Holahan CJ, Schutte KK, Brennan PL, North RJ, Holahah CK, Moos BS, Moos RH. Wine consumption and 20-year mortality among late-life moderate drinkers. J Stud Alcohol Drug 2012; 73: 80–88.
Objective: This study examined level of wine consumption and total mortality among 802 older adults ages 55–65 at baseline, controlling for key sociodemographic, behavioral, and health status factors. Despite a growing consensus that moderate alcohol consumption is associated with reduced total mortality, whether wine consumption provides an additional, unique protective effect is unresolved.
Method: Participants were categorized in three subsamples: abstainers, high-wine consumption moderate drinkers, and low-wine-consumption moderate drinkers. Alcohol consumption, sociodemographic factors, health behavior, and health problems were assessed at baseline; total mortality was indexed across an ensuing 20-year period.
Results: After adjusting for all covariates, both high-wine-consumption and low-wine-consumption moderate drinkers showed reduced mortality risks compared with abstainers. Further, compared with moderate drinkers for whom a high proportion of ethanol came from wine, those for whom a low proportion of ethanol came from wine were older, were more likely to be male, reported more health problems, were more likely to be tobacco smokers, scored lower on socioeconomic status, and (statistical trend) reported engaging in less physical activity. Controlling only for overall ethanol consumption, compared with moderate drinkers for whom a high proportion of ethanol came from wine, those for whom a low proportion of ethanol came from wine showed a substantially increased 20-year mortality risk of 85%. However, after controlling for all covariates, the initial mortality difference associated with wine consumption was no longer significant.
Conclusions: Among older adults who are moderate drinkers, the apparent unique effects of wine on longevity may be explained by confounding factors correlated with wine consumption.
Background: There are consistent data showing that moderate consumers of alcohol have lower risk of cardiovascular disease and many other diseases of ageing, as well as lower risk of mortality, than do abstainers.1,2 Experimental data in animals and humans have defined a large number of mechanisms for such an effect.3 There is still some inconsistency as to differential effects according to the type of beverage consumed. In most epidemiologic studies, wine consumers have been shown to have higher levels of education and income, to consume a healthier diet, and have other characteristics that are associated with better health outcomes than consumers of other beverages.4,5 A recent meta-analysis by Constanzo et al6 showed that moderate consumers of both wine and beer had lower risk of cardiovascular disease than did people who generally consumed spirits.
Experimental data clearly show that polyphenols and other constituents in wine, in addition to the alcohol, have beneficial effects on cardiovascular risk in animals, including humans.3, 7-9 The question is whether epidemiologic studies comparing people who consume certain beverages (rather than comparing the beverages themselves) demonstrate such differences in terms of health effects.
A Forum reviewer points out that studies often show that “it is the type of alcoholic beverage which is consumed most frequently in a population which exerts the clearest protective effect. For example, in France it is moderate wine consumption and in Germany moderate beer consumption associated with the healthiest outcomes.10-11 Wine consumers in a customary non-wine drinking country like Denmark may be especially different from the general population. Hence, attempting to adjust for the potential confounding by other lifestyle factors is an ongoing challenge for epidemiologists who are seeking to determine if the consumption of one type of beverage containing alcohol has different effects on health from the consumption of other types of beverage.”
Comments on the present paper: This study of older American adults seems to have used traditional statistical methods within a logical analytic plan. The authors report data among “moderate wine drinkers,” those consuming 1 to <3 drinks/day. Those consuming up to one-third of their total alcohol as wine were classified as low-wine consumers (50% of moderate drinkers) and those with more than 30% of their total alcohol intake as wine were classified as high-wine consumers (31% of moderate drinkers). (Evidently, the other 19% of moderate drinkers did not consume wine.) No data were available on the pattern of drinking.
Among the weaknesses of the study is that alcohol consumption and covariates were assessed only at baseline, and changes that may have occurred during the 20-year observation period are not accounted for. Exposure variables assessed only once at baseline may change considerably over time.12 Most often this changing of exposure over time produces a dilution effect; in this case, the mortality-decreasing effect of alcohol would be diminished. As shown in the MONICA Augsburg cohort 1987-1997, “Hazard rate ratios for alcohol intake classified by two assessments consistently revealed a more pronounced beneficial effect of alcohol consumption than those for alcohol intake groups based on a single measurement.”13 Data on the pattern of drinking were not available in the present study, which could be important as most health outcomes are favourably associated with regular moderate consumption rather than episodic drinking, such as week-end binge drinking. Further, the authors of this paper did not have data on diet, which differs between wine drinkers and other drinkers,14
The authors state that adjustments were made for “health problems;” if these included diabetes and cardiovascular diseases, adjusting for these conditions may have diluted the effect of alcohol on total mortality (akin to adjusting for intermediate variables). It is interesting that covariates such as gender, marital status, former problem drinking, obesity, depressive symptoms, avoidance coping, number of close friends, and quality of friend support were all non-significant in the analysis.
Repeatedly the quality of the assessment of alcohol intake is discussed and the authors make the statement ” . . . future research would be strengthened by including objective indices or collateral information on alcohol consumption.” However, these “objective indices” are not available for epidemiological studies on alcohol and longevity, and therefore we must rely on self-reports and validations of these self-reports. This has been done many times and the validity of the self-reports are quite good and reliable, at least in cohorts from general populations.
It is unclear what the rationale is for adjusting for total ethanol intake when assessing wine intake in relation to mortality; this seems to be testing if the other components in wine, rather than ethanol, affect mortality. If we are interested in the total effect of wine on mortality, it may be preferable not to adjust for total alcohol, but perhaps present stratified results according to total alcohol.
Forum reviewers were concerned that the results of the analyses are not sufficiently detailed in the paper. One would prefer if the investigators could provide the data on whether alcohol effect varied by sex, given that sex is associated with death and women may have different drinking patterns from men. (In the current study, the number of women may have been too small.) In any case, it is always a good approach to perform stratified analysis, making sure there the findings are consistent across sub-categories.
Difficulties in studying beverage-specific effects: A Forum reviewer commented: “It is difficult to study beverage choice in a population with non-homogeneous usual choice. The method used in this article dichotomizes moderate drinkers into persons drinking largely wine and persons drinking largely other beverage types. Thus the comparison appears to be between wine drinkers and those drinking mostly beer and liquor. This type of comparison maximizes the confounding issue, so a large difference in potential confounders is plausible.
“That said, the truly huge differences in traits between drinkers of more versus less wine in this study population seem out of line. Also, as epidemiological studies go, this cohort of 802 persons is pretty small., and some of those 802 were excluded from the analysis. Going back to the 1990 Brennan & Moos article15 cited as describing the study cohort it is apparent that they are not a representative cross section of the population. In fact, a large proportion were considered to be problem drinkers. How this interacts with beverage choice is not described. Further, it is not specified the apparent causes of death in the population, or how many died of alcohol-related illnesses in each group.”
Differences between low-wine and high-wine groups: As stated, there were large differences in lifestyle habits between low-wine and high-wine groups (e.g., for the high-wine group, there were more females, much higher SES, fewer health problems, and a strikingly lower rate of smoking). With such large differences between groups, it is not surprising that when the authors adjusted for all of these factors, there were large changes in the estimates of effect on mortality. This raises the question as to whether or not there may have been over-adjustment for confounders in the multivariable analyses. (An example might be adjusting for diabetes and coronary disease, which are probably important intermediary factors in the effects of drinking on mortality.)
While editors generally try to shorten manuscripts and prefer few tables and figures, it would have been helpful to have more detailed data from this analysis. Without further data on how each covariate modified the final equation, it is difficult to know how important the level of wine consumption was. It would have been useful to see stratified results according to gender (even though the statement by the authors that an OR of 0.70 was not statistically significant; exactly what was being compared is not described). Also, it would be interesting to see stratified results by levels of SES, and particularly comparing effects among smokers versus non-smokers.
References from Forum review
1. Holahan CJ, Schutte KK, Brennan PL, Holahan CK, Moos BS, Moos RH. Late-life alcohol consumption and 20-year mortality. Alcoholism: Clinical and Experimental Research 2011;34:1961-1971.
2. Ronksley PE, Brien SE, Turner BJ, Mukamal KJ, Ghali WA. Association of alcohol consumption with selected cardiovascular disease outcomes: a systematic review and meta-analysis. BMJ 2011;342:d671; doi:10.1136/bmj.d671
3. Brien SE, Ronksley PE,Turner BJ, Mukamal KJ, Ghali WA. Effect of alcohol consumption on biological markers associated with risk of coronary heart disease: systematic review and meta-analysis of interventional studies. BMJ 2011;342:d636; doi:10.1136/bmj.d636.
4. Grønbæk, M., Deis, A., Sorenson, J.A., Becker, U., Schnohr, P., Jensen, G. Mortality associated with moderate intake of wine, beer, or spirits. BMJ 1995;310:1165–1169.
5. Klatsky, A.L., Friedman, G.D., Armstrong, M.A., Kipp, H., 2003. Wine, liquor, beer, and mortality. Am J Epidemiol 2003;158:585–595.
6. Costanzo S, Di Castelnuovo A, Donati MB, Iacoviello L, de Gaetano G. Wine, beer or spirit drinking in relation to fatal and non-fatal cardiovascular events: a meta-analysis. Eur J Epidemiol 2011;DOI 10.1007/s10654-011-9631-0
7. Huang PH, Chen YH, Tsai HY, Chen JS, Wu TC, Lin FY, Sata M, Chen JW, Lin SJ. Intake of red wine increases the number and functional capacity of circulating endothelial progenitor cells by enhancing nitric oxide bioavailability. Arteriosclerosis Thrombosis and Vascular Biology 2010;30:869-877.
8. Zoechling A, Liebner F, Jungbauer A. Red wine: A source of potent ligands for peroxisome proliferator-activated receptor γ. Food & Function, Journal of the Royal Society of Chemistry 2011;2:28-38.
9. Dal-Ros S, Zoll J, Lang AL, Auger C, Keller N, Bronner C, Geny B, Schini-Kerth VB. Chronic intake of red wine polyphenols by young rats prevents aging-induced endothelial dysfunction and decline in physical performance: Role of NADPH oxidase. Biochem Biophys Res Commun 2011;404:743-749.
10. Brenner H, Rothenbacher D, Bode G, März W, Hoffmeister A, Koenig W. Coronary Heart Disease Risk Reduction in a Predominantly Beer-Drinking Population. Epidemiology 2001;12:390-395.
11. Keil U, Chambless LE, Döring A, Filipiak B, Stieber J. The relation of alcohol intake to coronary heart disease and all-cause mortality in a beer-drinking population. Epidemiology 1997;8:150-156.
12. Molander RC, Yonker JA, Krahn DD. Age-related changes in drinking patterns from mid- to older age: results from the Wisconsin Longitudinal Study. Alcoholism: Clinical and Experimental Research 2010;34:1182-1192.
13. Wellmann J, Heidrich J, Berger K, Döring A, Heuschmann PU, Keil U. Changes in alcohol intake and risk of coronary heart disease and all-cause mortality in the MONICA/KORA-Augsburg cohort 1987–97. Eur J Cardiovasc Prevention Rehab 2004;11:48–55.
14. Johansen D, Friis K, Skovenborg E, Grønbæk M. Food buying habits of people who buy wine or beer: cross sectional study. BMJ 2006;332:519-522.
15. Brennan PL, Moos RH. Life stressors, social resources, and late-life problem drinking. Psychology and Aging 1990;5:491-501.
Wine consumers, especially in comparison with spirits drinkers, have been shown to have higher levels of education and income, to consume a healthier diet, be more physically active, and have other characteristics that are associated with better health outcomes. However, epidemiologic studies have been inconsistent in showing that, after adjustment for all associated lifestyle factors, consumers of wine have lower risk of cardiovascular disease and mortality than do consumers of other beverages.
The present study concluded that the associated lifestyle habits and environmental factors of wine consumers largely explain their better health outcomes. Forum reviewers were concerned about some of the methodological approaches used, and believed that the data presented in the paper were inadequate to support such a conclusion. This was a small study, had only a single estimate of alcohol intake (at baseline but not throughout 20 years of follow up), and the authors may have over-adjusted for large differences in lifestyle factors between what they termed as “low-wine” and “high-wine” consumers. The study did confirm a lower mortality risk for alcohol consumers than for non-drinkers.
Experimental studies have clearly indicated that the polyphenols and other constituents that are present in wine and some beers have independent protective effects against most cardiovascular risk factors. Whether or not such advantages are seen among moderate drinkers of wine (or beer) in epidemiologic studies is difficult to determine, as comparisons are not being made between wine, beer, and spirits but between humans who consume one or other such beverage. In almost all populations, drinkers of a specific type of beverage differ in many ways other than just the type of beverage they consume.
* * *
Comments on this paper have been provided by the following members of the International Scientific Forum on Alcohol Research:
Yuqing Zhang, MD, DSc, Epidemiology, Boston University School of Medicine, Boston, MA, USA
Gordon Troup, MSc, DSc, School of Physics, Monash University, Victoria, Australia
Erik Skovenborg, MD, Scandinavian Medical Alcohol Board, Practitioner, Aarhus, Denmark
Arthur Klatsky, MD, Dept. of Cardiology, Kaiser Permanente Medical Center, Oakland, CA, USA
Ulrich Keil, MD, PhD, Institute of Epidemiology and Social Medicine, University of Münster, Münster, Germany
Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA
R. Curtis Ellison, MD, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA