《alcohol consumption and all-cause mortality》由会员分享，可在线阅读，更多相关《alcohol consumption and all-cause mortality（11页珍藏版）》请在金锄头文库上搜索。
1、Addiction (1995) 90, 47180RESEARCH REPORTAlcohol consumption and all-cause mortalityJURGEN REHM Beaglehole Duffy, 1992a; Rehm, 1993). While many studies found protective effects of moderate alcohol con- sumption on the leading cause of death, cardio- vascular disease (e.g. Blackwelder et al., 1980;
2、Boffetta Cullen, KtiuimannFriedmann Gordon Gordon Klatsky era/., 1981; Kozarevic era/., 1982; Miller era/., 1990; Rehm, Fichter Rimm et al., 1991; Stampfer et al., 1988; but see Camacho, Kaplan Dyer et al., 1980), the effects of alcohol consumption on all cause mortality are less clear. However, two
3、 largeCiirrcspondence to: Dr J. Rehm, Addiction Research Foun- dation, Social Evaluation Klatsky, AmstrongRehm, 1993). (2) Confounding may also be a problem. Thus, it has been suggested that nutritional factors may be the true causal agents and not al- cohol (Kromhout Rehm Kozlowski Cox et al., 1992
4、), most of those criticisms were addressed. Thus, sample size was sufficiently high to conduct separate analyses for both sexes in two age groups (25-59 years, 60- 75 years), as well as for the three different levels of smoking status at baseline (i.e. current smoker, former smoker, never smoker). S
5、ec- ondly, NHANES 1 incorporated dietary variables as well as physical examination (e.g. serum chol- esterol, BMI) to directly and indirectly control for nutritional factors. Thirdly, history of heart disease was assessed to give a rough control for the causal link between abstaining and illness.Met
6、hods Description of study design and sample The NHEFS cohort was drawn from the sample used in the first National Health and Nutrition Examination Survey, or NHANES I. Detailed descriptions of NHANES I and the NHEFS have been published elsewhere (Miller, 1973; Engel et al, 1978; Cox et al, 1992). NH
7、ANES I was conducted in 1971-75 and involved exten- sive interviews and examinations of a representa- tive sample of the civilian, non-institutionalized, United States population 1-74 years of age. The NHEFS was designed to investigate relationships between clinical, nutritional and behavioural fac-
8、 tors measured in NHANES I (baseline) and subsequent morbidity and mortality. The NHEFS cohort consists of adults who were 25- 74 years of age at the time of NHANES I n= 14 407). Approximately 93% were success- fully traced in the first NHEFS follow-up wave conducted in 1982-84. Additional waves wer
9、e conducted in 1986 (for those aged 55 years and older at baseline) and in 1987 (for the entire cohort). In each wave, the subjects (or their proxies) were interviewed again and death certificates were obtained for decedents. Our analyses were restricted to white men andwomen (w = 12 036). This rest
10、riction was due to race-specific mortality patterns in the US during the study period (National Center for Health Statistics, 1990). Separate analyses are planned for black men and women. In analyses incorpo- rating dietary information the sample was restric- ted to the 9279 white people who partici
11、pated in the first phase of NHANES I (1971-74) in which dietary information was collected. The first phase of NHANES I was also designed to be a nationally representative sample. Approxi- mately 96% of the people eligible for analytic sample were successfully traced for vital status through 1987, le
12、aving 11600 available for analysis.Alcohol intake and covariables Alcohol intake was measured by four variables (see the Appendix). The first question asked if at least one drink of beer, wine or liquor had been consumed in the last year. People who re- sponded in the negative to this question would
13、 skip all subsequent questions and were defined as abstainers in the present survey. The following questions concerned fi-equency of drinking occa- sions, average quantity of alcohol and the kind of alcohol consumed. The two former questions were combined into an index of drinks per week which serve
14、d as a major independent variable in our analysis. For some analyses, categories for drinking behaviour were formed. All covariables were measured at baseline and were defined as follows: age in years; education as a proxy measure for social status with three levels: less than 9th grade, 9th grade t
15、o 12th grade, at least 1st year of college; smoking status with three levels: current smoker, ex-smoker, never smoker; BMI was calculated from measured height and weight:weight in kg height in m blood pressure and serum total cholesterol; percentage of total energy intake from total fat intake as de
16、rived from a 24-hour recall; and reported history of heart disease at baseline with the levels NoAes (either prior heart attack, heart failure or weak heart).Alcohol and mortality 473Statistical analyses Two different regression techniques were used to estimate relative risk by level of alcohol con- sumption: Cox proportional hazards model (1972) and logistic regression analysis (Kahn no = Number of deaths. OR (odds ratios) were derived from logistic regressi
