Marsha L. Eigenbrodt, 1 Thomas H. Mosley, Jr., 2 Richard G. Hutchinson, 3 Robert L. Watson, 4 Lloyd E. Chambless, 5 and Moyses Szklo 6

Transcription

1 American Journal of Epidemiology Copyright 2001 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 153, No. 11 Printed in U.S.A. Changes in Alcohol Consumption Eigenbrodt et al. Alcohol Consumption with Age: A Cross-sectional and Longitudinal Study of the Atherosclerosis Risk in Communities (ARIC) Study, Marsha L. Eigenbrodt, 1 Thomas H. Mosley, Jr., 2 Richard G. Hutchinson, 3 Robert L. Watson, 4 Lloyd E. Chambless, 5 and Moyses Szklo 6 Previous cross-sectional and longitudinal studies assessing the association between age and drinking are inconsistent. Evaluating 15,425 Black and White men and women from four communities, this study sought to determine whether there was a consistent relation between age and drinking in cross-sectional and longitudinal analyses and to determine change in drinking status and level of consumption (occasional, light to moderate, and heavier drinkers) at follow-up. Cross-sectional analyses of drinking were performed for Atherosclerosis Risk in Communities examinations 1 ( ) and 3 ( ). The changes in drinking status and level were determined for the 12,565 persons with information at both examinations. Prevalence of drinking was generally inversely associated with age in the cross-sectional analyses for all ethnic/gender groups, and drinking prevalence decreased over the 6 years of follow-up for all except Black women. Only among Black drinkers was younger age associated with a higher level of alcohol consumption in both cross-sectional and prospective analyses. Thus, whether drinking prevalence declines, the amount consumed by drinkers is decreased, or whether both factors contribute to the decrease appears to vary with ethnicity and gender. The change in drinking level was substantial with more than 40% of baseline drinkers reporting drinking cessation or a different level of consumption at follow-up. Am J Epidemiol 2001;153: aging; alcohol drinking; prevalence Alcohol consumption is a lifestyle behavior that has been extensively studied for potential benefits and risks to health (1 6). It has been suggested that older persons may be at increased risk of adverse consequences of alcohol because of a diminished volume of distribution for alcohol due to a decrease of lean body mass (7) or because of an increased sensitivity to blood alcohol levels (8). Alcohol consumption has been associated with negative outcomes such as cognitive decline, repeated falls, stroke, malnutrition, and social Received for publication January 21, 2000, and accepted for publication October 5, Abbreviations: ARIC, Atherosclerosis Risk in Communities; oz, ounce. 1 Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC. 2 Department of Medicine (Geriatrics), University of Mississippi Medical Center, Jackson, MS. 3 Department of Medicine, University of Mississippi Medical Center, Jackson, MS. 4 Division of Epidemiology, University of Mississippi Medical Center, Jackson, MS. 5 Department of Biostatistics, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC. 6 Department of Epidemiology, Johns Hopkins School of Hygiene and Public Health, Baltimore, MD. Reprint requests to Dr. Marsha L. Eigenbrodt, Department of Epidemiology, Cardiovascular Disease Program, 137 E. Franklin Street, Bank of America Center, Suite 306, University of North Carolina at Chapel Hill, Chapel Hill, NC ( marsha.eigenbrodt@sph.unc.edu). isolation when intake is heavy (9, 10) as well as beneficial effects for cardiovascular disease, resulting in a decrease in all cause mortality when intake is light to moderate (1, 3, 6). Although there are numerous studies of the association between age and alcohol consumption (11 21), research into the stability of drinking at an intraindividual level comparing cross-sectional and longitudinal associations is rather sparse and contradictory (8, 16, 22 24). Cross-sectional studies have, in general, shown a lower level of drinking at older ages (11, 13 15, 19, 25). Although some longitudinal studies have also found a decrease (8, 12, 16, 17) in either the prevalence of drinking or the amount of consumption, other longitudinal studies have reported either little influence of age on drinking change or an increase in drinking over time (22 24, 26). Adams et al. (8) suggested that some of the discrepancies in the previous studies could be due to differences in the age, gender, geographic location, or other characteristics of the populations studied, or because of the inherent biases found in cross-sectional and longitudinal studies. In addition, although investigators have indicated the importance of variation in drinking by ethnicity and gender (8, 12, 27, 28), many of the studies of intraindividual variation were limited by having a restricted population (8, 16, 22, 23, 26, 29). Evaluating drinking change is of importance to epidemiologic studies because it is often the persistence of lifestyle behaviors, such as smoking and drinking, that influences health (29). As Mulder et al. (29) pointed out, most epidemiologic studies assume a relatively stable lifestyle behavioral 1102

2 Changes in Alcohol Consumption 1103 pattern over time and do not consider fluctuations in behavior when predicting the effects of the behavior. Moreover, it has been suggested that understanding age-related drinking patterns is important because alcohol problems often correlate with the number of years of heavy drinking (19). Identifying whether drinking change occurs among persons with heavy or lighter alcohol consumption is important because a decrease in consumption by heavy drinkers would be beneficial, while drinking cessation by light drinkers might not be of benefit (2, 30, 31). In view of the aging US population, understanding the influence of age on alcohol consumption will take on increased importance for determining the possible risk and benefits of alcohol use and for estimating the potential for alcohol-related problems. The Atherosclerosis Risk in Communities (ARIC) population provides an opportunity to compare cross-sectional and longitudinal associations of aging and drinking in both Black and White men and women from several regions of the United States. MATERIAL AND METHODS Study population The ARIC Study is a prospective study of clinical and subclinical atherosclerosis in four US communities: Jackson, Mississippi; Forsyth County, North Carolina; selected suburbs of Minneapolis, Minnesota; and Washington County, Maryland. At the baseline examination, the study included population samples totaling 15,792 persons. Approximately 12 percent of the Forsyth County study population and all of the Jackson participants were Black, while the Minneapolis and Washington County cohorts were predominantly White. Details of the study design and procedures have been previously published (32). The current study includes information obtained from an initial home interview and from two clinic examinations separated by approximately 6 years: examination 1 ( ) and examination 3 ( ). Baseline exclusions from the current study included ethnicity other than Black or White (n 48), age of <45 or >64 (n 158) years, drinking status or ethanol consumption missing (n 106), and Blacks from Minneapolis or Washington County (n 55). The present analyses are based on the 15,425 Black and White participants who remained after the exclusions (full baseline population) and on a subset of 12,565 participants who had information on drinking at both examinations 1 and 3 (restricted population). Study variables Trained interviewers obtained all data. Data on ethnicity, gender, and level of education were collected at the initial home interview. Information on age, alcohol consumption, and drinker status was obtained at each examination. The main dependent variables were drinking status and grams of alcohol consumed per week. At each examination, participants were asked if they currently drank alcoholic beverages. Current drinkers (drinkers) were asked how many glasses of wine (4 ounces (oz); 1 oz 40 ml), servings of beer (12 oz), and drinks of hard liquor (1.5-oz shots) they usually consumed per week. The amount of alcohol consumed in grams per week was calculated, assuming the following alcohol content: 4 oz of wine, 10.8 g; 12 oz of beer, 13.2 g; and 1.5 oz of hard liquor, 15.1 g. A drinker who reported less than one drink per week was classified as an occasional drinker, and the alcohol consumption was recorded as zero grams per week. The level of education was used as a marker of socioeconomic status. The highest level of education completed was based on each participant s self-report. The population was divided into those with less than a high school education (did not complete high school), a high school education (completed high school, received a graduate equivalency diploma, and/or completed 1 year of vocational training), and more than a high school education (completed 1 year of college). The locations of the four field centers were evaluated as markers of community or regional effects. Statistical analyses SAS statistical software was used for all analyses (33). Descriptive statistics were determined for the baseline population, and baseline characteristics were compared for those returning for examination 3 with those lost to follow-up. Because the association between age and drinking status was not linear for all ethnic/gender groups, four age categories (45 49, 50 54, 55 59, and years) were used in the evaluation of the association. The proportion of drinkers and, among drinkers, the amount of alcohol consumed in grams per week were determined for each examination 1 age category cross-sectionally for the total baseline population and for the population restricted to participants with information on drinking at both examinations. Examination 3 crosssectional analyses were performed on the restricted subset. Logistic regression was used to determine adjusted proportions, and linear regression was used for adjusted means (34). Analyses where age was not the primary focus were adjusted to age 55, and analyses adjusting for education were adjusted to the baseline population distribution. Regional or community adjustment varied, depending on whether the analyses were for the total population or were ethnic specific. For further evaluation of the longitudinal drinking change, the examination 3 drinking category (status or level) was determined for each examination 1 category of drinking. Direct adjustment was used because of sparse data in some categories. Participants were classified as nondrinkers, occasional drinkers (drinkers who reported consuming <1 drink per week), and regular drinkers (reported consuming at least one drink per week). For all ethnic/gender groups except Black women (where numbers were too small), regular drinkers were further divided into light to moderate drinkers ( g of alcohol or 1 14 drinks per week) and heavy drinkers (more than g of alcohol or more than 14 drinks per week). RESULTS The baseline population consisted of 26.5 percent Blacks and 55.3 percent women. The population was roughly

3 1104 Eigenbrodt et al. evenly distributed across the four age categories (45 49, 50 54, 55 59, and years) and among the four ARIC communities (data not shown). Over three fourths of the population (76.4 percent) had at least a high school education. Whites (69 percent of men and 61 percent of women) reported being a drinker more often than Blacks (49 percent of men and 20 percent of women), and male drinkers reported consuming more alcohol than female drinkers (Whites: g/week for men and 39.9 g/week for women; Blacks: g/week for men and 51.5 g/week for women). Although Blacks reported being a drinker less often than Whites, Black drinkers consumed more alcohol than their White gender counterparts. Most ethnic and gender differences in drinking characteristics persisted at examination 3 (data not shown). Cross-sectional comparisons Prior to evaluating the association between age and drinking, we tested the association between drinking and two factors, educational level (as a measure of socioeconomic status) and center as a measure of community or regional differences, which previous studies had indicated were associated with drinking (13, 22). The variation in drinking across educational levels is shown in table 1. The proportion of drinkers was higher at higher levels of education at examination 1 for each ethnic/gender group in unadjusted analyses (data not shown) and after adjusting for age and community. The amount of alcohol consumed by drinkers was inversely associated with educational attainment (table 1), except among White women. The center, as a measure of community or regional differences, was also associated with significant variation in the proportion of current drinkers for both Blacks and Whites in unadjusted analyses (data not shown) and after adjusting for age and education (table 2). Jackson Black men and women reported being drinkers less often than Forsyth County Blacks, and Minneapolis Whites reported being drinkers more often than Forsyth County and Washington County Whites. The proportion of drinkers did not vary between the White centers in the South, that is, Forsyth and Washington counties. The amount of alcohol consumption reported by drinkers was not significantly different across communities, although Forsyth County Black and White participants reported slightly lower levels of consumption than the respective ethnic/gender groups from the other communities. Variation in baseline drinking prevalence and alcohol consumption across examination 1 age categories is shown in table 3. Because the distribution of current drinkers TABLE 1. Distribution of drinking prevalence and mean alcohol consumption reported by drinkers by level of education* at examination 1 ( ) for each ethnic/gender group, Atherosclerosis Risk in Communities Study Ethnicity, gender, and educational level* All <High school High school >High school Black women <High school High school >High school No. 3,633 6,305 5,464 1, Examination 1 drinkers Alcohol (g/week) % 95% CI Mean 95% CI , 46 53, 56 63, 66 15, 20 14, 20 19, , , , , , , 40.4 Black men <High school High school >High school , 50 49, 60 53, , , , White women <High school High school >High school 973 3,049 1, , 47 58, 62 70, , , , 46.8 White men <High school High school >High school 971 2,100 2, , 57 65, 69 75, 78 * Three levels based on self-report of highest level of education completed. For current drinkers. CI, confidence interval. Adjusted for age, center (community), and ethnic/gender group. Adjusted for age and center (community) , , , 102.7

4 Changes in Alcohol Consumption 1105 TABLE 2. Variation in drinking prevalence and mean alcohol consumption reported by drinkers at examination 1 ( ) for each ethnic/gender-specific community, adjusting for age and education, Atherosclerosis Risk in Communities Study Ethnicity, gender, and community Black women Jackson, MS Forsyth County, NC No. 2, Examination 1 drinkers Alcohol (g/week)* % 95% CI Mean 95% CI , 18 31, , , 53.6 Black men Jackson, MS Forsyth County, NC 1, , 54 54, , , White women Forsyth County, NC Minneapolis, MN Washington County, MD 1,841 2,040 2, , 52 81, 84 47, , , , 45.5 across age categories was similar in the adjusted and unadjusted analyses, only the adjusted values are reported. In general, for all ethnic/gender groups, the baseline prevalence of drinking was inversely associated with age (table 3). The association between the amount of alcohol consumed and age category varied, depending upon ethnicity and gender (table 3). Black women aged years reported higher levels of alcohol consumption than those over 50 years of age. For Black men, the two younger age groups reported more alcohol consumption than the older age groups. In contrast, among Whites there was no significant variation in alcohol consumption across age groups for women, and male drinkers in the youngest age category reported the lowest level of consumption. The inverse association between drinking prevalence and age described above for the entire baseline population was also present in the restricted population at both examinations 1 and 3 as seen by the lower point estimates for drinking prevalence with increasing age category for each examination (figure 1). The amount consumed by drinkers at examination 3 showed similar cross-sectional patterns as described above for baseline drinkers (data not shown). Longitudinal comparisons White men Forsyth County, NC Minneapolis, MN Washington County, MD * For current drinkers. CI, confidence interval. In unadjusted analyses, 19 percent of drinkers and 12 percent of nondrinkers reported a change in drinking status at examination 3. The proportion of examination 1 drinkers who reported not drinking at examination 3 was directly associated with examination 1 age category and ranged from 17.0 percent of drinkers aged years to 21.5 percent of drinkers aged years. The report of new drinking at examination 3 was inversely associated with age. Among 1,612 1,900 1, , 64 79, 83 62, , , , examination 1 nondrinkers, 15.9 percent of the youngest age group (aged years) and 8.2 percent of the oldest examination 1 age group (aged years) reported drinking at examination 3. Moreover, more Whites (85 percent of men and 82 percent of women) continued to report drinking at examination 3 than did Blacks (71 percent of men and 57 percent of women), and more men (14 percent of Whites and 18 percent of Blacks) who were nondrinkers at examination 1 reported drinking at examination 3 than did women (10 percent of Whites and 11 percent of Blacks). Overall, the unadjusted drinking prevalence decreased between examination 1 and examination 3 for each ethnic/gender group, except for Black women (figure 2A). The decreased drinking prevalence was found for each White examination 1 age category after adjustment for education and center but was found only for the two oldest age categories of Black men (figure 1). Among Black women, no age category had a significant change in drinking prevalence (figure 1). With the exception of White women, examination 3 drinkers reported consuming slightly less alcohol than did examination 1 drinkers (figure 2B). Figure 3 presents information on examination 3 drinking category (status and level of consumption) for each baseline drinking category by ethnicity and gender after adjustment for baseline age and educational categories. The most stable baseline drinking category was nondrinkers with percent of nondrinkers continuing to abstain at examination 3 (figure 3A). Drinking cessation was, in general, inversely associated with the baseline level of alcohol consumption for each ethnic/gender group. Drinking cessation was higher for Blacks than for Whites for a given level of consumption (figure 3A). Not shown is the fact that, of the 1,367 examination 1 drinkers who reported not drinking at follow-up,

5 1106 Eigenbrodt et al. TABLE 3. Proportion of drinkers and mean alcohol consumption reported by drinkers at examination 1 ( ) and 95% confidence interval for each ethnic/gender group by examination 1 age category, Atherosclerosis Risk in Communities Study Ethnicity, gender, and age group (years) All No. 15,402 4,118 4,040 3,799 3,445 Examination 1 drinkers* Alcohol (g/week)* % 95% CI Mean 95% CI , 63 54, 57 52, 55 50, , , , , 79.6 Black women , 31 16, 22 13, 19 9, , , , , 59.2 Black men White women White men ,624 1,614 1,467 1,274 1,214 1,340 1,430 1, percent were occasional drinkers, 34.3 percent were light to moderate drinkers, and only 8.3 percent were heavy drinkers at baseline (unadjusted values). Although the interaction was not significant at the 0.01 level, persons aged over 55 years were more likely to become nondrinkers than were persons younger than 55 years (data not shown). Of White drinkers and Black drinkers, percent and percent, respectively, continued to report the same level of consumption at examination 3 as at examination 1 (figure 3, B D). A substantial proportion of heavy drinkers became light to moderate drinkers (23 38 percent) (figure 3C), and Black male heavy drinkers aged >55 years were more likely to do so (48 percent; 95 percent confidence interval: 32, 65) than were younger drinkers (28 percent; 95 percent confidence interval: 19, 38) (not shown in figure). Relatively few heavy drinkers became occasional drinkers (2 7 percent) or nondrinkers (5 14 percent) (figure 3, B and A, respectively). Likewise, relatively few nondrinkers became occasional drinkers (8 10 percent), light to moderate drinkers (2 7 percent), or heavy drinkers (0 1 percent) 56, 65 40, 50 41, 51 39, 51 63, 68 60, 65 57, 63 55, 61 69, 74 65, 70 65, 70 64, 69 * Restricted to examination 1 drinkers. CI, confidence interval. Adjusted for ethnicity, center (community), gender, and education. Adjusted for education and ethnicity-specific center (community). (figure 3, B D). Fewer than 10 percent of drinkers within any baseline category of drinking increased to the heavy drinking category at examination 3 (figure 3D). Overall without adjustment, more persons decreased their heavy drinking (n 453) than increased into the heavy drinking category (n 286) at examination 3. DISCUSSION , , , , , , , , , , , , Previous cross-sectional studies (11, 13 15, 19, 25), as well as a study on alcohol sales (18), have found an association between older age and lower levels of alcohol consumption. However, longitudinal studies have produced varying results (8, 12, 16, 17, 22 24, 26). It has been suggested that population characteristics that are associated with differences in drinking behavior, as well as inherent differences in cross-sectional and longitudinal studies, could contribute to variations in study results (8, 16, 22). The current study provides insight into the effect of aging on drinking behavior by comparing results from the two methods. In

6 Changes in Alcohol Consumption 1107 FIGURE 1. Drinking prevalence at examination 1 (solid line) and examination 3 (dashed line) for the total population and each ethnic/gender group by examination 1 age category, for persons with information on drinking at both examinations, Atherosclerosis Risk in Communities Study, Connecting lines are used as a visual aid to demonstrate the trend with age cross-sectionally. +, adjusted for ethnicity, gender, education, and center; ++, adjusted for education and center.

7 1108 Eigenbrodt et al. FIGURE 2. Unadjusted prevalence of drinking at examinations 1 and 3 by ethnicity and gender (A) and unadjusted mean alcohol consumption in grams per week (g/wk) reported by drinkers at each examination by ethnicity and gender (B), Atherosclerosis Risk in Communities Study, Bars, confidence interval. addition, the study provides specifics about changes in status and level of alcohol consumption and the relative stability of drinking among Black and White men and women. In the total population, there were consistent associations between older age and lower drinking prevalence in crosssectional and longitudinal analyses. The lack of decrease in drinking prevalence among Black women during the 6-year follow-up may have been the result of secular trends that obscured the effect of aging (22, 35). Thus, the consistent findings for the total population and for three of the four ethnic/gender groups of the current study indicate that aging (or factors associated with aging, such as changes in health status (36)) contributes to decreased alcohol consumption through a decrease in the prevalence of drinking. In contrast, the variation in alcohol consumption by drinkers was less consistent. In cross-sectional analyses, only Black drinkers had a significantly lower level of alcohol consumption at older ages. However, in longitudinal analyses that compared examination 1 drinkers with examination 3 drinkers, all ethnic/gender groups except White women (who already consumed low levels) decreased their overall consumption. The relatively consistent lower consumption at older ages for Blacks in both cross-sectional and longitudinal studies supports aging, or age-related factors (36), as contributing to the lower consumption. Secular trends or cohort effects may have contributed to the differences in cross-sectional and longitudinal associations. For White women, the level of drinking was consistent across ages in cross-sectional analyses and changed little over time, which is consistent with the greater long-term stability among women drinkers, reported by Filmore (16). Similar to findings of other studies, Whites reported being a drinker more frequently than did Blacks (37), and male drinkers reported higher levels of alcohol consumption than did female drinkers (13, 14, 19, 38). However, community comparisons indicate that some of the ethnic variation in drinking prevalence in the current study may be due to the lower prevalence of drinking within the communities in which the Black ARIC participants resided. Regional variation in drinking prevalence has been reported by others who have suggested that social processes are involved in aggregate drinking change (13, 17, 39). We, as Klein and Pitman (40), found no cross-sectional variation in the amount of alcohol consumption reported by drinkers in different communities of the United States. In the current study, the educational level was directly associated with the

8 Changes in Alcohol Consumption 1109 FIGURE 3. Prevalence of examination 3 drinking categories by ethnicity and gender for each examination 1 drinking category, adjusted for age and education, Atherosclerosis Risk in Communities Study, *, combined light to moderate and heavy drinkers; bars, confidence interval; BF, Black females; BM, Black males; WF, White females; WM, White males. proportion of drinkers for all ethnic/gender groups. Among drinkers, a higher level of education was associated with a lower level of alcohol consumption for everyone except White women, who reported low levels of alcohol intake for all levels of education. Cahalan and Cisin (13) reported similar cross-sectional associations between education and drinking. Although the present study is in general agreement with the studies by Adams et al. (8), Fillmore (16), and Fillmore et al. (17), the findings appear to disagree with two other major longitudinal studies of intraindividual change within the United States. The Framingham Study found an increase in the percentage of drinkers and the amount of ethanol consumed over a period of 20 years (23), and the Boston Normative Study found that age was less important than generational or attitudinal influences (22); however, both studies indicate that aging tends to lower drinking relative to other age groups (22, 23). In the Boston Normative Aging Study, those over 65 years had a decrease in drinks consumed per year (22). Moreover, although all cohorts in the Framingham Study showed an increase in alcohol consumption, presumably from the relaxation of attitudes toward drinking, the rate of increase was lower for older members of the original cohort (23). The present study suggests that additional reasons for study variations include differences in ethnicity, educational attainment, and community or regional differences. It is important to note how much change in drinking categorization occurred within just 6 years and the variability across ethnic/gender groups. Although the vast majority of nondrinkers continued to abstain, only percent of White and percent of Black drinkers (depending upon baseline level of consumption) continued to drink at the same level at follow-up. These changes could be real or due, at least in part, to misclassification. Some, but not all, of the drinking changes that occurred during follow-up (if real) would be hypothesized to be beneficial to the health of the population. In absolute numbers, more persons decreased from, than increased into, the heavy drinking category (>2 drinks/day). Because recent studies indicate that consuming two or more drinks per day places a person at increased risk of death (1, 6,

9 1110 Eigenbrodt et al. 41), this change should be beneficial. However, most drinking cessation occurred among light to moderate and occasional drinkers, which could be detrimental if light to moderate drinking is beneficial as suggested by many (2, 5, 6, 42) but not all (43) studies. There are several limitations to the current study. Alcohol is self reported and so accuracy of consumption may vary with community views on drinking behavior. However, the method for ascertaining alcohol consumption in the ARIC Study, which includes questions about the frequency and amount consumed for beer, wine, and liquor separately, has been reported to yield the most realistic levels of intake (44). Retention of the ARIC cohort at examination 3 was about 86.9 percent. Of the baseline participants who would have been eligible for inclusion in the current study, there were 2,860 who did not return at examination 3. In a comparison of the group who did not return at examination 3 with those who did, the groups varied by ethnicity (43.7 percent and 22.5 percent Black, respectively), examination 1 drinking prevalence (46.4 percent and 58.0 percent), and educational level (24.5 percent and 38.0 percent with more than a high school education). Among drinkers, the amount of alcohol consumption reported at examination 1 was less for those returning at examination 3 (71.8 g/week) than for those who did not return (100.9 g/week). Although the difference in mean consumption for those lost to follow-up and those returning for examination 3 was large, there was similarity in the pattern of consumption across age categories between the entire baseline population and the population restricted to those who returned at examination 3. Whether this relative consistency would have held for the longitudinal analyses cannot be determined. Another limitation is the fact that the ARIC Study is a closed cohort. Although the comparison of cross-sectional and longitudinal analyses adds support to our findings, use of data from a closed cohort does not allow the separation of an aging effect from a secular trend in our longitudinal analyses. Finally, adjustment for three levels of education is unlikely to have removed all potential confounding associated with socioeconomic status. In conclusion, the findings of the current study indicate that aging (or factors related to aging) is associated with a decrease in drinking. Much of the decrease in the ARIC population is due to a decrease in the prevalence of drinking, although decreased consumption may play a role for Blacks. Also important is the substantial change in drinking levels that occurred within just 6 years of follow-up. Although the decrease in the prevalence of heavy drinking should benefit the population s health, the drinking cessation by light drinkers could potentially be detrimental among persons at risk for heart disease. ACKNOWLEDGMENTS Support for this research was provided by National Heart, Lung, and Blood Institute contracts N01-HC-55015, N01- HC-55016, N01-HC 55018, N01-HC-55019, N01-HC , N01-HC-55021, and N01-HC Support was also provided to the lead author (M. L. E.) by a National Heart, Lung, and Blood Institute National Research Service Award grant (5T32HL07055) from the National Institutes of Health, Cardiovascular Disease Epidemiology Training Program. The authors thank the staff and participants in the ARIC Study for their important contributions. The questionnaires/data collection forms used in the present study are available at the following web site: REFERENCES 1. Doll R, Peto R, Hall E, et al. Mortality in relation to consumption of alcohol: 13 years observations on male British doctors. BMJ 1994;309: Klatsky AL, Armstrong MA, Friedman GD. Alcohol and mortality. Ann Intern Med 1992;117: Liao Y, McGee DL, Cao G, et al. Alcohol intake and mortality: findings from the National Health Interview Surveys (1988 and 1990). Am J Epidemiol 2000;151: Marmot MG, Rose G, Shipley MJ, et al. Alcohol and mortality: a U-shaped curve. Lancet 1981;1: Rehm J, Sempos CT. Alcohol consumption and all-cause mortality. Addiction 1995;90: Thun MJ, Peto R, Lopez AD, et al. Alcohol consumption and mortality among middle-aged and elderly US adults. N Engl J Med 1997;337: Vestal RE, McGuire EA, Tobin JD, et al. Aging and ethanol metabolism. Clin Pharmacol Ther 1976;21: Adams WL, Garry PJ, Rhyne R, et al. Alcohol intake in the healthy elderly. Changes with age in a cross-sectional and longitudinal study. J Am Geriatr Soc 1990;38: Gill JS, Shipley MJ, Tsementzis SA, et al. Alcohol consumption a risk factor for hemorrhagic and nonhemorrhagic stroke. Am J Med 1991;80: Parker DA, Parker ES, Brody JA, et al. Alcohol use and cognitive loss among employed men and women. Am J Public Health 1983;73: Barnes GM. Alcohol use among older persons: findings from a western New York State general population survey. J Am Geriatr Soc 1979;27: Caetano R, Kaskutas LA. Changes in drinking patterns among Whites, Blacks, and Hispanics. J Stud Alcohol 1995;56: Cahalan D, Cisin IH. American drinking practices: summary of findings from a national probability sample. I. Extent of drinking by population subgroups. Q J Stud Alcohol 1968;29: Clapp JF, Segars LB. Alcohol consumption patterns and related problems: results of a county survey. J Community Health 1993; 18: Douglass RL. Drinking patterns and abstinence among the elderly. Int J Addict 1988;23: Fillmore KM. Women s drinking across the adult life course as compared to men s. Br J Addict 1987;82: Fillmore KM, Hartka E, Johnstone BM, et al. A meta-analysis of life course variation in drinking. Br J Addict 1991;86: Hilton ME. Trends in U.S. drinking patterns: further evidence from the past 20 years. Br J Addict 1988;83: Johnson FW, Gruenewald PJ, Treno AJ, et al. Drinking over the life course within gender and ethnic groups: a hyperparametric analysis. J Stud Alcohol 1998;59: Midanik LT, Klatsky AL, Armstrong MA. Changes in drinking behavior: demographic, psychosocial, and biomedical factors. Int J Addict 1990;25: Treno AJ, Parker RN, Holder HD. Understanding U.S. alcohol consumption with social and economic factors: a multivariate time series analysis, J Stud Alcohol 1993;54:

10 Changes in Alcohol Consumption Glynn RB, Bouchard GR, LoCastro JS, et al. Aging and generational effects on drinking behaviors in men: results from the Normative Aging Study. Am J Public Health 1985;75: Gordon T, Kannel WB. Drinking and its relation to smoking, BP, blood lipids, and uric acid. The Framingham Study. Arch Intern Med 1983;143: Klatsky AL, Siegelaub AB, Landy C, et al. Racial patterns of alcoholic beverage use. Alcohol Clin Exp Res 1983;7: Goodwin JS, Sanchez CJ, Thomas P, et al. Alcohol intake in a healthy elderly population. Am J Public Health 1987;77: Hubert HB, Eaker ED, Garrison RJ, et al. Lifestyle correlates of risk factor change in young adults: an eight-year study of coronary heart disease risk factors in the Framingham offspring. Am J Epidemiol 1987;125: Dawson DA, Grant BF, Chou SP, et al. Subgroup variation in U.S. drinking patterns: results of the 1992 national longitudinal alcohol epidemiologic study. J Subst Abuse 1995;7: Herd D. Racial differences in women s drinking norms and drinking patterns: a national study. J Subst Abuse 1997;9: Mulder M, Ranchor AV, Sanderman R, et al. The stability of lifestyle behavior. Int J Epidemiol 1998;27: Fuchs CS, Stampfer MJ, Colditz GA, et al. Alcohol consumption and mortality among women. N Engl J Med 1995;332: Maskarinec G, Meng L, Kolone LN. Alcohol intake, body weight, and mortality in a multiethnic prospective cohort. Epidemiology 1998;9: The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. The ARIC Investigators. Am J Epidemiol 1989; 129: SAS user s guide. Cary, NC: SAS Institute, Inc, Zhao DY. Logistic regression adjustment of proportions and its macro procedure. Presented at the 22nd Annual Meeting of the SAS Users Group International Conference, 1997: Robertson C, Gandini S, Boyle P. Age-period-cohort models: a comparative study of available methodologies. J Clin Epidemiol 1999;52: Eigenbrodt ML, Fuchs FD, Hutchinson RG, et al. Healthassociated changes in drinking: a period prevalence study of the Atherosclerosis Risk in Communities (ARIC) cohort ( ). Prev Med 2000;31: Herd D, Grube J. Black identity and drinking in the US: a national study. Addiction 1996;91: Dawson DA. Patterns of alcohol consumption: beverage effects on gender differences. Addiction 1993;88: Hupkens CLH, Knibbe RA, Drop MJ. Alcohol consumption in the European community: uniformity and diversity in drinking patterns. Addiction 1993;88: Klein H, Pittman DJ. Regional differences in alcohol consumption in drinkers attitudes toward drinking. Am J Drug Alcohol Abuse 1993;19: Camargo CAJ, Hennekens CH, Gaziano JM, et al. Prospective study of moderate alcohol consumption and mortality in US male physicians. Arch Intern Med 1997;157: Nanchahal K, Ashton W, Wood DA. Alcohol consumption, metabolic cardiovascular risk factors and hypertension in women. Int J Epidemiol 2000;29: Shaper AG, Wannamethee SG. Alcohol intake and mortality in middle aged men with diagnosed coronary heart disease. Heart 2000;83: Fuenekes GI, van t Veer P, van Staveren WA, et al. Alcohol intake assessment: the sober facts. Am J Epidemiol 1999;150: