Critique 195: How much of the “health benefits” of alcohol relate to higher socio-economic status of drinkers? — 19 December 2016

Towers A, Philipp M, Dulin P, Allen J. The “Health Benefits” of Moderate Drinking in Older Adults may be Better Explained by Socioeconomic Status.  Pre-publication: J Gerontol B Psychol Sci Soc Sci 2016. doi:10.1093/geronb/gbw152

Authors’ Abstract

Objectives: To assess whether a relationship between alcohol use and health exists for older adults before and after controlling for proxy and full indicators of socioeconomic status (SES).

Method: Secondary analysis of data from 2,908 participants in the New Zealand Longitudinal Study of Ageing (2012) completing measures of alcohol use, health, SES proxies (income, education) and SES. Sample mean age was 65, 52% were female, more than 80% were drinkers, and more than 75% had educational qualifications.

Results: Moderate drinkers had better health and SES than heavier or nondrinkers. The positive influence of moderate alcohol consumption on health was observed for men and women when controlling for SES proxies, but was substantially reduced in women and completely disappeared for men when controlling for full SES.

Discussion: SES plays a key role in presumed “health benefits” of moderate alcohol consumption for older adults. It accounts for any alcohol–health relationship in a sample of men of whom 45% consume at least one drink daily, and substantially attenuates the association between alcohol and health in a sample of women who are not frequent drinkers. Prior research may have missed the influence of SES on this alcohol–health relationship due to the use of incomplete SES measures.

Forum Comments

It has long been recognized that both the positive and negative health effects of alcohol are modified by the socio-economic status (SES) of individuals. Truly light-to-moderate drinkers consistently are shown to be more highly educated that non-drinkers and heavy drinkers, and also tend to have higher incomes and job classifications.

The mechanisms that underlie such findings are poorly understood, although many studies have shown that SES is associated with differences in the actual alcohol intake among people who self-reported themselves as “moderate” drinkers. Lower SES subjects more often under-report their intake and are more frequently binge drinkers than subjects with high measures of SES.  Thus, the “exposure” (level of alcohol intake) in epidemiologic studies is usually shown to be affected by SES; further, the “outcome” in such studies (indices of health) clearly show that lower SES subjects have greater morbidity and mortality that higher SES subjects.

The authors of the present paper from New Zealand claim that when they used a “better” measure of SES (than the commonly used one that is based primarily on education, job level, and income), known as the Economic Living Standard Index-Short Forum (ELSI-SF), they found that reported alcohol intake correlated almost exactly with the results of the ELSI-SF on a self-reported assessment of physical health. They conclude that there are no health benefits of moderate drinking, only that people with higher SES are more likely to drink moderately.  Forum member Finkel remarked: “The Towers et al paper confirms yet again, and again, the J-shaped curve for drinking and health, and, lest we overlooked it, the same J-shaped relationship between drinking and ‘SES.’ The triangular interlocking influences of each one upon the other two are obvious.  It is willful for the authors to select just one directional vector of just one side of the triangle to declare, in the face of a mountain of well-developed evidence, both epidemiological and in the laboratory, including experimental animal research, over the last 30+ years, that now, at last, we –Towers et al. — alone have understood the truth.”  Finkel considered that this paper “does not add to our knowledge or perspective on this topic.”

Using single subjective measures of SES and health to make broad conclusions: The key differences in the effect of alcohol consumption on a self-reported index of health presented in this paper were related to the addition to the analysis of results of a self-reported measure of SES, the ELSI-SF.  Reviewer Djoussé commented: “This tool (the ELSI-SF) uses a series of questions, many of which are subjective and/or rely on the participant’s ability to recall.  It is obvious that decline in cognition is highly prevalent with aging.  Hence, ELSI-SF (and indirectly, the assessment of SES by this instrument) is subject to inaccuracy or imprecision that could bias the findings.  Moreover, physical health was assessed using SF-12, which suffers the same shortcomings as SES as detailed above.  It appears that both outcome (physical health) and the key confounder (SES) are not measured reliably in this study, thereby opening the door to biases that could limit the utility of the findings.”

Forum member Zhang stated: “We know that the SF-12 instrument has been well validated, but we know of little data on the ELSI-SF.  Nevertheless, the association of alcohol and SES is an interesting topic. If data show that participants with high SES in general drink alcohol and are more likely than others to drink it moderately, we must state that SES can be an important confounder of alcohol’s effects.  Having said that, I do have some concerns about the paper.  First, this is a cross-sectional analysis: it is hard to make the conclusion that the association between alcohol intake and Sf-12 is entirely accounted for by SES.  Since this study was conducted among an elderly population, survival bias (i.e., some people who may have died at an earlier age but are still alive owing to their alcohol consumption) could be a factor.  In addition I would like to see the relation of alcohol consumption to SF-12 in the different SES categories.  But only providing the results generated from regression analyses do not provide insights on this important issue.”

The net health effects of alcohol intake, including mortality, and not just self-reported “physical health”: Forum members considered it unusual that the authors of the present paper did not comment on the effects of alcohol consumption on mortality.  Surely, mortality (and other hard outcomes such as cardiovascular disease, dementia, diabetes, etc.) should be an important consideration when judging the “health benefits” of moderate drinking.  As noted by Djoussé, “The authors never examined any hard endpoint (i.e., mortality, CVD, cancer, motor vehicle accidents, or even biological markers) in this report; yet their title and conclusion are referring to ‘health benefits of moderate drinking.’  It is very difficult to follow such reasoning.  Self-assessed physical health is not synonymous with hard endpoints that can be validly measured.  To me, the authors are over-interpreting their data by referring to an endpoint that was not objective and which they could not measure directly.”

Importance of pattern of drinking, and not just the average alcohol intake: It appears that the authors took only the average intake of subjects as a measure of their alcohol consumption, and did not include data in their analyses on binge drinking or other information on the pattern of drinking.  It is well known that regular moderate intake of a given amount of alcohol has mainly beneficial effects on health, but such protection is lost when alcohol is consumed in binges, even of the same total amount of alcohol (Mukamal et al).  Further, almost all epidemiologic studies have shown that lower SES subjects tend to be more likely to binge drink, while higher SES subjects are more likely to be regular moderate drinkers, and consumers of wine rather than spirits as well. This was not discussed in the paper.

Reviewer Stockley summarized some of the key concerns noted by Djoussé and other Forum members: “I fully support the comments of others on the lack of hard end points, or even biological markers, and that the authors used only a self-reported general physical health measure as the key outcome. Among data missing from the discussion are a lack of a standard definition of a drink, no differentiation between type of alcoholic beverage consumed, the pattern of consumption, and when alcohol was consumed, such as with/without a meal.  Further, there was no discussion regarding differences in physical activity, demographics, employment status and type, and age (range 52 to 68 years, which includes individuals working full or part time, to potentially fully retired).  All of these factors could affect the health outcomes related to alcohol consumption.”

Forum member Mattivi was concerned that the authors’ assertions that their measure of SES, alone, gives the complete picture of factors affecting health actually confuses, rather than clarifies, our understanding of the relation of alcohol to health. He closed: “”Compito della scienza non è complicare le cose semplici, semmai semplificare quelle complesse” (Loosely translated: “The task of science is not to complicate the simple things; if anything, to simplify those that are complex.”)

Comparisons between the present study and the Dubbo Study of the Elderly:   Forum member Stockley commented on the decision of the authors not to discuss their results in comparison with those of an Australian study of very similar design that has been evaluating the effects of alcohol in an ageing population for many decades.  “Instead of using cross-sectional data from a single evaluation (as done in the present paper), The Dubbo Study has been repeatedly examining an ageing population of approximately 2800 men and women over age 60 who were first examined in 1988-89.  Hospitalisation and death records have been monitored continuously, and postal surveys were conducted every two years to confirm vital status. The survey in 1997 successfully traced more than 98% of surviving participants.

“As early as 1996, Simons et al judged the effects among the participants in the Dubbo Study of many socio-economic and lifestyle factors, including alcohol consumption, smoking, being married, poor-fair self-rated health, and physical activity, for their effects on mortality. They concluded: ‘By the end of almost 10 years follow-up, men consuming any alcohol lived 7.6 months longer and women lived 2.7 months longer than their abstaining counterparts.’   A follow-up paper by Simons et al in 2000 further elaborated on the causes of death in more than 800 subjects, concluding, ‘Moderate alcohol intake in the elderly appears to be associated with significantly longer survival in men 60-74 years and in all elderly women.’  In another report from the Dubbo Study in 2014, Simons stated: ‘All-cause mortality was related to quantity of alcohol intake in the familiar ‘U-shaped’ relationship, being 20% and 28% reduced in the low and moderate intake categories respectively, compared with nil intake.  Any alcohol intake added 12 months survival time in men and women over the follow-up period.  Alcohol intake in the low to moderate range appeared to offer protection against the onset of dementia.’”  Forum members considered that the similarity of the design of the two studies suggest that this might have been discussed in the present paper.

If SES is important to health, what are the mechanisms?  The results of the present paper add to results of previous research emphasizing the importance of SES in determining health.  However, the key question is “how?”  If the estimation of alcohol intake is based only on reported average intake, do lower and higher SES strata people consume alcohol differently (more or less binge drinking, spirits versus wine, with meals or not, etc.), as other data suggest?  Do other lifestyle habits (smoking, use of illegal drugs, chronic infections, etc.) differ according to SES?  Do subjects in differing strata of SES have different access to health care, or do they respond differently to information provided to the public on what constitutes a “healthy lifestyle”?  It is incumbent on scientists to evaluate the reasons for such differences by SES in their efforts to improve the health of the population.

References from Forum review

Mukamal KJ, Chen CM, Rao SR, Breslow RA. Alcohol Consumption and Cardiovascular Mortality Among U.S. Adults, 1987 to 2002.  J Am Coll Cardiol 2010;55:1328–1335.

Simons LA. Alcohol intake and survival in Australian seniors: The Dubbo Study and cardiovascular assessments, including examination of fasting blood samples.  Nutrition and Aging 2014;2:85-90.

Simons LA; McCallum J; Friedlander Y; Ortiz M; Simons J. Moderate alcohol intake is associated with survival in the elderly: The Dubbo study. Med J Aust 2000;173:121–124.

Simons LA, McCallum J, Friedlander Y, Simons J. 1996. Predictors of mortality in the prospective Dubbo study of Australian elderly. Aust N Z J Med 1996;26:40-48.

Forum Summary

It has long been recognized from epidemiologic studies that both the positive and negative health effects of alcohol consumption are modified by the socio-economic status (SES) of individuals. Higher SES subjects (higher education, income, job status, etc.) are more likely to be regular moderate drinkers, while lower SES subjects are more likely to binge drink and under-report their alcohol intake.  Truly moderate drinkers tend to have better health outcomes and fewer adverse effects from alcohol, and consistently show lower total mortality risks than non-drinkers or heavy drinkers.

The present study, from the New Zealand Longitudinal Study of Ageing, used data from the second examination of the cohort to do a cross-sectional analysis judging the relation of alcohol consumption to “health”. When controlling for age, income, and education (the latter two as measures of SES), they found a significant “J-shaped” association between reported alcohol intake and self-reported “physical health”.  However, the authors state that they then used another measure of SES, the Living Standard Index-Short Forum (ELSI-SF), and found that the reported alcohol intake of subjects correlated with their measure of physical health almost exactly the same as with the results of the ELSI-SF.  They conclude that there are no health benefits of moderate drinking, only that people with higher SES are more likely to drink moderately.

Forum members considered it unfortunate that the assessment of the exposure (alcohol intake) used in this study was based only on self-reported average intake and did not include data on the pattern of drinking (binge versus regular moderate intake, etc.).  Further, their outcome was based only on a self-reported questionnaire of “physical health,” and did include any hard data; an overall index based on assessments of functional status, biomarkers of and the occurrence of disease, and mortality would be preferable when judging “health benefits” of alcohol.  The Forum considered that the authors used inadequate indices of both the exposure to alcohol and their assessment of “health” (based on a questionnaire) to reach a conclusion that moderate drinking does not have any beneficial effects on health.  This was done ignoring the massive amount of not only observational data but results from extensive experimental studies over more than four decades.  Such research has found that moderate drinking not only has beneficial effects on assessments of physical health and disease occurrence, but results in significantly lower total mortality when moderate drinkers are compared with abstainers.

Regardless, given that this study, as have almost all epidemiologic studies, has shown the importance of SES on health outcomes, a key challenge to scientists is to seek to determine the mechanisms by which these differences occur.  One potentially important factor is that the differing health outcomes relate to errors in judging the exposure, i.e., inadequate assessments of alcohol: not evaluating for binge drinking versus regular intake, consuming alcohol with or without food, type of beverage consumed, under-reporting of intake, etc.  Also, much broader definitions of “health” are needed (rather than the results of a single self-administered questionnaire) that include the effects on functional status, disease states, and mortality.  And, it is important that other lifestyle factors (smoking, drug use, etc.) that affect health be properly evaluated as determinants of health.  Only with such data will we be able to fully judge the overall effects of alcohol consumption on health outcomes.

*                        *                      *                      *                      *                      *

Comments on this critique by the International Scientific Forum on Alcohol Research have been provided by the following members:

Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy

Luc Djoussé, MD, DSc, Dept. of Medicine, Division of Aging, Brigham & Women’s Hospital and Harvard Medical School, Boston, MA, USA

R. Curtis Ellison, MD, Professor of Medicine & Public Health, Boston University School of Medicine, Boston, MA, USA

Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA

Fulvio Mattivi, MSc, Head of the Department of Food Quality and Nutrition, Research and Innovation Centre, Fondazione Edmund Mach, in San Michele all’Adige, Italy

Linda McEvoy, PhD, Department of Radiology, University of California at San Diego (UCSD), La Jolla, CA, USA

Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia

Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway

Fulvio Ursini, MD, Dept. of Biological Chemistry, University of Padova, Padova, Italy

David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa

Yuqing Zhang, MD, DSc, Clinical Epidemiology, Boston University School of Medicine, Boston, MA, USA


Following the publication of Critique 195 on the Forum’s website earlier today, some additional relevant comments have been received. Forum member Erik Skovenborg has written: “Besides the general critique of ‘The “Health Benefits” of Moderate Drinking in Older Adults may be Better Explained by Socioeconomic Status’, I have some comments on the use of references used to support statements in the paper:

“Towers et al state: ‘Many older adults actively self-medicate with alcohol based on the widespread assumption that it is beneficial (Aira, Hartikainen, & Sulkava, 2008)’. In case you wonder why Towers et al went all the way round the globe for this reference, the answer is given by the Finish authors: ‘To our knowledge, there is no research on alcohol use as self-medication by the elderly to prevent or cure diseases in the community.’  The authors also explain why it is common among the aged in Finland to drink alcohol for medicinal purposes: ‘some find it an unstigmatizing way to mention their alcohol consumption’.  Anyway the resultant amount of alcohol drunk in the context of medicinal use is hardly worrying: ‘68% drink half a unit or less and 30% drink one unit.  The frequency was less than once a month in 38%, once a month in 28%, weekly in 27% and daily in 13%’.  The authors explain why their study cannot be generalized to other drinking cultures: ‘In Finnish drinking culture daily drinking is rare and most alcohol is consumed in binge drinking sessions’. Towers et al may not have read their reference, which would be negligence, but if they have read and understood the special situation concerning alcohol use in Finland, it amounts to intellectual dishonesty.

“Another statement from the Towers et al paper: ‘SES is a better predictor of older adults’ health and drinking level than education and income (Stephens, Alpass, Towers, Noone, & Stevenson, 2011; Towers et al., 2011).’ The paper of Stephens et al is published in a NZ journal that is not indexed in PubMed and not to be found in the Danish State and University Library. Towers et al found that ‘The odds of hazardous drinking were consistently higher for high earners and those in good living standards’. How does that agree with the argument that ‘higher SES levels may simply facilitate lifestyles that enable both regular moderate drinking and the capacity for better health’?  Towers et al also found that ‘binge drinkers were more likely to be Maori’, which introduces a possible bias in the study because Maori as a group tend to have less favorable standards of living.

“Towers et al state further: ‘We utilized the Economic Living Standard Index-Short Form (ELSI-SF; Jensen, Spittal, & Krishnan, 2005)’. The ELSI-SF has not been validated by international studies of other populations. The variables used by ELSI-SF are mentioned in another reference (Grundy & Holt, 2001) with this comment: ‘None of the variables we have considered are ideal on their own and most are subject to “reverse causation” problems.’”

In response to the above message, Forum member Dag Thelle commented: “Thanks to Skovenborg for bringing us up to par regarding the references used (or abused) in this paper. The research issue of interest is why higher socioeconomic echelons fare better regarding both somatic psychic health, not whether SES can ‘explain away’ the potential effects of a moderate alcohol intake.” Forum member Finkel also responded: “There are worlds of possibilities involved in ‘SES’ including, especially, examples of reverse causation that, if operative, would give the authors fits. And it’s so obvious that using a population skewed towards a distinct group is asking for trouble.”

Additional Forum members contributing:

Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark

Dag S. Thelle, MD, PhD, Department of Biostatistics, Institute of Basic Medical Sciences, University of Oslo, Norway; Section for Epidemiology and Social Medicine, Sahlgrenska Academy, University of Gothenburg, Sweden