Association of Alcohol Intake and Smoking with Malignant Lymphoma Risk in Japanese: A Hospital-Based Case-Control Study at Aichi Cancer Center

Several case-control and cohort studies have investigated the association of alcohol intake and smoking with the risk of Hodgkin's and non-Hodgkin's lymphoma in non-Asian populations, with most suggesting an inverse association for alcohol consumption with the risk of malignant lymphoma (1-6) versus inconsistent results for smoking (2-12). However, little research has been conducted in Asian populations, particularly with regard to the effects of smoking and alcohol consumption on the risk of malignant lymphoma subtypes. Although the incidence of malignant lymphoma in Japan has almost doubled in the last three decades (13, 14), it remains substantially lower than in western countries (15), with age-adjusted incidence rates in 2002 of 7.7 for men and 4.9 for women (13), almost half those in western countries (15). Further, the distribution of malignant lymphoma differs between Asian and western countries: marginal zone B-cell lymphoma is frequent in Asia, for example, whereas Hodgkin's lymphoma and chronic lymphocytic leukemia/small lymphocytic lymphoma are rare (15-17). These differences highlight the importance of epidemiologic investigation of malignant lymphoma in Asian countries.

Here, we conducted a case-control study to investigate the effect of alcohol and smoking on the risk of malignant lymphoma subtypes in a Japanese population using two large data sets, the first and second versions of the Hospital-based Epidemiological Research Program at Aichi Cancer Center Hospital (HERPACC-I and HERPACC-II, respectively).

Subjects were enrolled in HERPACC-I between January 1988 and December 2000 and in HERPACC-II between January 2001 and November 2005. Details of HERPACC have been described elsewhere (18, 19). In brief, HERPACC-I was initiated at Aichi Cancer Center Hospital in 1988. All outpatients, including cancer and noncancer patients, were asked about their lifestyle at first visit, specifically information regarding their lifestyle before the onset of symptoms for those with symptoms and up to the time of interview for those without. Information from the questionnaire was systematically collected and checked by trained interviewers. HERPACC-II was launched in 2001, with all first-visit outpatients requested to provide 7 mL blood and more detailed information on lifestyle factors than that provided in HERPACC-I. Approximately 95% of eligible subjects, including cases and controls, completed the questionnaire. The data were loaded into the HERPACC database and periodically linked with the hospital cancer registry system to update the data on cancer incidence. We previously confirmed that the lifestyle patterns of first-visit outpatients without cancer were in accordance with those in a general population of subjects randomly selected from Nagoya City in both the HERPACC-I (20) and the HERPACC-II,⁷

A total of 452 and 330 patients with no prior history of cancer were histologically diagnosed with malignant lymphoma at our center in HERPACC-I and HERPACC-II, respectively. All lymphoma subtypes were reclassified based on the WHO classification of 2001 by pathologists in our center (21). Sex- and age-matched (±2 years) control groups for cases of HERPACC-I and HERPACC-II were independently selected from noncancer outpatients in HERPACC-I and HERPACC-II, respectively, with an overall case-control ratio set at 1:5 to maximize statistical power. Several of these cases and controls in the HERPACC-I were included in our previous article (6).

Daily alcohol consumption (in grams) was determined by totaling the amount of pure alcohol in the average daily consumption of Japanese sake (rice wine), shochu (distilled spirit), beer, wine, and whiskey. The amount of ethanol in each beverage has been described elsewhere (22). The subjects were categorized based on drinking status into the three groups of never, former, and current drinkers. Owing to a difference in questionnaire items, weekly drinking frequency was categorized into three groups in HERPACC-I (0, <4, and ≥4 times per week) and five groups in HERPACC-II (0, <1, 1-2, 3-4, and ≥5 times per week). For the pooled analysis, frequent drinkers were defined as subjects who drank ≥4 times per week in HERPACC-I and ≥5 times per week in HERPACC-II. The subjects were further categorized into four groups based on alcohol intake: never drinkers, occasional drinkers, frequent but moderate drinkers (<50 g/d alcohol), and frequent and heavy drinkers (≥50 g/d alcohol).

With regard to smoking, the two HERPACC studies used the same questionnaire items. Subjects were categorized into the three groups of never, former, and current smokers. Other categories evaluated were the number of cigarettes smoked per day (<10, 10-19, and ≥20), duration of smoking (<20, 20-39, and ≥40 years), years since smoking cessation (never smokers; former smokers with <5, 5-9, and ≥10 years since cessation; and current smokers), and the number of pack-years, defined as the product of the average number of packs per day and the number of years of smoking (<5, 5-19, 20-39, and ≥40 pack-years).

Owing to the slight difference between the two data sets in questionnaire items for alcohol, the effect of drinking-related variables on the risk of malignant lymphoma was analyzed separately for the two HERPACC data sets. For malignant lymphoma subtypes, data were pooled when findings across sexes and data sets were closely similar. Odds ratios (OR) and 95% confidence intervals (95% CI) were estimated using a conditional logistic regression model, which included smoking exposure, expressed in pack-years, and alcohol intake. All statistical analyses were done using Stata version 10 (Stata), with P < 0.05 considered statistically significant.

The proportion of men among all subjects in HERPACC-I and HERPACC-II was 59.3% and 51.5%, respectively, and median age (years) was 55 (range, 18-79) and 56 (range, 18-80), respectively. With regard to histologic subtypes, respective distributions in HERPACC-I and HERPACC-II were diffuse large B-cell lymphoma in 167 (36.9%) and 113 (34.2%), follicular lymphoma in 69 (15.3%) and 82 (24.8%), marginal zone lymphoma in 76 (16.8%) and 60 (18.2%), Hodgkin's lymphoma in 26 (5.8%) and 29 (8.8%), peripheral T-cell lymphoma of unspecified type in 23 (5.1%) and 2 (0.6%), mantle cell lymphoma in 10 (2.2%) and 11 (3.3%), precursor T-lymphoblastic leukemia/lymphoma in 8 (1.8%) and 1 (0.3%), angioimmunoblastic T-cell lymphoma in 7 (1.5%) and 1 (0.3%), extranodal natural killer/T-cell lymphoma of nasal type in 7 (1.5%) and 1 (0.3%), lymphoplasmacytic lymphoma in 5 (1.1%) and 2 (0.6%), anaplastic large cell lymphoma in 5 (1.1%) and 1 (0.3%), chronic lymphocytic leukemia/small lymphocytic lymphoma in 1 (0.2%) and 4 (1.2%), other non-Hodgkin's lymphoma in 5 (1.1%) and 1 (0.3%), non-Hodgkin's lymphoma not otherwise specified in 33 (7.3%) and 11 (3.3%), and malignant lymphoma not otherwise specified in 10 (2.2%) and 11 (3.3%).

In both data sets, results showed a consistently inverse association between current drinking and malignant lymphoma [OR (95% CI), 0.66 (0.52-0.84) for HERPACC-I and 0.60 (0.46-0.78) for HERPACC-II; Table 1]. Further, trends for the effect of drinking frequency on malignant lymphoma risk were significant in both data sets (P_trend = 0.011 for HERPACC-I and 0.001 for HERPACC-II). Compared with nondrinking, consumption of ≥50 g/d in frequent drinkers was significantly associated with the risk of malignant lymphoma in both data sets [OR (95% CI), 0.70 (0.53-0.93) for HERPACC-I and 0.40 (0.23-0.68) for HERPACC-II]. Further, to explore the interaction between alcohol and smoking, we analyzed the effect of alcohol consumption among either never or ever smokers in the pooled data sets and found no interaction between drinking and smoking [OR (95% CI), 0.74 (0.55-0.99) for ever drinking versus never drinking among never smokers and 0.66 (0.51-0.86) for that among ever smokers, respectively].

The effect of smoking-related variables on malignant lymphoma risk was analyzed in the pooled data set of HERPACC-I/II (Table 2). Compared with never smoking, current smoking was associated with an increased risk of malignant lymphoma [OR (95% CI), 1.27 (1.03-1.57)]; this finding was consistent in both male and female subjects, albeit without statistical significance. Longer smoking duration (≥40 years) was associated with a significantly increased risk in both male and female subjects. Further, the number of pack-years was also associated with malignant lymphoma risk [OR (95% CI), reference for <5 pack-years, 1.32 (1.02-1.71) for 5-19 pack-years, 1.39 (1.07-1.80) for 20-39 pack-years, and 1.48 (1.12-1.95) for ≥40 pack-years], with P_trend for malignant lymphoma risk significant in female subjects (P_trend < 0.001) and marginally significant in males (P_trend = 0.079).

Finally, we also evaluated the effect of alcohol and smoking on the histologic subtype of malignant lymphoma (Table 3). With regard to drinking versus nondrinking, the point estimates of OR were below unity for the main malignant lymphoma subtypes in the pooled data sets. However, due to the small sample size of our study, a significant association was observed only for diffuse large B-cell lymphoma. The association with smoking was less apparent for all subtypes, except Hodgkin's lymphoma, for which a significant increase in risk was observed in heavy smokers with ≥40 pack-years [OR (95% CI), 3.24 (1.01-10.41)].

In this study, we identified an inverse association between alcohol and malignant lymphoma risk in a group of Japanese subjects. This finding appeared consistent across malignant lymphoma subtypes. Although smoking appeared to be positively associated with the risk of malignant lymphoma, this association was less clear when analyzed within subtypes. The present study confirmed our previous report in a limited number of cases in the HERPACC-I (6) and further extended the analysis with regard to the effect of malignant lymphoma subtypes.

In a large pooled analysis of nine case-control studies conducted by the International Lymphoma Epidemiology Consortium, ever drinking was associated with a lower risk of non-Hodgkin's lymphoma than never drinking [OR (95% CI), 0.83 (0.76-0.89)] and the protective effect of alcohol varying across malignant lymphoma subtypes (1). However, this pooled analysis did not include Asian populations. Further, the distribution of malignant lymphoma differs between Asian and western countries. The incidence of follicular lymphoma and marginal zone lymphoma was high, whereas that of chronic lymphocytic leukemia/small lymphocytic lymphoma and Hodgkin's lymphoma was extremely low in our cohort compared with western countries (1, 17). In the present study, we found that malignant lymphoma risk was inversely associated with drinking status [OR (95% CI), reference for never drinking and 0.65 (0.54-0.77) for ever drinking] and alcohol intake across both sexes and that this finding was consistent in the two data sets investigated. We also found OR below unity for all the major malignant lymphoma subtypes examined. These findings suggest the presence of a shared protective mechanism against the development of lymphoma irrespective of ethnic background. The potential protective effects of alcohol against lymphoma might be attributable, at least in part, to the effects of alcohol on the immune system (23). Light to moderate drinking might improve the immunologic function by increasing cellular and humoral immune responses. Moreover, light to moderate drinking is known to improve insulin sensitivity (24, 25), which might reduce the incidence of diabetes mellitus, a risk factor for malignant lymphoma (26). Compared with abstinence from alcohol, however, we found that malignant lymphoma risk was inversely associated with not only frequent and moderate drinking but also frequent and heavy drinking, suggesting that the decreased risk of lymphoma may involve other mechanisms. On the other hand, heavy drinkers might be exposed to high amounts of acetaldehyde, which plays an important role in the pathogenesis of other cancers (27). In this regard, we evaluated the effect of acetaldehyde on malignant lymphoma risk by assessing the effect of alcohol using self-reported reactions to alcohol in HERPACC-II based on the fact that flushing after drinking is mainly dependent on the speed of catalysis of acetaldehyde by aldehyde dehydrogenase 2 (28). Results showed that alcohol intake affected the risk of lymphoma irrespective of the reaction to alcohol (data not shown). This finding in turn suggests that acetaldehyde does not play a major role in the pathogenesis of lymphoma.

Although several studies have reported no association between smoking and the risk of lymphoma (3, 10, 11), the pooled analysis of the International Lymphoma Epidemiology Consortium study revealed a positive association between smoking and the risk of non-Hodgkin's lymphoma [OR (95% CI), reference for never smokers and 1.07 (1.00-1.15) for ever smokers], suggesting that smoking has carcinogenic effects (9). Consistent with this, our present study also showed a positive association between smoking exposure and lymphoma [OR (95% CI), reference for never smokers and 1.24 (1.02-1.51) for ever smokers] along with heterogeneity of this risk between each of the major histologic subtypes: follicular lymphoma, Hodgkin's lymphoma, marginal zone lymphoma, and diffuse large B-cell lymphoma. Given the significant effect observed among patients with follicular lymphoma (9, 11, 12), several studies have speculated that the inconsistent results across studies might be attributable to the different mechanisms of smoking on the etiology of each histologic subtype of lymphoma. Other studies have failed to find a significant association with follicular lymphoma (3, 10), however, and one even reported an inverse association (4). In the present study, we saw no significant association between pack-years of smoking and follicular lymphoma, but the point estimate in follicular lymphoma seemed considerably higher than that in diffuse large B-cell lymphoma, suggesting a possible association between smoking exposure and the incidence of follicular lymphoma. Consistent with other reports, we also found a significant association between Hodgkin's lymphoma risk and heavy smoking (2, 4, 7, 8). With regard to other major subtypes, we found no increased risk among patients with marginal zone lymphoma and diffuse large B-cell lymphoma. These findings relating smoking with malignant lymphoma risk and the heterogeneity of this risk among major subtypes require validation in larger studies in Asian populations.

With regard to the methodology of our study, one important factor was selection of the control base population. The control group was composed of noncancer patients who enrolled in the HERPACC-I/II study at the Aichi Cancer Center Hospital and had high response rates to questionnaires. Because our cases also belonged to this population, selection of this group warranted the internal validity of the study. Further, we confirmed previously that this population is similar to the general population in Nagoya City in terms of a range of exposures that are of interest in the HERPACC-I (20) and HERPACC-II,⁷ warranting its external validity. A second potential source of bias was the medical background of the controls, but our previous study in females showed that this had only limited effect: >66% of the noncancer outpatients at Aichi Cancer Center Hospital have no specific medical condition, whereas the remainder have benign diseases (29). The situation is comparable for men. Any bias from this issue, if present, would therefore be limited. Further, in contrast to typical hospital-based studies, the HERPACC system is less prone to information bias because the data for most or all patients were collected before diagnosis. Third, known or unknown risk factors were not considered as residual confounders in the present analysis. Fourth, because the number of cases in each stratified category was relatively small, statistical power may not be sufficient, and any interpretation of the findings in each histologic subtype should thus be undertaken with caution. Fifth, the distribution of malignant lymphoma cannot be compared between our center and the rest of Japan, because no population-based study has been conducted in Japan. However, the distribution in our center is largely compatible with the previous report of 2,260 malignant lymphoma cases collected from various regions in Japan, except for the higher frequency of marginal zone lymphoma in our center (30). Finally, the results of the pooled analysis of the two data sets should be interpreted cautiously owing to several differences in questionnaire items and definitions between the programs.

In conclusion, our results suggest that alcohol had an inverse association with the risk of malignant lymphoma in a group of Japanese subjects. This finding appeared consistent across malignant lymphoma subtypes. Although smoking appeared to be positively associated with the risk of malignant lymphoma, this association was less clear when analyzed within subtypes. Further studies are required to determine the mechanisms underlying these findings.

No potential conflicts of interest were disclosed.

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 the doctors, nurses, technical staff, and hospital administration staff at Aichi Cancer Center Hospital for the daily administration of the HERPACC study.