 30 years of smoking and women who smoked 20 cigarettes and more per day (data not shown).
Tables IV and V show the association of alcohol consumption 5–10 years prior to the date of interview and the risk of all lymphoma subtypes combined (Table IV) and individual subentities (Table V). Men who reported regular alcohol consumption 5–10 years prior to the date of interview had a statistically significant lower risk of lymphoma than nondrinkers (defined as men who consumed less than 2 g of ethanol per day). For women, the association of alcohol consumption and lymphoma was less pronounced. In comparison to nondrinking women (defined as < 0.5 g alcohol/day), alcohol consumption was associated with a 32% decreased risk of all lymphoma subtypes combined. The inverse relationship was particularly strong for follicular lymphoma and B-NHL for men and HL for both genders (Table V).
|Drinking characteristics||Cases/controls||OR1||95% CI|
|Amount of alcohol|
|2 to < 10 g ethanol/day||117/125||0.52||0.33–0.81|
|10 to < 40 g ethanol/day||129/162||0.41||0.26–0.65|
|≥ 40 g ethanol/day||41/42||0.50||0.28–0.91|
|Amount of alcohol|
|0.5 to < 2 g ethanol/day||87/96||0.67||0.42–1.07|
|2 to < 10 g ethanol/day||93/107||0.66||0.41–1.08|
|> 10 g ethanol/day||53/52||0.73||0.42–1.27|
|Cases/controls||OR1||95% CI||Cases/controls||OR1||95% CI|
|Diffuse large B-cell lymphoma|
Approximately 48–59% reduced ORs were seen for all alcohol categories among men and a 27–34% reduction of borderline significance among women (Table II). Men who drank 10–40 g ethanol per day had a statistically significant 57% reduced risk of HL (data not shown).
In the unconditional stratified analysis by educational level, the inverse association between alcohol consumption and lymphoma risk was less prominent for men of low educational level (OR = 0.72; 95% CI = 0.41–1.27) than for men of medium or high educational level (OR = 0.41; 95% CI = 0.25–0.67; data not shown). The same tendency was seen for women; however, the ORs were not statistically significant in both strata. Risk estimates were not substantially different between ever smokers and nonsmokers (data not shown). The inclusion of several variables that were associated with an increased or decreased lymphoma risk in previous analyses,21, 22 such as allergies, contact with animals in childhood, or vaccinations, had only marginal effects on the association between alcohol consumption and lymphoma risk (data not shown).
In Tables VI and VII, we present lymphoma risks associated with the consumption of different types and amounts of alcoholic beverages adjusted for only smoking and educational level as well as adjusted additionally for the total amount of alcohol. In the first model, all beverage types were associated with a statistically significant decreased risk of lymphoma among men. In the second model, the highest category of consumption showed a clearly decreased OR only for wine consumption. In both models, none of the beverage types correlated with a statistically significant change of lymphoma risk among women (Table VII).
|Drinking characteristics||Cases/controls||OR1||95% CI||OR2||95% CI|
|≤ 5 bottles/months or < 1 bottle/week||90/106||0.44||0.27–0.72||0.55||0.30–0.98|
|1–5 bottles/week or < 1 bottle/day||113/138||0.44||0.28–0.69||0.61||0.32–1.14|
|≥ 1 bottle/day||69/76||0.42||0.24–0.72||0.60||0.28–1.31|
|< 0.5 bottle/month or < 1 glass/week||75/92||0.42||0.26–0.69||0.50||0.27–0.90|
|1 glass to 0.5 bottle/week or < 1 glass/day||90/100||0.44||0.27–0.72||0.56||0.28–1.13|
|≥ 1 glass/day||73/110||0.33||0.20–0.55||0.40||0.19–0.85|
|≤ 5 glasses/month or < 1 glass/week||125/176||0.40||0.26–0.62||0.57||0.32–1.00|
|≥ 1 glass/week||83/89||0.48||0.29–0.79||0.77||0.38–1.55|
|< 1 glass/month||44/100||0.21||0.12–0.36||0.28||0.14–0.55|
|≥ 1 glass/month||40/57||0.40||0.22–0.71||0.53||0.26–1.08|
|Drinking characteristics||Cases/controls||OR1||95% CI||OR2||95% CI|
|≤ 2 bottles/month||93/116||0.63||0.39–1.02||0.58||0.32–1.07|
|3–5 bottles/month or < 1 bottle/week||42/27||1.09||0.58–2.06||1.05||0.46–2.38|
|≥ 6 bottles/month or ≥ 1 bottle/week||31/34||0.63||0.33–1.20||0.66||0.28–1.53|
|≤ 2 glasses/month||83/89||0.68||0.42–1.09||0.58||0.33–1.03|
|≤ 0.5 bottle/week or < 1 glass/day||91/112||0.64||0.39–1.04||0.52||0.23–1.16|
|≥ 1 glass/day||43/40||0.73||0.41–1.33||0.75||0.28–1.97|
|≤ 2 glasses/month||97/107||0.67||0.41–1.09||0.54||0.29–1.02|
|> 2 glasses/month||33/38||0.60||0.33–1.08||0.53||0.24–1.14|
|< 1 glass/month||73/80||0.65||0.38–1.12||0.53||0.27–1.05|
|≥ 1 glass/month||36/45||0.59||0.32–1.08||0.50||0.24–1.05|
For men, a simultaneous model (adjusted for educational level and smoking) including all beverage types in 4 categories for wine and beer and 3 categories for aperitif and spirits revealed decreased OR for wine, beer and aperitif consumption, statistically significant for the category of highest consumption (wine, beer) and the medium consumption of aperitif (data not shown). Spirit consumption was not associated with lymphoma risk.
In this population-based case-control study on the etiology of lymphoma, we have identified smoking as a risk factor for all lymphoma subtypes combined and presumably several subentities, potentially with the exception of diffuse large B-cell lymphoma. The risk was increased with long duration of smoking (statistically significant for men) and heavy smoking (statistically significant for women). Alcohol consumption 5–10 years prior to the date of interview decreased the risk of lymphoma (statistically significant for men; of borderline significance for women). The inverse association was already seen at low amounts of alcohol and did not appear to be restricted to a particular type of beverage.
So far, epidemiologic studies have provided inconsistent evidence for the association of smoking with lymphoma risk. For most entities, some studies showed a positive association, while some other studies presented null results (NHL reviewed in Peach and Barnett25; HL reviewed in Briggs et al.8; multiple myeloma citations in Adami et al.26 and Stagnaro et al.13).
Some cohort studies have investigated the association between smoking and NHL risk.25 One cohort study reported a strong association between past smoking and follicular lymphoma.11 However, the causal interpretation was questionable due to a missing trend of increased risk with increasing duration or intensity of smoking.11 Linet et al.27 found ever smoking associated with an elevated mortality from NHL (RR = 2.1; CI = 0.9 – 4.9) among men. A statistically significant 3.8-fold risk of dying from NHL was seen for the heaviest smokers. A recent review evaluating the evidence from epidemiologic studies concluded that cohort studies found no association between past smoking and NHL.25 The authors, however, stated that on the basis of data from cohort studies it was difficult to reach a conclusion about the association between NHL and current smoking.25 A number of case-control studies have investigated the relationship between smoking and NHL risk (some reviewed previously13, 14, 19, 25, 28, 29, 30, 31). In 6 studies, smoking was not associated with NHL risk.17, 19, 28, 31, 32, 33 The remaining studies presented some indication for an elevated disease risk associated with smoking.13, 14, 15, 16, 30, 34, 35 However, due to small sample sizes of some studies, a missing dose-response relationship and the possibility of the operation of different biases, a causal interpretation of the association between smoking and NHL is weakened.
Our study provides further evidence for an association between smoking and lymphoma risk. A long duration of smoking was a risk factor for all lymphoma subtypes combined. Interestingly, the effect was more pronounced for men than for women and reached statistical significance only in the highest category (≥ 30 years) among men. Thus, studies that evaluated both genders combined may have failed to detect a gradient, especially if they were unable to consider the category of 30 years or more of duration of smoking (e.g., due to too small sample sizes). Nevertheless, a gradient of lymphoma risk with duration of smoking was also seen in several other studies.13, 16, 34, 36 These results support the notion that long latency periods are needed before a smoking-related effect on lymphoma risk will be evident.
The relationship between intensity of exposure and lymphoma risk may provide a clue to understand the lower overall effect of smoking on lymphoma risk among women. A strong statistically significant positive association could only be seen within the combined 2 highest categories of smoking (≥ 20 cpd), which comprised only about 1/3 of all smoking cases. Thus, the lack of association in the lower categories may cover the association in the overall evaluation. The relevance of smoking is underlined by the trend toward elevated risks (statistically significant among men) associated with increasing pack-years.
We observed differences in the effect of smoking on lymphoma subentities, with strongest positive associations seen for HL (men) and multiple myeloma, and an inverse relationship with DLBCL. Several studies have addressed smoking as a risk factor for HL, mostly with null or inconclusive results (reviewed in Briggs et al.8). Some recent studies, however, presented some evidence for a positive correlation of tobacco consumption with HLrisk.8, 9, 10 A study by Briggs et al.8 found a significant 80% increased risk of HL for current male smokers. Greater intensity of smoking and longer smoking duration were associated with a 2.5-fold increased risk of HL, especially of the mixed cellularity subtype. A recent study by Glaser et al.9 reported a significant, almost 3-fold elevated risk of Epstein-Barr virus (EBV)-positive HL among current female smokers. Furthermore, they found household environmental tobacco smoke exposure to be positively correlated with risk for EBV-positive HL. Two prospective studies among men found evidence for an increased risk associated with current smoking, which showed or approached significance in a stratified analysis.26, 37
In previous studies,11, 12, 13, 14, 29, 30 smoking was associated with the risk of follicular lymphoma, particularly among women.12, 13, 29 In this study, only current smoking correlated in both genders with an > 80% increased risk (not significant), whereas ever and former smoking were no risk factors. The stronger effect of current smoking was also seen in other recently conducted studies.12, 13, 29 The significant effect of smoking intensity and pack-years on FL risk was only confined to women.
In this study, current smoking correlated with significantly elevated OR for multiple myeloma. A role of cigarette smoking in the etiology of multiple myeloma was also suggested in 2 European studies38, 39 and in an American cohort study based on a few cases.40 Most other cohort and case-control studies, however, found no association.13, 26
Heterogeneity in the study populations and selection of controls, in the classification of cases based on the working formulation or Revised European-American Lymphoma (REAL)/WHO, or in the definition of smoking variables as well as chance may have contributed to the inconsistent findings between smoking variables and the risk of individual lymphoma subentities.
A further finding to be addressed is the apparent lack of association or potentially the even inverse relationship between smoking and the risk of diffuse large B-cell lymphoma. Similar results have also been reported by Zahm et al.36 It could be speculated that differences in body mass index (BMI) among smokers contribute to these findings. An increased body mass index in young adulthood and adulthood was recently shown to be positively associated with the risk of diffuse large B-cell lymphoma.41
Concerning the mechanisms that may underlie an increased risk of lymphoma for smokers, folate deficiency,42 oxidative stress43, 44 and an upregulation of antiinflammatory cytokines and impairment of cellular immunity45, 46, 47 may be discussed. Furthermore, several lines of evidence suggest a modulating role of nicotine on apoptotic processes.44, 48, 49 Apoptosis is of great importance for the maintenance of homeostasis, particularly in the immune system, where new cells are continuously generated and destroyed. Nicotine was recently shown to induce phosphorylation of Bcl-2 and to suppress apoptosis in human small cell lung cancer cells.48 The antiapoptotic gene BCL2 is often affected by translocation in B-cell lymphoma. A dysregulation of cell apoptosis and proliferation in smokers was also demonstrated by Tomita et al.50
Most of the epidemiologic studies on lymphoma etiology that addressed alcohol consumption indicated an inverse relationship, but were inconsistent about the nature of the effect potentially due to the restriction to only women or specific NHL types.16, 17, 18, 19, 20 One particular study found an increased risk for high consumption of beer and maté drinking,15 another study reported an association between alcohol consumption and NHL risk for men with a family history of hematolymphoproliferative cancer51 and 2 other studies reported null results.31, 52
Several studies related the decreased risk of lymphoma to wine consumption.20, 33, 53 In 2 of these, the association was specific for wine, as no comparable relationship could be determined for beer or spirits consumption.20, 53 The study by Morton et al.,20 restricted to women, reported additionally a consistently inverse association between wine consumption across all major NHL subentities without, however, presenting the results in detail.
In contrast, other studies have cast doubts on a protective role of specifically wine consumption. In 2 studies,17, 18 a protective effect of alcohol consumption was not restricted to wine drinking. In these analyses, however, the authors failed to control for the consumption of other alcoholic beverage types when evaluating each type separately. A further study conducted in Japan found also a negative association of current drinking on the risk of malignant lymphoma, whereby wine drinking was left out of consideration because it plays no major role in Japan.19
Our data also support an effect of alcohol, and not of a specific type of alcoholic beverage, on the risk of lymphoma, which was evident already at low intensities of consumption. Increasing amounts of alcohol did not appear to strengthen the inverse relationship. Interestingly, the lowest ORs occurred for alcohol consumption levels (10–40 g for men and 2–10 g for women), which range within the public health recommendation, derived from the inverse effects of alcohol consumption on coronary heart disease risk. For cardiovascular disease risk, a protective effect of alcohol is already seen at relatively low intake levels and follows aJ-shaped curve.54 Due to the few individuals with very high intake levels and the assessment mode of high consumption in this study, we could not explore how heavy drinking affects lymphoma risk. Initially, we hypothesized a positive relationship between alcohol consumption and lymphoma risk because of its frequently cited immunosuppressive effect due to, e.g., the downregulation of Th1 cytokines important for antibacterial, antiviral and antitumor immunity.55, 56 However, moderate consumption of alcohol appears to have also beneficial effects on cellular and humoral immune responses, as was recently summarized by Diaz et al.57 Alternative mechanisms underlying a potential protective effect of alcohol on lymphoma risk may include antioxidants such as resveratrol in wine or flavonoids in beer, improvement of insulin sensitivity by alcohol,58, 59 or the interaction between retinol and ethanol metabolism.60, 61
For the overall interpretation of our results, some strengths and limitations of this study should be pointed out. To the strengths we count the relatively large sample size, its population-based design and the presentation of exposure information from standardized questionnaires obtained by trained interviewers. Furthermore, nearly all diagnoses (98.9%) were confirmed by pathology or physicians' reports and in about 46% of the cases additionally verified by an independent source.
We asked for information on drinking habits 5–10 years prior to the date of interview to evaluate the etiologic relevance of alcohol consumption. Misclassification of exposure may have distorted our results; however, substantial underreporting of the cases is not expected, since smoking and alcohol consumption were not considered as substantial risk factors for lymphoma in the past.
Of concern is the low participation rate among the controls and the related potential for selection bias. The level of nonparticipation was not equal for both genders and different age groups. Overall, participation was higher for men (46%) than for women (41%). The lowest participation was seen among women above 55 years of age. In this group of women, only 35% participated, whereas among the younger women 57% participated. However, the ORs for the association between the highest intensity of smoking and lymphoma were similarly elevated in both groups (OR = 3.4, 95% CI = 1.3–8.8 for women < 55 years of age; OR = 2.8, 95% CI = 0.97–7.9 for women ≥ 55 years of age). The largest difference in the percentage of current smoking between cases and controls was in the group of women between 45 and 55 years of age. In this group, 67% of women participated, which may speak against a strong compliance-related bias of our smoking-related results. It is also worth mentioning that we used a conservative approach to calculate participation rates.
To address the potential for selection bias, we tried to compare smoking and drinking habits of our control population with estimates from a population survey conducted in Germany. For male controls, the smoking prevalence in different age strata compared very nicely with estimates from the National German Health Survey.62 Also for female controls, the prevalence did not differ substantially, except for the age group of 30–44, for which smoking was more prevalent in our control population.62 The comparison of alcohol intake in our study with survey data, however, turned out to be difficult because of the substantial differences in intake levels and patterns between different geographic areas in Germany and consequently the lack of suitable estimates.
A matter of concern with regard to the analysis of alcohol consumption data is the reference category of nondrinkers. Although we did not restrict the reference category to abstainers by including individuals with low alcohol consumption, we cannot exclude that the individuals of the reference category differ from the rest with respect to other variables associated with lymphoma risk. However, when we excluded the abstainers and took as reference men with the consumption of < 2 g alcohol per day, the inverse association remained largely unchanged.
Furthermore, certain drinking patterns, such as the regular consumption of alcohol and wine in particular, correlate with specific behavior and lifestyle characteristics, including a better diet.63, 64, 65 The inverse association between alcohol consumption and lymphoma risk is strongest, and only there statistically significant, for individuals of higher educational level, suggesting an effect modification and/or confounding by socioeconomic status. A beneficial lifestyle associated with higher educational level may have contributed to the observed findings. Conversely, the small risk reduction by alcohol may have been overlaid by other risk factors more prevalent among individuals of lower educational level.
Before concluding that alcohol consumption is a protective factor for lymphoma, a more in-depth analysis of this topic, including larger sample sizes to allow for a better estimation of risks typical for specific subentities and population strata, is warranted. Also, the evaluation of lifelong alcohol consumption and information on binge drinking is suggested. Additionally, the consideration of genetic variants in alcohol metabolizing genes is indicated to improve our knowledge on the relationship between alcohol drinking and lymphoma risk.
In conclusion, our study suggests that long-term and heavy smoking increases the risk of lymphoma, especially of Hodgkin's lymphoma and multiple myeloma. Furthermore, the results of this study are in line with previous reports showing an inverse relationship between alcohol consumption and lymphoma risk.
The authors are indebted to the participants of the study and to the many colleagues who supported the performance of the study. A detailed list is presented in Becker et al.21, 22 Furthermore, they thank Jakob Linseisen for helpful discussions and critically reading the manuscript. The European Community supported the setup of a common protocol for assessment of occupational exposures (SOC 98 201307 05F02) and implementation of additional study areas (QLK4-CT-2000-00422). A recent workshop of the industrial hygienists in Heidelberg was funded by the German Research Foundation (4850/161/03).
- 3Atlas of cancer mortality in the Federal Republic of Germany (1981–1990), 3rd ed. Berlin: Springer, 1997., .
- 5Association of Population-Based Cancer Registries in Germany. Cancer in Germany. Saarbrücken: Association of Population-Based Cancer Registries in Germany, 2004.
- 6IARC. Alcohol drinking: epidemiological studies of cancer in humans. IARC Monogr Eval Carcinog Risks Hum 1988; 44: 153–250.
- 7IARC. Tobacco smoke and involuntary smoking. vol. 83. Lyon: IARC, 2004.
- 23Pathology and genetics of tumours of haematopoietic and lymphoid tissues. Lyon: IARC, 2001., , , .
- 24IARC. European multicentre case-control study of lung cancer in non-smokers. 1998. IARC technical report no. 33. Lyon: IARC.