Relation of Cigarette Smoking to Non-Hodgkin's Lymphoma among Middle-aged Men

Transcription

1 American Journal of Epidemiology Copyright 998 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 48, No. 9 Printed in U.S.A. Relation of Cigarette Smoking to Non-Hodgkin's Lymphoma among Middle-aged Men David S. Freedman, Paige E. Tolbert, Ralph Coates, Edward A. Brann, and Carl Ft. Kjeldsberg 4 Because of previous inconsistencies in the observed relation of cigarette smoking to non-hodgkin's lymphoma, this association was investigated in the Selected Cancers Study, a population-based case-control study of,9 non-hodgkin's lymphoma cases and,90 controls, conducted between 984 and 988. Study subjects were men, and the median age of non-hodgkin's lymphoma cases was 50 years (range, -60 years). As compared with the risk among men who had never smoked cigarettes, the risk among ever smokers was not increased (odds ratio (OR) =.05, p ~ 0.50), but the risk was significantly elevated among men who reported smoking > / packs per day and among men who had smoked for 0- years (OR =.45 in each group, p < 0.05). The estimated odds ratio among the 50 heavy smokers (>50 pack-years) was.4 (95% confidence interval ) after controlling for educational achievement, various occupational and medical exposures, and other potential confounders. The observed associations, however, tended to vary by age, with the odds ratio among heavy smokers decreasing from.8 among - to 44-year-olds to. among men over 55 years of age. These age-related differences, which may account for some of the inconsistencies seen in previous studies of cigarette smoking and non-hodgkin's lymphoma, should be considered in future investigations. Am J Epidemiol 998; 48:8-4. case-control studies; lymphoma, non-hodgkin's; men; smoking Although the risk for non-hodgkin's lymphoma is greatly increased among persons with various immunodeficiencies, most cases of non-hodgkin's lymphoma cannot be attributed to these conditions (-). Other factors, such as socioeconomic status, occupations involving the use of various chemicals, radiation exposure, and military service in Vietnam, have been associated with moderate increases (odds ratio (OR) ~.5-) in the risk for non-hodgkin's lymphoma (4-8). However, the findings for many of these characteristics are quite inconsistent across studies, and the causes of most cases of non-hodgkin's lymphomas are uncertain (, ). Although the relation of cigarette smoking to non- Received for publication August 4, 997, and accepted for publication March 0, 998. Abbreviations: AIDS, acquired immunodeficiency syndrome; Cl, confidence interval; OR, odds ratio. National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, GA. Departments of Environmental and Occupational Health and of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA. National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA. 4 Department of Pathology, University of Utah Medical Center, Salt Lake City, UT. Reprint requests to Dr. David S. Freedman, Division of Nutrition and Physical Activity, Centers for Disease Control and Prevention, Mailstop K-6, 4770 Buford Highway NE, Atlanta, GA 04. Hodgkin's lymphoma has not been studied extensively, moderate (but not always statistically significant) increases in risk have been reported among various groups (8-), and some results suggest that the relation to high-grade lymphomas may be stronger (9). Furthermore, the single cohort study () that examined the relation of smoking to non-hodgkin's lymphoma found an overall relative risk of. that increased to.8 among men who smoked > pack per day. An association is biologically plausible; that is, cigarette smoking is associated with various immunologic alterations (-6), and tobacco smoke contains chemicals that are carcinogenic for lymphoid tissue (7). However, several studies (7, 8, 9), including a recent investigation of >,000 non-hodgkin's lymphoma cases (0), have found no difference in risk between smokers and nonsmokers. The goal of this case-control study was to examine the relation of various aspects of cigarette smoking, such as intensity and duration, to the risk of non- Hodgkin's lymphoma among men. MATERIALS AND METHODS Cases This case-control study was originally designed to examine the association between military service in 8

2 84 Freedman et al. Vietnam and various cancers (including non- Hodgkin's lymphoma and Hodgkin's disease). The enrollment of subjects has been described (-), and previous analyses found that men who had served in the US military in Vietnam had a modestly increased (OR =.47) risk for non-hodgkin's lymphoma (). Men were considered eligible for the study if they ) lived in an area covered by one of eight participating cancer registries; ) had an initial diagnosis of non-hodgkin's lymphoma, Hodgkin's disease, or "lymphoma, not otherwise specified" between 984 and 988; and ) were between 5 and years of age in 968. This age criterion, which ensured that men would have been eligible to have served in Vietnam, resulted in the enrollment of relatively young cases (median, 50 years; range, -60 years). A total of,54 men with an initial diagnosis of lymphoma were identified and contacted by registry personnel;,07 (88 percent) of these ascertained cases were interviewed, on average, ~00 days after diagnosis (). Data were obtained from a proxy respondent for cases who died before an interview could be conducted, and the possible effect of these proxy interviews on the observed associations was examined. Slides or tissue blocks were obtained for,004 (97 percent) men who had been interviewed, and three pathologists with expertise in the diagnosis of lymphoma reached a consensus diagnosis after independent and blinded evaluation of the specimens. The initial diagnosis was confirmed for 9 percent of the diagnoses;,5 specimens were diagnosed as non- Hodgkin's lymphoma and 4, as Hodgkin's disease; 95 percent of the unconfirmed diagnoses were due to an "inadequate specimen." Non-Hodgkin's lymphoma cases were further categorized according to the Working Formulation (, 5), a widely used grading system that is based on differences in prognosis (6). Additional exclusions were made if the subject had acquired immunodeficiency syndrome (AIDS) {n = 90), reported a previous diagnosis of lymphoma (n 0), or was missing information on smoking intensity or duration (n = 8). These 8 exclusions resulted in a final sample size of,9 non-hodgkin's lymphoma cases. were frequency matched to lymphoma (non-hodgkin's lymphoma and Hodgkin's disease) cases within each of the eight registries and within 5-year age groups. Households were selected by random digit dialing and, when the presence of an eligible control was established, personnel at the participating tumor registry conducted the interview. Information concerning the eligibility of a control was provided by 8 percent of the 5,768 contacted households, and,90 (8 percent) of the,99 identified controls completed the interview. An additional seven men were excluded because of AIDS or missing information on cigarette smoking, resulting in a final sample size of,90 controls. Because of the differing age distributions of non-hodgkin's lymphoma and Hodgkin's disease, the frequency matching resulted in the non-hodgkin's lymphoma cases' being slightly older (mean, years) than were controls. Sources of data A standardized questionnaire, which required about 50 minutes to complete, was used during the telephone interview to collect data on demographic characteristics and on suspected risk factors for the various cancers; these characteristics included various medical conditions and treatments, behaviors, and occupational exposures (). In-person interviews were conducted when necessary to secure participation (58 cases, controls). For cigarette smokers, data were obtained on the average number of packs smoked daily (intensity), cumulative number of years of smoking (duration), age at which smoking began, and for former smokers, the number of years since quitting. Several analyses focussed on pack-years, an index constructed by multiplying the number of packs smoked daily by the number of years smoked. Additional analyses examined the relative importance of smoking intensity versus duration and whether pack-years adequately summarized information on intensity and duration. Because smokers in poor health tend to stop smoking (7), most analyses considered men who had stopped smoking within the previous year to be current smokers. Of the men who had ever smoked, percent of the 556 cases and 7 percent of the 74 controls reported quitting within the last year. Statistical analyses Men were excluded from the analyses if the questionnaire or registry report suggested that they had AIDS or a related illness, a previous diagnosis of lymphoma, or if information on either smoking intensity or duration was not available; this resulted in a total of,9 cases and,90 controls for most analyses. All analyses accounted for the frequencymatched design through the use of indicator variables for the five age categories and eight registries in unconditional logistic regression models; year of birth within each age category was also included as a co-

3 Cigarette Smoking and Non-Hodgkin's Lymphoma 85 variate. Very similar results, however, were obtained using conditional logistic regression. Because the relation of smoking duration to non- Hodgkin's lymphoma tended to be nonlinear, smoking characteristics were modeled as either categorical variables or with natural splines in S-Plus (8). (Splines are piecewise polynomials that are modeled separately within regions of the predictor variable and are forced to join "smoothly.") Several characteristics, such as educational achievement and various medical conditions, were examined as potential confounders of the smoking-non-hodgkin's lymphoma association. Exact analyses were performed (9) if the number of cases or controls within an exposure category was <5. The relation of cigarette smoking to non-hodgkin's lymphoma subtypes (5) was examined through the use of polytomous logistic regression (BMDP PR) and by comparing the observed and expected numbers of heavy smokers (:50 pack-years) within each non- Hodgkin's lymphoma subgroup. The expected numbers of heavy smokers for each non-hodgkin's lymphoma subtype were first calculated within 5-year age groups and then summed over the age categories. RESULTS Nonsmokers were slightly younger (mean age, 45 years) than were former smokers (49 years) or current smokers (47 years). As compared with men who had never smoked (table ), current (but not former) smokers tended to have a slightly higher risk for non- Hodgkin's lymphoma (OR =.7, p = 0.0). Non- Hodgkin's lymphoma risk also varied by smoking intensity (average daily number of packs) and duration (years of smoking); that is, odds ratios of.45 were observed among men who smoked >!/ packs per day and among those who had smoked for 0- years. Although men who had smoked for > years TABLE. Relation of cigarette smoking to non-hodgkin's lymphoma (NHL), Selected Cancers Study, Smoking status^ Never Ever Past Current Packs per dayh < >.5 Years smoked < > Pack-years <0 (light smokers) 0-49 (moderate smokers) >50 (heavy smokers) (n=,90) 68, Cases (n =,9) No %t Adjusted odds ratio* (referent).05 (0.89-.) 0.94( ).7(0.97-.) 0.87( ) 0.90( ).05( ).09(0.8-.).6( ).45 (.0-.08) 0.79( ) 0.9( ).0(0.89-.).45(.4-.84) 0.97 ( ) 0.87( ).09(0.90-.).6 (.-.08) Linear trend.(.0-.)# **. (.09-.5)tt * Adjusted for registry and age. t Percentage of men with the specified characteristic who have NHL. $ Men were considered to be former smokers if they reported having stopped smoking > year before diagnosis (cases) or interview (controls). Forty-six men (8 controls and 8 cases) were excluded from the analysis of smoking status because they reported smoking < pack-year of cigarettes. Numbers in parentheses, 95% confidence interval. Smokers who were missing information on years smoked or packs per day have been excluded from subsequent analyses. #Per pack. **, as assessed using natural splines, the relation of smoking duration to NHL was not linear (p for nonlinear component). The fitted regression model indicated that a man who had smoked cigarettes for 5 years had an odds ratio of.8 (95% confidence interval ) as compared with a nonsmoker. tt Per 0 pack-years.

4 8 Freedman et al. were not at increased risk for non-hodgkin's lymphoma, this subgroup tended to be older (mean age, 56 years), and additional analyses indicated that the importance of smoking duration in non-hodgkin's lymphoma risk decreased with age (p ~ 0.00 for the interaction term). Figure (left panel) illustrates the interaction between smoking duration and age; that is, whereas non-hodgkin's lymphoma risk increases with smoking duration among younger (e.g., -year-old) men, there is little association among older men. For example, stratified analyses indicated that, while the odds ratio associated with smoking for 0- years was. among men ^45 years of age, it was 0.86 among men ^55 years of age. The relation of smoking duration to non-hodgkin's lymphoma was also modified by the number of packs smoked per day (middle panel, p < 0.00 for interaction). Smoking duration was not related to non-hodgkin's lymphoma risk among men who smoked ^'/ pack daily, but the estimated risk increased.7-fold for each 0 years of smoking among men who smoked ^ packs daily. Although the relation of pack-years, defined as smoking intensity (packs per day) X smoking duration (years), to non-hodgkin's lymphoma risk also tended (p ~ 0.0) to vary by age, a positive association was seen even among older men (figure, right panel). Subsequent analyses (table ) that categorized men as nonsmokers (referent group), light smokers (<0 pack-years), moderate smokers (0-49 pack-years), or heavy smokers (s50 pack-years) indicated that non-hodgkin's lymphoma risk was increased only among the latter group (OR =.6, 95 percent confidence interval (CI).-.08). Furthermore, as assessed by differences in log-likelihood ratios between models, this categorization of pack-years adequately summarized the effects of both smoking duration and intensity on non-hodgkin's lymphoma risk. Neither cigar («= 88) nor pipe (n = 45) smokers were at increased risk for non-hodgkin's lymphoma (data not shown). Bivariate analyses suggested that the age at which men began smoking cigarettes and the number of years since smoking cessation were also related to non- Hodgkin's lymphoma risk (table ). As compared with nonsmokers, for example, the odds ratio decreased from. among men who began smoking before age 5 to 0.69 among men who started smoking after age (p < 0.05 for trend). Men who began smoking at a relatively young age, however, also tended to smoke more intensely and for a longer period, and controlling for pack-years reduced the differences in odds ratios across smoking categories (final column, p = 0. for trend). Somewhat similar results were seen for years since smoking cessation. Adjustment for pack-years reduced the trend in non-hodgkin's lymphoma risk across categories of years since quitting. After having 0.5 Age (years) 0.5 i Packs per day O.Sn Age (years) X CD 0.4" \ ' 0.4- g 0.- O Q. O 0. I ' I ' I ' I ' Duration of Smoking (years) I ' rn ' Duration of Smoking (years) ' I ' I ' I ' I ' I Pack-years FIGURE. The relation of smoking duration to the risk of non-hodgkin's lymphoma, by age (left) and intensity (middle), and the relation of pack-years to non-hodgkin's lymphoma risk by age (right), Selected Cancers Study, Each panel shows the predicted probability (from logistic regression models) of non-hodgkin's lymphoma according to the characteristic shown along the x-axis for a -, 48-, or 56-year-old man (left and right panels) or for men who smoked 0.5,, or packs daily (middle panel). (The overall proportion of men with non-hodgkin's lymphoma in the study is,9/(,9 +,90) = 0..) The left panel shows that, whereas the probability of non-hodgkin's lymphoma increased with smoking duration among younger men, little association was seen among older men. The middle panel shows that the relation of smoking duration to non-hodgkin's lymphoma risk was strongest among men who smoked a packs per day. Both interactions (duration x age and duration x intensity) were statistically significant at the 0.0 level. The panel on the right shows the relation of pack-years to non-hodgkin's lymphoma (NHL) risk for men of three ages; p 0.0 for the pack-year x age product term.

5 Cigarette Smoking and Non-Hodgkin's Lymphoma 8 TABLE. Relation of age of smoking initiation and years since quitting to non-hodgkin's lymphoma, Selected Cancers Study, Never smokers Age (years) of smoking initiation < > Years since quitting Current smokers < >6 (n-,90) Cases (n =,9) No. % Mean Pack-yeats SDJ Odds ratios (adjusted for pack-years)f No Yes.0 (referent) * * * p < 0.05 as assessed by test for linear trend. When statistical significance was assessed, analyses of age at initiation were limited to men who had ever smoked, and analyses of smoking cessation were limited to former smokers. t Among smokers, the odds ratios forage at smoking initiation have been adjusted to pack-years, the mean number of pack-years among men who began smoking before age 5. Among former smokers, the odds ratios for years since smoking cessation have been adjusted to pack-years, the mean number among former smokers. $ SD, standard deviation stopped smoking for 5-9 years, former smokers had a non-hodgkin's lymphoma risk comparable to that among men who had never smoked. The relation of several potential confounders to non- Hodgkin's lymphoma risk was then examined (table ). Non-Hodgkin's lymphoma risk was found to be significantly lower among men who had completed 4 years of college than among those who did not finish high school (OR = 0.75) and among Hispanics than among whites (OR = 0.60). In contrast, the risk for non-hodgkin's lymphoma was elevated (ORs of.5-.7) among men who had never married or who had used intravenous drugs, possibly reflecting the incomplete reporting of AIDS, and among men who had been raised in a Jewish household (OR =.47, p ). In general, associations with the various medical conditions were inversely related to the prevalence of the characteristic. For example, organ transplantation and systemic lupus erythematosus increased the risk for non-hodgkin's lymphoma by more than fivefold, but each was reported by fewer than 0 men. Odds ratios of - were observed among men who reported having enlarged lymph nodes years before the interview or a first-degree relative with lymphoma or leukemia. Unexpectedly, non-hodgkin's lymphoma risk also varied according to a history of chicken pox (OR = 0.80, p = 0.0) and tonsillectomy (OR =.7, p = 0.05). Controlling for several of the demographic, behavioral, and medical characteristics shown in table reduced the odds ratio among heavy smokers from.6 (table ) to.4 (95 percent CI ); this difference was largely due to confounding by educational achievement. (Various models included educational achievement, race/ethnicity, alcohol consumption, a relative with leukemia or lymphoma, raised in a Jewish household, never having married, previously enlarged lymph nodes, tonsillectomy, chicken pox, organ transplant, systemic lupus erythematosus, and military service in Vietnam.) In addition to its relation to non-hodgkin's lymphoma risk, heavy cigarette smoking was reported by 7 percent of men who did not finish high school versus 5 percent of men who had completed ^4 years of college. To investigate whether cigarette smoking was preferentially associated with specific non-hodgkin's lymphoma subtypes, we compared the observed and expected numbers of heavy smokers within various morphologic categories (table 4). Heavy smokers tended to be overrepresented among men with either low- or high-grade lymphomas, but differences between the observed and expected cell frequencies were quite small. After adjustment for age and educational achievement, estimated odds ratios for heavy smoking and non-hodgkin's lymphoma subtypes were.65 (low grade),.4 (intermediate grade), and.7 (high grade) as compared with controls. Polytomous logistic regression analyses indicated that this difference in risk across subgroups was not statistically significant (p ~ 0.5).

6 88 Freedman et al. TABLE. Relation of potential confounders to non-hodgkin's lymphoma (NHL), Selected Cancers Study, Educational achievement! <High school High school Some college (- years) >4 years of college Ethnicity/racet White Black Hispanic Asian Alcohol consumption (oz-yr)f Nondrinker < >00 Never marriedh Intravenous drug use Raised in a Jewish household Reported medical histories Organ transplantation Systemic lupus erythematosus Enlarged lymph nodes# Relative with lymphoma Relative with leukemia Tonsillectomy Infectious mononucleosis Medical irradiation# Chicken pox Reported occupational exposures Military service in Vietnam Solvents Sprayed/mixed herbicides on farm Radiation (n=,90) , , No , Cases (n =,9) % 45* Odds ratios* (0.7-.) 0.8(0.6-.) 0.8(0.6-.0).0 0.9(0.7-.) 0.6 ( ) 0.6(0.4-.0).0 0.9(0.7-.).0(0.8-.).0(0.8-.). (0.8-.4).5(.-.0).7(.0-.9).5(.0-.) 6.5(.4-46) 5.7(.-^0).(.6-.).9(.7-5.). (.-.5).(.0-.).(0.9-.7).(0.8-.8) 0.8(0.7-.0).4(.-.9).0(0.9-.) 0.9(0.7-.) 0.8(0.6-.0) * All odds ratios have been adjusted for age and registry. If the number of either exposed controls or exposed cases was <5, estimated confidence intervals were calculated using exact (mid-p) statistical procedures. t Information on educational achievement was missing for two men, and information on oz-yr was missing for 9 men. Thirteen men who reported race/ethnicity as other than one of the four indicated categories have been excluded. (One oz-yr is the usual number of ounces of alcohol consumed per week x the number of years the men drank; oz = 9.6 ml). $ Percentage of men with specified exposure who are NHL cases. Numbers in parentheses, 95% confidence interval. H Fordichotomous characteristics, the number of controls and cases with the specified characteristic is shown; the referent group is men without the characteristic. # Enlarged lymph nodes > years before diagnosis or interview; irradiation 5 years before diagnosis or interview. DISCUSSION The results of this population-based case-control study indicate that heavy cigarette smoking may be an important risk factor for non-hodgkin's lymphoma among young and middle-aged men. Compared with the risk among men who had never smoked cigarettes, the risk for non-hodgkin's lymphoma among heavy cigarette smokers ( 50 pack-years) was.4 (95 percent Cl.-.8); this association was independent of educational achievement and several other potential confounders. This increased odds ratio, however, decreased rapidly and, after 5-9 years of smoking cessation, the risk among former smokers approached that of men who had never smoked. Several other findings,

7 Cigarette Smoking and Non-Hodgkin's Lymphoma 8 TABLE 4. Relation of cigarette smoking to non-hodgkin's lymphoma (NHL) subtypes based on the Working Formulation, Selected Cancers Study, NHL subtypes Unclassified NHL Intermediate cell} Mantle zone} Low grade (n = 4, 5%) (A)U Small lymphocytic (A) Small lymphocytic plasmacytoid Follicular (B) Small cleaved (C) Mixed, small cleaved and large cell Intermediate grade (n = 56, 47%) (D) Follicular, large cell Diffuse (E) Small cleaved (F) Mixed, small and large cell (G) Large (G) Large cleaved (G) Large noncleaved High grade (n = 07, 7%) (H) Large cell, immunoblastic (I) Lymphoblastic (J) Small, noncleaved Burkitt's %of NHL cases No. of pack-years Observed n lu Expectedt All men (nrirts ratios').0 (referent) (.-.5) (0.8-.6) (.-.8) * All odds ratios are for a heavy smoker (>50 pack-years) relative to a nonsmoker and have been adjusted for age and educational achievement. t Number of men with each NHL subtype that would be expected if heavy cigarette smoking were similarly related to all NHL subtypes. These values were calculated from the marginal distributions of cases and smokers within each 5-year age group; the (age-specific) expected numbers were then summed over the five age groups. t Generally considered low grade; not classified by the Working Formulation. Numbers in parentheses, 95% confidence interval. U Letters in parentheses, subtype names given in references (Cancer 98;49:-5) and 5 (Cancer 984; 54:45-8). such as the relation of non-hodgkin's lymphoma risk to tonsillectomy, chicken pox, educational achievement, and having been raised in a Jewish household, are of interest but need confirmation. In a recent 0-year cohort study of 49 non- Hodgkin's lymphoma deaths among 7,6 men (), odds ratios among cigarette smokers were. (overall) and.8 (> pack daily) as compared with those among nonsmokers. Although some case-control studies (8-) have also found cigarette smokers to be at slightly increased risk for non-hodgkin's lymphoma (ORs =.-.5), statistical significance (8, 0, ) or a dose-dependent relation (9, 0) has often been absent. Several other case-control studies (7, 8, 9), including one with,77 cases (0), have shown no association between cigarette smoking and non- Hodgkin's lymphoma. As many of the negative studies (7, 9, 0) had sufficient power to detect a modestly increased risk (e.g., OR =.5), it has frequently been asserted that cigarette smoking is unrelated to non-hodgkin's lymphoma (0). Although it is also possible that some of the variability in the published estimates concerning the smoking-non-hodgkin's lymphoma association may be due to chance, the effect modification by age that was observed in the current study should also be considered. The observed interaction between smoking duration and age was statistically significant at the 0.00 level, and we found that the odds ratio for heavy smokers was.6 among men <45 years of age versus. among men over 50 years of age. The average age at diagnosis for non-hodgkin's lymphoma is ~65 years (), and cases in several previous studies (7, 8, 9) have been substantially older than those in the current analyses. The single investigation that mentioned age stratification used a cutpoint of 65 years (7).

8 8 Freedman et al. Although not widely studied, age may modify the relation of cigarette smoking to various cancers. Some evidence suggests that, whereas low doses of a carcinogen are most effective in producing tumors among older animals, younger animals are more sensitive to higher doses (). Age-related differences have also been seen in observational studies, with the lung cancer risk among male cigarette smokers (vs. nonsmokers) decreasing from (among 50- to 59-year-olds) to 0 (ages years) (). Although this variability may reflect a decrease in the importance of cigarette smoking, it is also possible that the cumulative effect of other factors increases with age, or that susceptible men die at relatively young ages. The importance of smoking duration versus intensity in lung cancer risk may also differ by age (4). In addition to cancers of the respiratory tract, cigarette smokers are at increased risk for cancers of the kidney, pancreas, urinary tract, and cervix and for leukemia (0, 5, ). Although carcinogenic constituents of tobacco smoke (7) may be absorbed into the circulation, intense and prolonged antigenic stimulation, particularly when combined with immune dysfunction, may also influence non-hodgkin's lymphoma risk (, -). Cigarette smoking, which can lead to chronic inflammation of the airways, is associated with numerous changes in the immune system (), including altered CD4+ counts and T-cell function (4-6). Furthermore, in a process that parallels antigenic stimulation, a glycoprotein from the tobacco leaf can interact with a membrane component of lymphocytes (). Although based on small numbers, we found that heavy smokers were.7 times as likely to have highgrade lymphomas as were nonsmokers, and previous studies of men have also found that smoking was most strongly associated (odds ratios of. (9) and.7 (0)) with this subtype. Although the differences that we observed in non-hodgkin's lymphoma risk across subtypes were not statistically significant (p ~ 0.5), it should be emphasized that these analyses had relatively low power (~ percent power to detect an OR of.5 between non-hodgkin's lymphoma subgroups) and that the interobserver agreement in the diagnosis of high-grade lymphomas may have been only moderate (4). Because the Working Formulation is based on morphologic and prognostic differences rather than pathogenesis (6), it may be useful to include immunologic markers in future studies of non-hodgkin's lymphoma subtypes. Because of the relatively modest association that we observed between smoking and non-hodgkin's lymphoma, it is important to consider the possible effects of various biases. Participation among the controls was 8 percent and, if healthy male smokers were less likely to participate than were nonsmokers (), the observed associations could be biased upward. This overestimation could have been further exacerbated if some men with non-hodgkin's lymphoma attributed their disease to cigarette smoking, leading to an overreporting of either duration or intensity. However, the exclusion of the deceased cases did not alter our results even though proxies may have been most likely to overestimate the extent of cigarette smoking. The conflicting results of previous casecontrol studies of smoking and non-hodgkin's lymphoma suggest that these possible biases may not be widespread, and cigarette smokers were found to be at increased risk in the single cohort study that examined this association (). In summary, we found that heavy cigarette smokers were at a moderately increased risk (OR ~.4) for non-hodgkin's lymphoma as compared with nonsmokers. Although this association was independent of several potential confounders, the interactions that we observed suggest that the role of cigarette smoking in the development of non-hodgkin's lymphoma may be most important among relatively young men. As emphasized in reports of smoking and lung cancer (), possible interactions with age should be explored in future studies of non-hodgkin's lymphoma. REFERENCES. Hartge P, Devesa SS. Quantification of the impact of known risk factors on time trends in non-hodgkin's lymphoma incidence. Cancer Res 99;5(suppl):5566s-9s.. Scherr PA, Mueller NE. Non-Hodgkin's lymphomas. In: Schottenfeld D, Fraumeni JF, eds. Cancer epidemiology and prevention. nd ed. New York: Oxford University Press, 996: Tucker MA, Coleman CN, Cox RS, et al. Risk of second cancers after treatment for Hodgkin's disease. N Engl J Med 988;8: Seidman H. Cancer death rates by site and sex for religious and socioeconomic groups in New York City. Environ Res 970;: LaVecchia C, Negri E, Franceschi S. Education and cancer risk. Cancer 99;70: Hoar SW, Blair A, Holmes FF, et al. Agricultural herbicide use and risk of lymphoma and soft-tissue sarcoma. JAMA 986;56: Cartwright RS, McKinney PA, O'Brien C, et al. Non- Hodgkin's lymphoma: case-control epidemiologic study in Yorkshire. Leuk Res 988;: Franceschi S, Serraino D, Bidoli E, et al. The epidemiology of non-hodgkin's lymphoma in the northeast of Italy: a hospitalbased case-control study. Leuk Res 989; : Brown LM, Everett GD, Gibson R, et al. Smoking and risk of non-hodgkin's lymphoma and multiple myeloma. Cancer Causes Control 99;: Williams RR, Horn JW. Association of cancer sites with tobacco and alcohol consumption and socioeconomic status of patients: interview study from the Third National Cancer Survey. J Natl Cancer Inst 977,58:55-47.

9 Cigarette Smoking and Non-Hodgkin's Lymphoma 84. Nelson RA, Levine AM, Marks G, et al. Alcohol, tobacco, and recreational drug use and the risk of non-hodgkin's lymphoma. Br J Cancer 997;76:5-7.. Linet MS, McLaughlin JK, Hsing AW, et al. Is cigarette smoking a risk factor for non-hodgkin's lymphoma or multiple myeloma? Results from the Lutheran Brotherhood Cohort Study. Leuk Res 99;6:6-4.. Holt PG. Immune and inflammatory function in cigarette smokers. Thorax 987;: Tollerud DJ, Kurman CC, Nelson DL, et al. Racial variation in serum-soluble interleukin- receptor levels: a populationbased study of healthy smokers and nonsmokers. Clin Immunol Immunopathol 994;70: Mili F, Flanders WD, Boring JR, et al. The association of race, cigarette smoking, and smoking cessation to measures of the immune system in middle-aged men. Clin Immunol Immunopathol 99 ;59; Freedman DS, Flanders WD, Barboriak JJ, et al. Cigarette smoking and leukocyte subpopulations among men. Ann Epidemiol 996;6: World Health Organization. Tobacco smoking. IARC Monogr Eval Carcinog Risks Chem Hum 985;8: Paffenbarger RS, Wing WL, Hyde RT. Characteristics in youth predictive of adult-onset malignant lymphomas, melanomas, and leukemias: brief communication. J Natl Cancer Inst 978;60: Tavani A, Negri E, Franceschi S, et al. Smoking habits and non-hodgkin's lymphoma in northern Italy. Prev Med 994; : Zahm SH, Weisenburger DD, Holmes FF, et al. Tobacco and non-hodgkin's lymphoma: combined analysis of three casecontrol studies (United States). Cancer Causes Control 997; 8: Selected Cancers Cooperative Study Group. The association of selected cancers with service in the U.S. military in Vietnam. I. Non-Hodgkin's lymphoma. Arch Intern Med 990; 50:7-8.. Selected Cancers Cooperative Study Group. The association of selected cancers with service in the U.S. military in Vietnam. III. Hodgkin's disease, nasal cancer, nasopharyngeal cancer, and primary liver cancer. Arch Intern Med 990; 50: Centers for Disease Control. The association of selected cancers with service in the U.S. miliary in Vietnam: final report. Atlanta, GA: US Department of Health and Human Services, National Cancer Institute sponsored study of classifications of non-hodgkin's lymphomas: summary and description of a working formulation for clinical usage. The Non-Hodgkin's Lymphoma Pathologic Classification Project. Cancer 98; 49: Percy C, O'Conor G, Ries LG, et al. Application of the International classification of diseases for oncology (ICD-O) to the Working Formulation. Cancer 984;54: Weisenburger DD. Pathological classification of non- Hodgkin's lymphoma for epidemiological studies. Cancer Res 99;5(suppl):5456s-64s. 7. Hammond EC, Garfinkel L. The influence of health on smoking habits. Natl Cancer Inst Monog 966;9: Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluation assumptions and adequacy, and measuring and reducing errors. Stat Med 996;5: StatXact for Windows. Cambridge, MA: Cytel Software Corporation, Baron JA, Rohan TE. Tobacco. In: Schottenfeld D, Fraumeni JF, eds. Cancer epidemiology and prevention. nd ed. New York: Oxford University Press, 996: Maksymiuk AW, Bratvold JS, Ezzat W, et al. Non-Hodgkin's lymphoma in Saskatchewan. A review of 0 years' experience. Cancer 994;7:7-9.. Anisimov VN. Age-related mechanisms of susceptibility to carcinogenesis. Semin Oncol 989;6:0 9.. Thun MJ, Myers DG, Day-Lally C, et al. Age and exposureresponse relationships between cigarette smoking and premature death in Cancer Prevention Study II. In: Smoking and tobacco control, monograph 8: changes in cigarette-related disease risks and their implication for prevention and control. Washington, DC: National Cancer Institute, 997: (NIH publication no. 97-). 4. Moolgavkar SH, Dewanji A, Luebeck G. Cigarette smoking and lung cancer: reanalysis of the British doctors' data. J Natl Cancer Inst 989;8: Siegel M. Smoking and leukemia: evaluation of a causal hypothesis. Am J Epidemiol 99;8:-9.. Barton SE, Jenkins D, Cuzick J, et al. Effect of cigarette smoking on cervical epithelial immunity: a mechanism for neoplastic change? Lancet 988;: Hoover RN. Lymphoma risks in populations with altered immunity a search for mechanism. Cancer Res 99; 5(suppl):5447s-78s. 8. Magrath I. Molecular basis of lymphomagenesis. Cancer Res 99;5(suppl):559s-s.. Bahler DW, Zelenetz AD, Chen TT, et al. Antigen selection in human lymphomagenesis. Cancer Res 99;5:5547s-5s.. Francus T, Klein RF, Staiano-Coico L, et al. Effects of tobacco glycoprotein (TGP) on the immune system. II. TGP stimulates the proliferation of human T cells and the differentiation of human B cells into Ig secreting cells. J Immunol 988; : Dick F, VanLier S, Banks P, et al. Use of the Working Formulation for non-hodgkin's lymphoma in epidemiologic studies: agreement between reported diagnoses and a panel of experienced pathologists. J Natl Cancer Inst 987;78: Benfante R, Reed D, MacLean C, et al. Response bias in the Honolulu Heart Program. Am J Epidemiol 989; 0: