|Home | About | Journals | Submit | Contact Us | Français|
Several previous studies found inverse associations between alcohol consumption and risk of non-Hodgkin lymphoma (NHL) and multiple myeloma. However, most studies were retrospective, and few distinguished former drinkers or infrequent drinkers from consistent nondrinkers. Therefore, the authors investigated whether history of alcohol drinking affected risks of NHL and multiple myeloma among 102,721 eligible women in the California Teachers Study, a prospective cohort study in which 496 women were diagnosed with B-cell NHL and 101 were diagnosed with multiple myeloma between 1995–1996 and December 31, 2007. Incidence rate ratios and 95% confidence intervals were estimated using Cox proportional hazards regression. Risk of all types of B-cell NHL combined or multiple myeloma was not associated with self-reported past consumption of alcohol, beer, wine, or liquor at ages 18–22 years, at ages 30–35 years, or during the year before baseline. NHL subtypes were inconsistently associated with alcohol intake. However, women who were former alcohol drinkers at baseline were at elevated risk of overall B-cell NHL (rate ratio = 1.46, 95% confidence interval: 1.08, 1.97) and follicular lymphoma (rate ratio = 1.81, 95% confidence interval: 1.00, 3.28). The higher risk among former drinkers emphasizes the importance of classifying both current and past alcohol consumption and suggests that factors related to quitting drinking, rather than alcohol itself, may increase B-cell NHL risk.
Alcohol intake can modulate immune function (1, 2), an important etiologic factor for lymphoid malignancies (3–6). A pooled analysis of 9 case-control studies of non-Hodgkin lymphoma (NHL) revealed that ever drinkers, compared with never drinkers, had a 17% lower risk of NHL, a finding potentially explained by a beneficial effect of moderate alcohol consumption on immune function (7). However, the results showed no evident dose-response relations, possibly arguing against a biologic basis for the inverse association. Results of case-control studies of multiple myeloma, another B-cell malignancy that may be etiologically related to immune perturbation (8), have been variable, with some investigators finding no association (9–12) and others reporting at least a marginal inverse association (13–15). In 6 large prospective cohort studies of alcohol and NHL, investigators found that moderate alcohol intake was associated with a 41% reduction in risk among Iowa women (16), while heavier alcohol intake was associated with risk reductions of 33% among United Kingdom women (17), 23% among retired US men and women (18), and 40% among Japanese men (19). However, alcohol intake was not associated with NHL risk among Finnish male smokers (20) or US men and women in a cancer screening trial (21) or with multiple myeloma risk in the Finnish and Japanese cohorts (19, 20).
Recent media reports have highlighted findings of a possible protective effect of alcohol against lymphoma development (7), contributing to public confusion over whether alcohol is good or bad for general health. Thus, it is important from a public health standpoint to clarify whether modifying alcohol consumption can reduce lymphoma risk. In support of the premise that alcohol use is not protective against lymphoma, it has been documented for decades that a sizeable proportion of persons with preclinical lymphoma experience acute, intense pain and/or intolerance upon ingestion of alcohol, usually leading to voluntary cessation of alcohol consumption (22, 23). In addition, several chronic diseases are associated with increased risk of lymphoid malignancies (24–26) and can also influence lifestyle choices, including cessation of alcohol use. Therefore, failure to distinguish former drinkers from never drinkers could result in an artifactual association between current alcohol consumption and reduced lymphoma risk.
Because most previous studies have not addressed this issue, we examined the association between alcohol intake and risk of lymphoid malignancies in a large, prospective cohort study, the California Teachers Study. In the California Teachers Study, we have collected detailed data on alcohol consumption at and before cohort entry, along with data from more than a decade of follow-up.
The California Teachers Study cohort includes 133,479 active and retired female public school teachers and administrators who completed a mailed, self-administered baseline questionnaire in 1995–1996 evaluating a range of factors related to cancer risk and women's health (27). For this analysis, we sequentially excluded participants who, at baseline, were not residents of California (n = 8,867); had an unknown prior history of cancer (n = 663); consented to participate only in analyses of breast cancer (n = 18); had previously been diagnosed with NHL, multiple myeloma, Hodgkin lymphoma, or leukemia (n = 536); were aged 85 years or older (n = 2,179) or under age 30 years (n = 5,373); or had invalid or inconsistent data on alcohol intake (n = 10,664) or missing data on alcohol intake during the year before baseline (n = 2,458). Of the 102,721 remaining women included in this analysis, 496 were diagnosed with B-cell NHL (including chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL); International Classification of Diseases for Oncology, Third Edition (ICDO-3), morphology codes 9591, 9670–9699, 9727, 9823, 9832, 9835, 9836, and 9940) and 101 were diagnosed with multiple myeloma (including plasmacytoma; ICDO-3 codes 9731–9734) (28) after joining the cohort, through December 31, 2007. The 3 most common B-cell NHL histologic subtypes were diffuse large B-cell lymphoma (DLBCL) (n = 139; ICDO-3 codes 9678–9680 and 9684), follicular lymphoma (n = 111; ICDO-3 codes 9690, 9691, 9695, and 9698), and CLL/SLL (n = 111; ICDO-3 codes 9670 and 9823).
Human subjects research in this study was approved by the institutional review boards at all participating institutions.
On the baseline questionnaire, participants reported average weekly consumption (0, ≤3, 4–10, 11–17, 18–24, or ≥25 drinks per week) of beer, wine/champagne, and cocktails/liquor during the year before study entry, at ages 30–35 years, and at ages 18–22 years. They also reported how many days per week they usually drank each beverage. A typical drink was defined as 1 bottle, can, or glass of beer; 1 glass of wine or champagne or 1 wine cooler; or 1 cocktail, shot, or mixed drink of liquor. A single drink of beer, wine, or liquor was assumed to contain 13.2 g, 13.3 g, or 20.0 g of alcohol, respectively, for women aged <30 years; 13.2 g, 11.1 g, or 15.0 g of alcohol, respectively, for women aged 30–59 years; and 13.2 g, 9.2 g, or 15.0 g of alcohol, respectively, for women aged ≥60 years (29, 30). On the basis of these standards, we calculated daily intake of alcohol in grams from each type of drink during each time period, and we classified intake using cutoffs approximately equivalent to half or full drinks of wine (the most commonly consumed alcoholic beverage in the cohort) per day. Alcohol intake as reported in our food frequency questionnaire was reproducible (ρ = 0.87) and valid (ρ = 0.74) in comparison with multiple 24-hour recalls in a subset of cohort members (31).
For analyses of alcohol intake during the year before baseline, we classified women as former drinkers if they reported no baseline alcohol consumption but did report having consumed alcohol at ages 18–22 years and/or 30–35 years. Women who reported not having consumed alcohol during all 3 time periods were classified as consistent nondrinkers. For analyses of alcohol intake at ages 30–35 years, we classified women as former drinkers if they reported no alcohol consumption at that age but did report having consumed alcohol at ages 18–22 years. Women who reported not having consumed alcohol at ages 18–22 or 30–35 years but who did consume alcohol during the year before baseline were classified separately. For analyses of beer, wine, and liquor intake, we classified women as current, former, or consistent nondrinkers of that specific type of alcoholic beverage, using the same logic.
Participants were followed from the date of completion of the baseline questionnaire to the date of a first diagnosis with a hematopoietic malignancy, relocation out of California, death, or December 31, 2007, whichever occurred earliest. Participants diagnosed with T-cell NHL (n = 43), NHL of unknown histologic type (n = 22), Hodgkin lymphoma (n = 33), or leukemia (n = 116; other ICDO-3 morphology codes between 9590 and 9989) during follow-up were censored at the date of diagnosis. Information on incident cancers was obtained through annual linkage of cohort members to the population-based California Cancer Registry, which is over 99% complete. Linkages with the California state mortality file and the national Social Security Administration death master file were used to ascertain date and cause of death. Address changes were obtained through record linkages with multiple sources, including the National Change of Address database, change-of-address forms from annual mailed newsletters, and proactive notifications by participants.
Associations between alcohol intake and risk of lymphoid malignancies were estimated using Cox proportional hazards regression, with age (in days) as the time scale and data stratified by age (in years) at baseline to adjust for calendar-year effects. Incidence rate ratios were estimated as hazard ratios and 95% confidence intervals comparing former or current alcohol drinkers with consistent nondrinkers (of total alcohol or a specific alcoholic beverage type). On the basis of prior knowledge and independent associations with risk of overall B-cell NHL, NHL subtypes, or multiple myeloma, we assessed a broad range of potential confounders, including race/ethnicity, birthplace, total energy intake, intake of fruits and vegetables, body mass index, sunburn history, family history of hematopoietic cancer, personal history of melanoma or other skin cancer, number of older siblings, age at menarche, menopausal hormone therapy, pesticide/herbicide/insecticide use at various ages, urban/rural residence, and neighborhood-level socioeconomic status. None of these factors altered the associations with alcohol intake by more than 10% after multivariable adjustment; therefore, none were included in the final regression models. In models of associations with specific alcoholic beverage types, all estimates were mutually adjusted for beer, wine, and liquor intake.
The proportional hazards assumption was not violated by any of the main categories of alcohol consumption, based on significance tests of interactions between the exposure and the time scale and visual assessment of the time-to-event curves. All tests of statistical significance were 2-sided. Analyses were performed using SAS, version 9.1.3 (SAS Institute, Inc., Cary, North Carolina).
Table 1 shows the distribution of selected baseline characteristics in the study population. As shown in Table 2, which presents the associations with alcohol intake in the year before baseline, we found a statistically significant positive association between former alcohol consumption, compared with consistent nondrinking, and risk of overall B-cell NHL. We observed no association of overall B-cell NHL risk with any level of alcohol consumption in the year before baseline. Likewise, baseline and former beer, wine, and liquor consumption were not statistically significantly associated with overall B-cell NHL risk, nor was frequency of consuming each type of beverage (1–4 days/week or 5–7 days/week vs. consistent nondrinking; latter data not shown).
For the sake of comparison with results from previous studies, when we combined former drinkers with consistent nondrinkers as the reference group, the rate ratios for associations with alcohol intake in the year before baseline decreased (for baseline total alcohol intake of <5 g/day vs. never/former drinking, rate ratio (RR) = 1.00 (95% confidence interval (CI): 0.79, 1.28); for 5–<10 g/day, RR = 0.84 (95% CI: 0.65, 1.09); for 10–<20 g/day, RR = 0.86 (95% CI: 0.66, 1.11); and for ≥20 g/day, RR = 1.00 (95% CI: 0.73, 1.39)). By contrast, when we combined former drinkers with persons who consumed alcohol during the year before baseline, the rate ratio for any drinking versus consistent nondrinking was increased (RR = 1.12, 95% CI: 0.90, 1.40).
In analyses of B-cell NHL subtypes, former consumption of total alcohol or wine at baseline was associated with increased risk of follicular lymphoma (Table 2). Average intake of <10 g/day (but not ≥10 g/day) of alcohol was marginally associated with reduced risk of DLBCL, whereas any alcohol intake during the year before baseline was associated with increased risk of CLL/SLL. We observed no apparent associations between alcohol consumption in the year before baseline and risk of multiple myeloma.
As Table 3 shows, we observed no association of total alcohol consumption at ages 18–22 years or 30–35 years with risk of overall B-cell NHL, DLBCL, follicular lymphoma, or multiple myeloma, whereas consumption of any alcohol during either age interval was associated with increased risk of CLL/SLL. When we examined intake of specific alcoholic beverages at ages 18–22 years and 30–35 years, we did not detect any consistent inverse or positive associations with risk of any lymphoid malignancies (see Web Table 1, which is posted on the Journal’s Web site (http://aje.oxfordjournals.org/)). The results for alcohol intake at ages 30–35 years were unchanged when the analysis was limited to women aged ≥40 years at baseline, thereby ensuring that reported intake was at least 5 years in the past (data not shown). When we examined changes in alcohol intake across the time periods assessed, including total alcohol intake of ≥20 g/day in all 3 periods (1% of the cohort; n = 5 cases), we observed no meaningful patterns of association (data not shown).
To assess whether the positive associations of former drinking during the year before baseline with risks of overall B-cell NHL and follicular lymphoma were due to prodromal symptoms, we performed a secondary analysis after excluding the first 3 years of follow-up (n = 381 B-cell NHL cases, n = 102 DLBCL cases, n = 85 follicular lymphoma cases, n = 82 CLL/SLL cases, and n = 78 multiple myeloma cases). The results were not appreciably affected, with the exception that former alcohol consumption was associated with increased risk of CLL/SLL (RR = 2.85, 95% CI: 1.08, 7.51) as well as overall B-cell NHL and follicular lymphoma (other data not shown).
To examine associations with recent alcohol drinking, we performed a secondary analysis of baseline alcohol consumption limited to the first 5 years of follow-up (through December 31, 2000; n = 201 B-cell NHL cases). In this analysis, the relation between recent alcohol consumption and overall B-cell NHL risk showed evidence suggestive of a U-shaped curve. The rate ratios for B-cell NHL associated with total alcohol intake, versus consistent nondrinking, were 1.19 (95% CI: 0.79, 1.79) for <5 g/day, 0.67 (95% CI: 0.41, 1.09) for 5–<10 g/day, 0.77 (95% CI: 0.48, 1.24) for 10–<20 g/day, and 1.38 (95% CI: 0.84, 2.27) for ≥20 g/day.
Finally, to examine whether associations with alcohol consumption changed after exclusion of cases with rapidly fatal disease, who are less likely to be included in case-control studies, we performed a secondary analysis excluding B-cell NHL cases who died within 18 months of diagnosis (n = 88 cases). The results did not differ appreciably from those in the primary analysis (data not shown).
In this large, prospective cohort study of California women, we found little evidence that alcohol consumption during various time periods in adulthood is associated with risk of overall B-cell NHL or multiple myeloma, although we may have lacked sufficient statistical power to detect associations. We detected a weak inverse association of moderate baseline alcohol intake with risk of DLBCL, but the association was not consistent across different types of alcoholic beverages, nor did risk decline with increasing total alcohol intake, suggesting that the observed association was not due to an effect of alcohol itself. Conversely, we found positive associations of alcohol intake at all assessed time points with risk of CLL/SLL, although most of these associations lacked evidence of a dose-response trend. Thus, the observed associations with risk of DLBCL and CLL/SLL may have been due to chance or confounding. Alternatively, these associations may reflect true biologic heterogeneity between NHL subtypes and could perhaps point to an effect of alcohol on B-cell differentiation.
Most notably, we observed positive associations between former intake of alcohol in the year before baseline and risk of overall B-cell NHL, follicular lymphoma, and possibly CLL/SLL. This result was not explained by the wide array of confounders examined and may point to an etiologic role of other factors, such as illness or efforts to improve general health, that lead to cessation of alcohol use. The persistence of the positive association with former drinking after we excluded diagnoses made within 3 years of baseline indicates that such factors are unlikely to be alcohol pain/intolerance or other preclinical symptoms of lymphoma or that such symptoms occur more than 3 years before diagnosis. This general premise is consistent with our observation that recent moderate alcohol consumption (within the past 5 years) was suggestively associated with decreased risk of overall B-cell NHL, indicating that current alcohol consumption may be a surrogate for better current health.
The limited evidence of an inverse association with DLBCL risk in our study accords with the inverse association detected in an InterLymph pooled analysis of 9 NHL case-control studies (7) and several other studies (10, 12, 32–34) and in 4 of 6 prospective cohort studies of NHL (16–19). A number of biologically plausible mechanisms could underlie such a protective effect. Whereas heavy alcohol intake has been shown to have immunosuppressive effects that result in increased susceptibility to infections and impaired host response to injury (1), light or moderate alcohol consumption can have a beneficial attenuating effect on proinflammatory cytokines and chemokines (35, 36) that may otherwise promote lymphomagenesis (37). Alcohol can also increase insulin sensitivity (38), which may in turn decrease NHL risk (39). Antioxidants such as resveratrol in wine and flavonoids in beer may have additional anticarcinogenic effects (40, 41), and alcoholics have been found to have better DNA repair capacity than nonalcoholics (42).
However, our overall finding of no association of alcohol consumption with risk of B-cell NHL or multiple myeloma is more consistent with the results of previous case-control studies (9, 11, 43–47) and 2 prospective cohort studies (20, 21) that similarly detected no association. Some of the heterogeneity among previous studies may be due in part to the substantial potential for recall and selection biases in retrospective case-control studies. Given that alcohol consumption has variable connotations of social and cultural desirability (48) and is well known to affect health, cases may report alcohol intake differently from controls, resulting in recall bias. In addition, early symptoms or the diagnosis itself may cause patients to stop drinking alcohol, leading to a false inverse association between alcohol consumption and disease risk at the time of the study interview. Finally, if alcohol consumption is associated not with NHL or multiple myeloma risk but with more severe or fatal disease, then case-control studies, in which deceased patients are usually excluded and those with debilitating disease often do not participate, may again detect a spurious inverse association with alcohol consumption. Therefore, prospective cohort studies generally have higher validity for assessing the etiologic role of alcohol intake.
Nonetheless, in both retrospective and prospective studies, it is important to collect information on not only current alcohol intake but also past alcohol intake to distinguish among current, former, and never drinkers. Given that poor health is a major reason for reducing or ceasing alcohol consumption (30), former drinkers may have higher disease risk than never drinkers or current drinkers. Therefore, regardless of study design, combining former drinkers with never drinkers may produce a spurious inverse association with current drinking, whereas combining former drinkers with current drinkers may generate a false-positive association or obscure a true inverse association. Indeed, in our study, when we combined never drinkers with former drinkers as the reference group, the association with current alcohol intake was decreased, whereas combining former drinkers with current drinkers inflated the association with ever drinking. Nearly all previous studies of alcohol and risk of lymphoid malignancies combined former drinkers with either never drinkers or current drinkers, and many investigators were unclear about whether “alcohol consumption” referred to current (excluding former) drinking or ever (including former) drinking. A few studies included in the InterLymph analysis (7) classified former drinkers as a separate group and found no association between former (versus never) drinking and risk of NHL or its histologic subtypes. However, none of the previous 5 prospective cohort studies of alcohol and risk of lymphoid malignancies analyzed former alcohol intake separately.
Another potential source of misclassification in previous studies of alcohol and risk of lymphoid malignancies is a failure to distinguish between never drinkers and occasional drinkers, who can differ markedly in behaviors, health status, and—particularly if abstinence is due to religious or cultural reasons—genetic traits (30). Especially in European studies, true abstainers have often been combined with infrequent drinkers (e.g., <1 drink monthly or weekly) for comparison with frequent drinkers (10, 12, 33, 47). Such variability in the definition of the reference group for relative risk estimates may explain some of the inconsistency in results among previous studies.
In the California Teachers Study, we prospectively collected information on recent and past alcohol consumption, and our questionnaire distinguished between women who consumed low quantities of alcohol and those who did not drink at all, enabling us to classify women according to both past alcohol intake and true abstinence. This information was necessary to reveal that risks of overall B-cell NHL and follicular lymphoma differed between former drinkers and never drinkers.
However, our study had some limitations. First, we did not assess alcohol intake continuously throughout the participant's lifetime and therefore could not construct a detailed history of alcohol consumption. Second, our analyses were constrained by the limited number of incident cases, which prevented us from examining numerous categories of alcohol intake, drinking patterns over time, or associations with less common lymphoma subtypes. Notably, 2 of the previous prospective cohort studies that observed an inverse association with alcohol consumption had substantially more incident cases than we did (17, 18), giving them considerably more statistical power to detect a true association. Thus, the absence of an association with risk of overall B-cell NHL, some NHL subtypes, and multiple myeloma in our study may have been due to insufficient statistical power. Third, we did not assess reasons for cessation of alcohol consumption and therefore could not compare women who quit drinking because of ill health with those who quit for other reasons. Finally, we could not examine the effects of changes in alcohol consumption (including cessation or resumption of drinking) after baseline, which may have influenced subsequent risk of B-cell NHL or multiple myeloma.
Despite these limitations, ours is the first prospective cohort study of lymphoid malignancies to examine alcohol intake before baseline and to distinguish current drinkers from former drinkers at baseline. Our results are strengthened by detailed covariate data and complete and valid follow-up for incident cancer among California residents. In summary, we found that alcohol consumption is not associated with risk of overall B-cell NHL or multiple myeloma and that former drinkers may have higher risks of follicular lymphoma and CLL/SLL, for reasons that may be related to illness, lifestyle changes, or other factors that prompt people to stop drinking prior to diagnosis. Moreover, the relation of alcohol consumption with B-cell NHL risk may vary by histologic subtype. Future studies of the role of alcohol in the development of lymphoid malignancies or any other health outcomes should account for former drinking and changing drinking patterns over time. Only with consistent attention to these distinctions can future investigators determine whether alcohol truly protects against lymphomagenesis or whether other factors explain the inverse associations detected in past studies.
Author affiliations: Cancer Prevention Institute of California (formerly the Northern California Cancer Center), Fremont, California (Ellen T. Chang, Christina A. Clarke, Alison J. Canchola, Dee W. West, Pamela L. Horn-Ross); Division of Epidemiology, Department of Health Research and Policy, School of Medicine, Stanford University, Stanford, California (Ellen T. Chang, Christina A. Clarke, Dee W. West, Pamela L. Horn-Ross); Division of Cancer Etiology, Department of Population Sciences, City of Hope National Medical Center, Duarte, California (Yani Lu, Sophia S. Wang, Leslie Bernstein); and Division of Epidemiology, Department of Preventive Medicine, Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, Los Angeles, California (Giske Ursin).
This work was supported by the National Cancer Institute (grants R03 CA135687, R01 CA77398, and K05 CA136967) and the California Breast Cancer Research Fund (contract 97-10500). The collection of cancer incidence data used in this study was supported by the California Department of Health Services as part of the statewide cancer reporting program mandated by the California Health and Safety Code (section 103885); the National Cancer Institute's Surveillance, Epidemiology, and End Results Program (contract N01-PC-35136 with the Cancer Prevention Institute of California, contract N01-PC-35139 with the University of Southern California, and contract N02-PC-15105 with the Public Health Institute); and the Centers for Disease Control and Prevention's National Program of Cancer Registries (agreement U55/CCR921930-02 with the Public Health Institute).
The funding sources did not contribute to the design or conduct of the study or to the writing or submission of the manuscript. The ideas and opinions expressed herein are those of the authors, and endorsement by the state of California, the California Department of Health Services, the National Cancer Institute, or the Centers for Disease Control and Prevention or their contractors and subcontractors is not intended and should not be inferred.
Conflict of interest: none declared.