Alcohol Consumption and All-Cause and Cancer Mortality among Middleaged Japanese Men: Seven-year Follow-up of the JPHC Study Cohort I

Transcription

1 American Journal of Epidemiology Copyright O 1999 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved vol. 0, No. 11 Printed in USA. Alcohol Consumption and All-Cause and Cancer Mortality among Middleaged Japanese Men: Seven-year Follow-up of the JPHC Study Cohort I Shoichiro Tsugane, 1 Michael T. Fahey, 1 Satoshi Sasaki, 1 and Shunroku Baba, 2 for the JPHC Study Group* To examine the association between alcohol consumption and mortality in Japan, where mortality and lifestyle differ substantially from Western countries, a population-based prospective study was conducted in four public health center areas as part of the Japan Public Health Center-based prospective study on cancer and cardiovascular disease (JPHC). After excluding subjects with self-reported serious diseases at baseline, 19,1 men aged years who reported their alcohol intake were followed from 1990 through 199, and 54 deaths were documented. The association between all-cause mortality and alcohol consumption was J-shaped. The lowest risk was observed for men who consumed 1-9 g/week (relative risk (RR) = 0.4, 95% confidence interval (Cl) 0.4, 0.), while the highest risk was seen for men who consumed S450 g/week (RR = 1.2, 95% Cl, 1.74), after adjustment for possible confounders. The association did not change after excluding deaths that occurred in the first 2 years of follow-up. However, the association was modified by smoking, and beneficial effects of moderate drinking were largely limited to nonsmokers. The risk of cancer death showed a similar trend, but increased more in heavy drinkers. The background characteristics of moderate drinkers were healthier than either nondrinkers or heavy drinkers. The authors conclude that moderate alcohol consumption was associated with the lowest risks of all-cause and cancer mortality, especially among nonsmokers. Am J Epidemiol 1999;0:11-7. alcohol drinking; lifestyle; middle age; mortality; neoplasms; prospective studies; smoking U- or J-shaped associations between alcohol consumption and all-cause mortality have been reported in several cohort studies, and a meta-analysis of 1 cohort studies reported that relative risks were less than unity among men who consumed less than 0 g of alcohol per day (1). Most of these reports were from Western countries, and the beneficial effect of moderate alcohol consumption was attributed to reduced mortality from coronary heart disease (2-4), while a higher risk of cancer was reported with increasing level of alcohol intake (5). There is little evidence regarding the beneficial effect of alcohol consumption among other ethnic groups, whose genetic background, preferred type of Received for publication July 1, 199, and accepted for publication March, Abbreviations: Cl, confidence interval; ICD-9, International Classification of Diseases, 9th Revision; JPHC Study, Japan Public Health Center Study-based prospective study on cancer and cardiovascular diseases; PHC, public health center; RR, relative risk. 1 Epidemiology and Biostatistics Division, National Cancer Center Research Institute East, Chiba, Japan. 2 Department of Preventive Medicine, National Cardiovascular Center, Osaka, Japan. * See acknowledgments for centers and investigators included in the JPHC Study Group. Reprint requests to Dr. Shoichiro Tsugane, Epidemiology and Biostatistics Division, National Cancer Center Research Institute East, -5-1 Kashiwanoha, Kashiwa, Chiba 7, Japan. alcohol, life-style characteristics, and mortality profile differ from those in Western countries. A recent prospective study of Chinese men () showed that moderate alcohol consumption reduced the risk of total mortality compared with nondrinkers, although coronary heart disease accounted for less than percent of all deaths. MATERIALS AND METHODS Study cohort As of January 1, 1990, we established a populationbased cohort of,0 men (the Japan Public Health Center-based prospective study on cancer and cardiovascular diseases (JPHC Study) Cohort I, which also included,45 women) who registered their address in administrative districts supervised by four public health center (PHC) areas:,0 from Ninohe city and Karumai town in the Ninohe PHC area of Iwate prefecture, 7,559 from Yokote city and Omonogawa town in the Yokote PHC area of Akita,, from eight districts of Minami-Saku county in the Saku PHC area of Nagano, and 7,09 from Gushikawa city and Onna village in the Ishikawa PHC area of Okinawa. All of the men were bom between 190 and 1949 (40-59 years of age). These four PHC areas were selected to repre- 11

2 12 Tsuganeetal. sent the extent of variation in the mortality rate of stomach cancer based on our previous ecologic study. The profiles of these four areas have been reported elsewhere (7-9). Baseline survey A self-administered questionnaire was distributed to all registered residents in 1990, and they were asked to report on their sociodemographic characteristics, personal medical history, smoking and drinking history, and diet. Of these persons,,5 men (7 percent) returned their questionnaires. Although the date of questionnaire completion ranged from January 1990 to May 1992, 54 percent were between February 1990 and March Only 4 percent were after October Usual alcohol intake was first reported as frequency of consumption in six categories: <1 day/month, 1- days/month, 1-2 days/week, -4 days/week, 5- days/week, and every day. Subjects who reported alcohol consumption at least once per week were also asked to report on the usual amount and type of alcohol. Weekly ethanol consumption was calculated by combining the amount of ethanol per day and frequency per week. Amount of ethanol was calculated in grams of ethanol as follows: 0 ml sake (rice wine) as g ethanol, 0 ml shochu and awamori (white spirits) as g ethanol, 0 ml whiskey or brandy as g ethanol, 0 ml wine as g ethanol, and ml beer as g ethanol. Alcohol intake was classified into six groups: nondrinkers (<1 day per month), occasional drinkers (1- days per month), weekly ethanol intake of 1-9 g/week, g/week, g/week, >450 g/week. Because it was not possible to separate exdrinkers from nondrinkers, they were grouped together. The average levels for serum gamma-glutamyltransferase in these six groups were, 2, 9, 4, 4, and 90 IU/liter, and for high density lipoprotein (HDL) cholesterol the levels were 4, 4, 50, 51, 5, and 5 mg/dl after adjustment for age, body mass index, cigarette use, and residential area in our previous study using the same questionnaire and in the almost same areas (n = 52) (7). The reproducibility of alcohol intake as measured by Pearson's correlation coefficient between 1990 and 1995 among subsamples of the male cohort was 0.7 for calculated weekly ethanol intake (n = ). The weekly intake frequency of food items was reported in four categories: rarely, 1-2 days/week, -4 days/week, and almost every day. The weekly frequency for each food item was calculated according to a score assigned to each frequency category (0, 1.5,.5, and, respectively). For rice, miso soup, and nine kinds of beverage, the daily amount consumed was also asked. After excluding subjects with self-reported serious diseases (cancer, cerebrovascular diseases, myocardial infarction, and chronic liver diseases) at baseline, 19,1 men who reported their alcohol intake were included in this study. Follow-up We followed up all registered residents from January 1, In Japan, all death certificates are submitted to a local government office, and forwarded to the PHC in the area of residence. Mortality data is then sent centrally to the Ministry of Health, and Welfare, and coded for the National Vital Statistics. The registration of deaths is required by the Family Registration Law, and is believed to be complete in Japan. Therefore, all deaths that occurred in the cohort were based on death certificates from a PHC, except for subjects who died after they moved from their original PHC area, in which case the subject was treated as a censored case. Changes in residence status were identified through the residential registry in each area. In addition to all-cause mortality, deaths from total cancer (International Classification of Diseases, 9th Revision (ICD-9) 0-) were also analyzed. Statistical analysis Mortality rates were calculated using exact person-years of follow-up as the denominators (), and were standardized to the entire cohort. Personyears of follow-up were counted from January 1, 1990 to December 1, 199 or until the date of death or moving from a PHC area, respectively. A proportional-hazards model was used to adjust for covariates, and to assess departure from risk-ratio multiplicativity. RESULTS Baseline characteristics by alcohol Intake The baseline characteristics according to each category of alcohol intake are shown in table 1. Current smoking status was highest in heavy drinkers (>450 g/week), and lowest in moderate drinkers (1-9 g/week). The background characteristics were healthier in moderate drinkers than either nondrinkers or heavy drinkers. Characteristics of alcohol consumption by categories of intake are shown in table 2. Average frequency of alcohol intake was.2 days/week for moderate drinkers, and was almost daily for heavy drinkers. The major source of ethanol was beer in Am J Epidemiol Vol. 0, No. 11, 1999

3 Alcohol and Mortality among Japanese TABLE 1. Baseline characteristics of alcohol consumption by categories of Intake In 19,1 men, Japan, Age (years) Characteristic Occasional Weekly ethanol Intake (9/week) Nondrinker drinker (n» 97). U1U 'T', CM49 S450 (n = 2,0) (n =,41) (n =,02) (n = 2,94) (n = 2,9) Smoking history Never (%) Past(%) Current (%) Educational background College or higher (%) Sport at leisure time 1 day/week (%) Medication Any drug (%) Medical history Hypertension (%) Diet Fruit (days/week) Yellow vegetables (days/week) Fresh fish (days/week) Miso soup (bowls/week) Pickled vegetables (days/week) TABLE 2. Characteristics of alcohol consumption by categories of Intake among, male weekly drinkers, Japan, Characteristic Ethanol intake, by measure In g/week In drinks/week* In days/week Source of ethanol intake (%) Sake Shochu Beer Whisky Other Type of alcoholic beverage (%) Sake only Shochu only Beer only Whisky only Mixed 1 One drink = g of ethanol. 1-9 (n =,41) Weekly ethanol Intake (g/week) (n =,02) ^49 (n «2,94) ;>450 (n = 2,9) moderate drinkers, and sake in heavy drinkers. Shochu was another important source of ethanol in heavy drinkers. Almost half of moderate drinkers consumed beer exclusively, and a similar proportion of heavy drinkers were consumers of mixed drinks. Am J Epidemiol Vol. 0, No. 11, 1999

4 Tsugane et al. Mortality by alcohol Intake During the follow-up of 7 years, 54 men died while 09 moved out of the PHC areas. Among all causes of death, 2 (9 percent) were due to cancer and only (5 percent) were due to ischemic heart disease (ICD-9 4-4) (table ). The number of deaths, age-standardized death rates, and relative risks of death are shown in table 4 according to each category of alcohol intake. The association between all-cause mortality and alcohol consumption was J-shaped. Compared with nondrinkers, moderate drinkers had a significantly lower overall relative risk of death after adjustment for study area, age, and smoking (0.7, 95 percent confidence interval (CI) 0.49, 0.90). After further adjustment for education, medication, past history of hypertension, sports activity, intake frequency of yellow vegetables, fruits, fish, miso soup, and pickled vegetables, the relative risk was 0.4 (95 percent CI 0.4, 0.). The relative risks increased with amount of alcohol intake to 1.2 (95 percent CI, 1.74) in heavy drinkers. The risk in moderate drinkers was even lower than for occasional drinkers. The overall association with amount of alcohol intake was still J-shaped after excluding deaths in the first 2 years. The association with total cancer mortality showed a similar trend, and the lowest relative risk occurred in moderate drinkers after adjustment for study area, age, and cigarette smoking (relative risk (RR) = 0., 95 percent CI 0., 1.05) and the highest in heavy drinkers (RR = 1.55, 95 percent CI 1.02, 2.5) (table 4). The pattern of results was similar after further adjustment for dietary factors and other indicators of a healthy life-style (table 4). In order to examine whether the association between alcohol intake and mortality was modified by either PHC area of residence or current smoking status (nonsmokers vs. smokers), indicator variables for these effects and their respective product terms with the six categories of alcohol intake were included in a model adjusting for age. There was very little improvement in model fit associated with the inclusion of product terms for area of residence and subsequent models were refit without these terms. However, risk ratio estimates were indicative of heterogeneity among nonsmokers and smokers, and there was a large improvement in model fit associated with the inclusion of the product terms involving smoking and alcohol (interaction p = 0.02). Consequently, these product terms and an additional variable to adjust for the number of cigarettes smoked per day (current smoking, or previous smoking for ex-smokers) were included in the model that describes effect modification. The association between alcohol and all-cause mortality was modified by current smoking status (interac- TABLE. Causes of death In 19,1 men, Japan, All causes Cause of death Cancer Stomach Lung Colon Liver Rectum Esophagus Heart diseases Ischemic heart disease Cerebrovascular diseases Injuries and external causes Other ICD-9* code E00-E999 No ICD-9, International Classification of Diseases, 9th Revision. tion p = 0.0, table 5). Among nonsmokers (never and ex-smokers), there was a U-shaped association with lowest risks in the men who drank 1-9 g/week (RR = 0.47, 95 percent CI 0.29, 0.74) and g/week (RR = 0.51, 95 percent CI 0.2, 0.1). However, among current smokers, there was no evidence of a beneficial effect of moderate alcohol consumption and the association was J-shaped with highest risk in the men who drank 450 g/week (RR = 1.92, 95 percent CI 1., 2.71). When the model was fit with separate categories of never, ex-, and current smokers, a similar pattern of effect modification among nonsmokers was present. The risk of total cancer mortality showed a similar trend (interaction p = 0.12), but the risks were higher among men who drank g/week (RR = 2.0, 95 percent CI 1.,.0) and 450 g/week (RR = 2., 95 percent CI 1.,.7). DISCUSSION In this population-based prospective study in Japan, moderate alcohol consumption had a beneficial effect on 7-year risk of all-cause and cancer mortality. Several large-scale prospective studies in Western countries (11-) have reported similar associations with all-cause mortality, however, the beneficial effect of moderate drinking on cardiovascular disease accounted for a large part of the observed association. In this cohort, the proportion of deaths due to ischemic heart disease was only 5 percent and it is not likely to Am J Epidemiol Vol. 0, No. 11, 1999

5 Alcohol and Mortality among Japanese 15 TABLE 4. Number and relative risks (RR) (95% confidence Intervals (Cl)) for death by cause and alcohol intake categories In 19,1 men, Japan, All-cause mortality (n = 54) Age-standardized death rates per 1,000 person-years Multtvariate RR' (59/19,0) Multivariate RRf (41/,70) All-cause mortality excluding deaths In in first 2 years (n = 452) Age-standardized death rates per 1,000 person-years Mutovariate RR* (444/,97) (9/,) Total cancer mortality No. erf deaths (n = 2) Age-standardized death rates per 1,000 person-years Multivariate RR' (211/19,0) (4/,70) PtonunnKBre Occasional drinkers (0., 1.2) 04 (0.59, 1.19) (0.5, 1.0) 0.7 (0.59, 1.) (0.5, 1.55) 0.79 (0.44, 1.44) (0.49, 0.90) 0.4 (0.4, 0.) (0.4, 0.90) 0.4 (0.45, 0.92) (0., 1.05) 0.5 (0.29, 0.94) Weekly ethand intake (g/week) (0.0, 1 0) 0.7 (0.5, 1.1) (0.5, 1.0) 0. (0.,1.19) (0.54,1.2) 0.90 (0.5, 1.45) 00-^ (0.72, 1.) 1.04 (0.77, 1.41) 7 71 (0.7, 1.) 1 07 (0.77, 1.4) (0.7, 2.0) 1.4 (0.94, 2.5) (1.0,1.72) 1.2 (, 1.74) (0.99,1.7) 1.1 (0.9,1.7) (1.02,2.5) 1.54 (0.9, 2.42) pfor norhinear association * Adjusted for study area (Iwate, AWta, Nagano, Okinawa), age fn 1990 (40-44, 45-49, 50-54, years), and cigarette smoking (never, past, 1-19/day, -29/day, >O/day). t Further adjusted for educational background (junior high school, high school, college or more), medication (none, any), past history erf hypertension (no, yes), sports at leisure time (<1 day/month, 1- days/months, 1 day/week), and four categories (<1 day/week, 1-2 days/week, -4 days/week, 5 days/week) erf selected dietary habits (yellow vegetables, fruits, fish, mbo soup, pickled vegetables). TABLE 5. Relative risks (RR) (95% confidence interval (Cl)) for all-cause and cancer mortality by smoking status and alcohol Intake categories in,90 men, Japan, Type of mortality All-cause mortality (n = 55) Nonsmokers* Current smokers Total cancer mortality (n = 2) Nonsmokers* Current smokers Nondrinkers (0., 1.7) 1. (0.59, 2.04) Occasional drinkers 1.0 (0., 1.57) 0.99 (0.1, 1.0) 1.12 (0.57, 2.) 0.70 (0.2, 1.75) (0.29, 0.74) (0.71, 1.59) 0.41 (0., 0.91) 1.02 (0.52, 2.00) Weekly ethanol intake (g/week) (0.2,0.1) 1. (0.7, 1.79) 0.54 (0.2, 1.) (0.72, 2.4) (0.44, 1.) (0.99, 2.0) 0.2 (0.9,1.74) 2.0 (1.,.0) <0.001 <0.001 < (0.9, 1.7) (1.,2.71) (0.5, 2.) 7 2. (1.,.7) Includes ex- and never smokers. t Adjusted for study area (Iwate, Akita, Nagano, Okinawa), age in 1990 (40-44, 45-49, 50-54, years), and number of cigarettes smoked per day. Am J Epidemiol Vol. 0, No. 11, 1999

6 1 Tsuganeetal. be responsible for the J-shaped relation observed for all-cause mortality. An earlier large-scale cohort study in Japan (1), in which alcohol consumption had been assessed in four categories of frequency intake in 195, similarly showed a percent reduction in risk of death among occasional male drinkers. The effect of alcohol intake on total cancer mortality has been shown to be linear in Western countries (11, ). A cohort study of male Japanese physicians also showed a linear increase in risk of total cancer (). However, in our male cohort, we observed a 50 percent reduction in the risk of total cancer mortality among moderate drinkers and a 50 percent increase in risk among heavy drinkers. Heavy alcohol consumption has been associated with an increased risk of upper aerodigestive and liver cancer (5), and such cancers may have accounted in part for the increased risk among heavy drinkers. However, as the number of cancer deaths was not sufficient for site-specific analysis, we could not examine this. In addition, heavy alcohol consumption is associated with reduced activity in immunologic response (), and consequently may increase the overall risk of cancer. The reason for the beneficial effect among moderate drinkers is not clear. The effect of alcohol on high density lipoprotein (HDL) cholesterol concentration (19) and platelet function () provide mechanistic interpretations for the reduced risk of cardiovascular disease, but not for cancer. Moderate drinking may be a marker of a healthy life-style. The baseline characteristics of the men who consumed moderate amounts of alcohol were suggestive of a relatively healthier lifestyle in comparison with nondrinkers or heavy drinkers. Moderate drinkers smoked less, studied more, exercised more, and ate more fruit and yellow vegetables. Although we controlled for such factors in the multivariate model, it is still possible that residual confounding by favorable life-style factors may have resulted in the lower risk in moderate alcohol drinks. It has also been suggested that social integration confounds the J-shaped association between alcohol consumption and health status (21). In Japan, alcohol drinking is an important social event especially among men, and social networks may be better among drinkers compared with nondrinkers or heavy drinkers. It has been argued that nondrinkers include abstainers due to previous or present ill health conditions and that this is responsible for the J-shaped curves (,). Our baseline questionnaire could not separate exdrinkers from nondrinkers. In our previous ecologic study of 51 randomly selected men (aged years) from the same four areas (7, ), men (5 percent) were ex-drinkers (2 percent of 97 nondrinkers). Of these men, only were abstainers due to illness (2 diabetes mellitus, 2 gastric ulcer, 1 hepatitis, and unknown diseases). The other reasons were constitutional ( men; not likely to have been former heavy drinkers), social, and economical, and for health promotion. Therefore, we expect that the reference category includes only a small proportion of ex-drinkers who have quit due to a serious disease or who were heavy drinkers. Moreover, lower risks among moderate drinkers were still observed when occasional drinkers were used as the reference category. An analysis of several prospective studies also showed that the apparent U-shaped association was not explained by abstention as a special risk factor (24, ). In this study cohort, 42 percent of total ethanol intake was from sake and less than 0.1 percent was from grape wine. Some studies have suggested that the reduction in coronary heart disease (2, ) or all-cause mortality was specific to grape wine (), while Rimm et al. (2) concluded that ethanol itself rather than the type of beverage was responsible for the beneficial effect of alcohol. Our data, which included few grape wine drinkers, suggests that the favorable effect of alcohol intake was not necessarily due to grape wine. We observed a different effect of alcohol among current smokers and nonsmokers at baseline. A risk reduction in excess of 50 percent for all-cause and cancer mortality was found among those who consumed up to 00 g/week among nonsmokers. However, among smokers, there was no evidence of risk reduction and excess risk among those who drank >450 g/week and even among those who drank g/week for total cancer mortality. Although no studies have reported that smoking modified the U-shaped association between alcohol intake and all-cause mortality in Western countries (, 2), an effect similar to that observed in this study was shown in Chinese men () and in a study of alcohol and subjective health (29). The beneficial effect of alcohol consumption may have not compensated for the harmful effect of cigarette smoking among Oriental populations. In conclusion, among middle-aged Japanese men, moderate alcohol drinkers had the lowest risk of allcause and total cancer mortality. However, there was no strong evidence of a beneficial effect among smokers. Studies with longer follow-up may be able to analyze specific causes of death and to assess how lifetime tobacco consumption modifies the association between alcohol consumption and mortality. ACKNOWLEDGMENTS This work was supported by the grants-in-aid for Cancer Research and for the second Term Comprehensive Ten-Year Am J Epidemiol Vol. 0, No. 11, 1999

7 Alcohol and Mortality among Japanese 17 Strategy for Cancer Control from the Ministry of Health and Welfare of Japan. The authors express their appreciation to local staffs in each study area, to Drs. Shaw Watanabe and Masamitsu Konishi who contributed to the initiation of the JPHC Study, and to Dr. Walter C. Willett for his helpful comments on the manuscript. The investigators and participating institutions in the JPHC Study Cohort I, a part of JPHC Study Group (Principal Investigator Dr. S. Tsugane) were as follows: Drs. S. Tsugane, S. Sasaki, and Y. Tsubono, Epidemiology and Biostatistics Division, National Cancer Center Research Institute East, Kashiwa; Drs. J. Ogata and S. Baba, National Center for Circulatory Diseases, Suita; Drs. K. Miyakawa and F. Saito, Ninohe Public Health Center, Ninohe; Drs. Y. Miyajima and N. Suzuki, Akita prefectural Yokote Public Health Center, Yokote; Drs. H. Sanada, Y. Hatayama, F. Kobayashi, H. Uchino, and Y. Shirai, Nagano prefectural Saku Public Health Center, Saku; Drs. Y. Kishimoto, E. Takara, and M. Kinjo, Okinawa prefectural Ishikawa Public Health Center, Ishikawa; Dr. S. Matsushima, Saku General Hospital, Usuda; Drs. S. Watanabe and M. Akabane, Tokyo University of Agriculture, Tokyo; Dr. M. Konishi, Ehime University, Matsuyama; Dr. S. Tominaga, Aichi Cancer Center Research Institute, Nagoya; Dr. M. Iida, Center for Adult Diseases, Osaka; and Dr. M. Yamaguchi, National Institute of Health and Nutrition, Tokyo. REFERENCES 1. Holman CD, English DR, Milne E, et al. Meta-analysis of alcohol and all-cause mortality: a validation of NHMRC recommendations. Med J Aust 199;14: Rimm EB, Klatsky A, Grobbee D, et al. Review of moderate alcohol consumption and risk of coronary heart disease: is the effect due to beer, wine, or spirits? BMJ 199;12:71-.. Shaper AG. Alcohol and coronary heart disease. Eur Heart J 1995;1: Marmot MG. Alcohol and cardiovascular disease: the status of the U-shaped curve. BMJ 1991;0: Blot WJ. Alcohol and cancer. Cancer Res 1992;52:2119s-21s.. Yuan JM, Ross RK, Gao YT, et al. Follow up study of moderate alcohol intake and mortality among middle aged men in Shanghai, China. BMJ 1997;:-. 7. Tsugane S, Gey F, Ichinowatari Y, et al. Cross-sectional epidemiologic study for assessing cancer risks at the population level. I. Study design and participation rate. Epidemiology 1992;2: Tsugane S, Tsuda M, Gey F, et al. Cross-sectional study with multiple measurements of biological markers for assessing stomach cancer risks at the population level. Environ Health Perspect 1992;9: Tsubono Y, Kobayashi M, Tsugane S. Food consumption and gastric cancer mortality in five regions of Japan. Nutr Cancer 1997;:O-4.. Macaluso M. Exact stratification of person-years. Epidemiology 1992;: Boeffeta P, Garfinkel L. Alcohol drinking and mortality among men enrolled in an American Cancer Society Prospective Study. Epidemiology 1990; 1: Doll R, Peto R, Hall E, et al. Mortality in relation to consumption of alcohol: years' observations on male British doctors. BMJ 1994;09: Gronbaek M, Deis A, Sorensen TI, et al. Mortality associated with moderate intakes of wine, beer, or spirits. BMJ 1995; : Fuchs CS, Stampfer MJ, Colditz GA, et al. Alcohol consumption and mortality among women. N Engl J Med 1995;2: Camargo CA Jr, Hennekens CH, Gaziano JM, et al. Prospective study of moderate alcohol consumption and mortality in US male physicians. Arch Intern Med 1997;7: Hirayama T. Life-style and mortality a large-scale censusbased cohort study in Japan. Basel: Karger, Kono S, Dceda M, Tokudome S, et al. Alcohol and mortality: a cohort study of male Japanese physicians. Int J Epidemiol 19;:5-2.. Baker RC, Jerrells TR. Recent developments in alcoholism: immunological aspects. Recent Dev Alcohol 199;11: Hulley SB, Gordon S. Alcohol and high density lipoprotein cholesterol: cause inference from diverse study designs. Circulation 191;4:57-.. Renaud SC, Beswick AD, Fehily AM, et al. Alcohol and platelet aggregation: the Caerphilly Prospective Heart Disease Study. Am J Clin Nutr 1992;55: Skog OJ. Public health consequences of the J-curve hypothesis of alcohol problems. Addiction 199;91:-7.. Shaper AG. Non-drinkers shouldn't be used as baseline. (Letter). BMJ 1995;:.. Marks DF. The dose-response relation is probably linear. (Letter). BMJ 1995;: Duffy JC. Alcohol consumption and all-cause mortality. Int J Epidemiol 1995,24:0-5.. Brenner H, Arndt V, Rothenbacher D, et al. The association between alcohol consumption and all-cause mortality in a cohort of male employees in the German construction industry. Int J Epidemiol 1997 ;2: Renaud S, de Lorgeril M. Wine, alcohol, platelets, and the French paradox for coronary heart disease. Lancet 1992; 9:-.. Klatsky AL, Armstrong MA. Alcoholic beverage choice and risk of coronary artery disease mortality: do red wine drinkers fare best? Am J Cardiol 199 ;71: Gronbaek M, Deis A, Sorensen TI, et al. Influence of sex, age, body mass index, and smoking on alcohol intake and mortality. BMJ 1994;0: Poikolainen K, Vartiainen E, Korhonen HJ. Alcohol intake and subjective health. Am J Epidemiol 199; 4:4-50. Am J Epidemiol Vol. 0, No. 11, 1999