Critique 022: Alcohol consumption decreases with the development of disease – 21 November 2010

Liang W, Chikritzhs T.  Reduction in alcohol consumption and health status.  Addiction 2010; in press (doi:10.1111/j.1360-0443.2010.03164.x)

Authors’ Abstract

Aims This study investigated the association between alcohol consumption and health status using cross-sectional national survey data.

Measurements and design This study relied upon self-report data collected by the 2004 and 2007 Australian National Drug Strategy Household (NDSH) surveys.  Households were selected using a multi-stage, stratified-area, random sample design.  Both surveys used combinations of the drop-and-collect and computer-assisted telephone interview approaches.  Respondents were questioned about their current and past drinking, the presence of formal diagnosis for specific diseases (heart disease, type 2 diabetes, hypertension, cancer, anxiety, depression) and self-perceived general health status.  Associations between drinking status, the presence of diagnoses and self-perceptions of general health status among respondents aged 18+ and 45+ were assessed using multivariate logistic regression. 

Setting and participants Males and females aged 18 years or older and resident in Australia.  The sample sizes for the 2004 and 2007 NDSH surveys were 24,109 and 23,356, respectively.

Findings Respondents with a diagnosis of diabetes, hypertension and anxiety were more likely to have reduced or stopped alcohol consumption in the past 12 months.  The likelihood of having reduced or ceased alcohol consumption in the past 12 months increased as perceived general health status declined from excellent to poor.

Conclusions Experience of ill health is associated with subsequent reduction or cessation of alcohol consumption. This may at least partly underlie the observed ‘J-shape’ function relating alcohol consumption to premature mortality.

Forum Comments

Background:  Most prospective epidemiologic studies show that subjects who develop disease tend to decrease their alcohol intake, as does this study, and the results of this paper are neither new nor surprising.  It has been realized for almost two decades that when comparing subjects’ health outcomes according to their alcohol consumption, it is important to exclude “sick quitters,” people who decrease or stop alcohol consumption due to disease (whether the latter is associated with alcohol intake or not).  Such subjects tend to have poorer outcomes in terms of both morbidity and mortality.  This study supports the premise that people who get ill decrease their alcohol intake.

Epidemiologists are very aware of the problem with classifying ex-drinkers.  The majority of studies now seek to compare consumers at various levels of intake with either “lifetime abstainers” (which avoids the potential problem with sick quitters) or with regular light drinkers.  Mixing lifetime abstainers and ex-drinkers in the same category is no longer an acceptable procedure in prospective studies.

In a study of changes in alcohol intake and risk of coronary heart disease (CHD) and all-cause mortality in the MONICA Augsburg cohort, the investigators could stratify subjects into categories of continuing drinkers and continuing non-drinkers, and in addition could classify quitters and starters.1  When diminishing the amount of misclassification of exposure (alcohol) by assessing alcohol consumption in a classical prospective cohort study twice and obtaining the above named groups, the inverse association between the group of continuing light to moderate drinkers and CHD becomes stronger.  This should be expected, because misclassification of the exposure variable in this case dilutes the association between alcohol consumption and a protective effect on CHD.

Forum Comments on Present Study 

Problems using this dataset for the analyses:  This is an observational study that is based on subjects’ reports on whether or not they decreased their alcohol intake over the past year (rather than comparing reports of concurrent drinking at two different points in time).  One Forum member comments: “The better the study design, the execution of the study, and the quality of the data, the clearer the protective effect of alcohol on CHD.  The Australian Health Survey is not suitable to shed any new light on the alcohol-CHD and all-cause mortality relationships.  Sometimes the intention is to fulfill some ideology or to simply confuse.”  Another Forum member stated that “this paper might be considered in the category of ‘circular reasoning’ and some philosophers like Friedrich Nietzsche would be reminded of ‘Die ewige Wiederkehr’ (‘eternal recurrence’).”

Specific comments on the analyses:  The reasons that the subjects in this study reduced their intake is not known.  Importantly, details on subjects in each category are not given, so it is not possible to ascertain whether those reporting a reduction in intake went from heavy drinking to light drinking or none (which would be expected to have favorable health outcomes), from light drinking to none (which could have adverse health outcomes), or somewhere in between. 

Due to the frequency of binge drinking and other irregular drinking patterns among younger subjects, it is difficult to know how to interpret the results for subjects ages 18+ in these analyses.  For those 45 years of age or older, the paper shows a greater frequency of reducing alcohol intake for those who develop diabetes or hypertension; these are two conditions for which medical advice usually includes limiting or avoiding alcohol, so these changes may well have been related to such recommendations.  While heavier drinkers with hypertension would benefit from decreasing their intake, if newly diagnosed diabetics were advised by their doctors to reduce a moderate consumption of alcohol, it may well have been ill advised.  Diabetics are at an increased risk of CHD and moderate-drinking diabetics are at much lower risk of CHD than are non-drinkers.2,3  In fact, the “protective” effect of moderate drinking against the development of CHD is even greater among diabetics than among non-diabetics.4

The authors classified as “ex-drinkers” those subjects who stated that they had not consumed any alcohol within the past year but who reported that they “had ever had a full serve of alcohol.”  It is probable that this was a very heterogeneous category, and the proportion of these ex-drinkers who may have had a previous problem with heavy consumption cannot be ascertained.

When relating changes in alcohol intake to reported state of health, the risk of subject’s reporting only “poor” or “fair” health (versus “good” to “excellent” health) was higher for subjects consuming 5-6 or more drinks/day than it was for those reporting 1-2 or 3-4 drinks/day.  The authors state that “being a recent abstainer, a life-long abstainer or drinking five or more drinks per day was associated with worse perceived general health status over the last 12 months.”  Had they elaborated on their findings, they would have pointed out that subjects meeting such criteria have worse outcomes than do people who drink 1 to 4 drinks a day; the risk of adverse health was actually higher (Odds ratio 1.33, 95% CI 1.07 – 1.66) for lifetime abstainers than it was for moderate drinkers.   Hence, among the conclusions of this paper is not only that people who develop disease tend to decrease their intake but that lifetime abstainers report poorer health than do moderate drinkers.

The authors state that their previous data indicate that the reported intake of alcohol in Australia is considerably lower than the actual intake.  In their discussion of the question of underestimation, the authors do not mention one of the logical consequences of underestimation: that the limits for “healthy” consumption would be too narrow and would have to be adjusted upwards.  Applying this principle to the present analyses would suggest that the association between heavier alcohol intake and poor health may not become apparent until the usual consumption is quite high.  Hence, while the authors state that a proportion of subjects reporting low levels of intake “actually drink at risky levels,” their data suggest that for at least some level classified as “risky drinking,” the health status of subjects is better than it is among lifetime abstainers.

The suggestion by the authors that people who change their intake might be classified according to their previous intake (analogous to the “intention to treat” principle in clinical trials) is interesting.  Unfortunately, since numbers of subjects in different categories are not given in the paper, it is not possible to estimate the effect such “reclassification” might have had in the present analyses.

Among the limitations alluded to by the authors is the low response rates (<50%) in the surveys upon which these analyses are based.  Further, they state that “A large number of subjects with missing values in controlled variables were excluded in the multivariate logistic regression.”  Again, a lack of numbers for each analysis makes it impossible to judge the effects of this. 

Forum Summary:  The key finding from this analysis is that people who state that they have developed disease within the past 12 months (especially diabetes or hypertension) report a reduction in their intake of alcohol.  This is in line with most previous studies, and epidemiologists always attempt to not include “sick quitters” within the non-drinking category in their analyses.

Further, while heavier drinkers are more likely to report less than good health status, the results of this study show that lifetime abstainers also report poorer health than do moderate drinkers.  Again, this is a usual finding in prospective cohort studies.

A weakness of this study is that changes in alcohol intake are based purely on self-report at one point in time (rather than on actual reports of intake at two or more points in time).  Further, a lack of presentation of the numbers of subjects within various categories makes it difficult to judge the implications of this study.

Most prospective studies in which alcohol intake is assessed at different times (rather than having “changes” based only on recall at one point in time, as was done in this study) usually indicate that subjects who decrease their intake are more likely to subsequently develop adverse health outcomes, especially related to cardiovascular disease, than those who continue moderate drinking.

References from Forum Review:

1.   Wellmann J, Heidrich J, Berger K, Döring A, Heuschmann PU, Keil U.  Changes in alcohol intake and risk of CHD and all-cause mortality in the MONICA-Augsburg cohort 1987-1997.  EJCPR 2004;11:48-55.

2.  Valmadrid CT, Klein R, Moss SE, Klein BE, Cruickshanks KJ.  Alcohol intake and the risk of coronary heart disease mortality in persons with older-onset diabetes mellitus.  JAMA. 1999;21;282:239-246.

3.  Solomon CG, Hu FB, Stampfer MJ, Colditz GA, Speizer FE, Rimm EB, Willett WC, Manson JE.  Moderate alcohol consumption and risk of coronary heart disease among women with type 2 diabetes mellitus.  Circulation 2000;102:494-949.

4..  Ajani UA, Gaziano JM, Lotufo PA, Liu S, Hennekens CH, Buring JE, Manson JE.  Alcohol consumption and risk of coronary heart disease by diabetes status.  Circulation. 2000;102:500‑505.

                                                                                           *                              *                              *

Comments included in this critique by the International Scientific Forum on Alcohol Research were provided by the following:

Ross McCormick PhD, MSC, MBChB, Associate Dean, Faculty of Medical and Health Sciences, The University of Auckland, Auckland, New Zealand.

Creina Stockley, clinical pharmacology, Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia.

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

Pierre-Louis Teissedre, PhD, Faculty of Oenology – ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France.

Erik Skovenborg, MD, Scandinavian Medical Alcohol Board, Practitioner, Aarhus, Denmark.

R. Curtis Ellison, MD, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA.

Ulrich Keil, MD, PhD, Institute of Epidemiology and Social Medicine, University of Münster, Münster, Germany.

Dominique Lanzmann-Petithory,MD, PhD, Nutrition/Cardiology, Praticien Hospitalier Hôpital Emile Roux, Paris, France.

David Vauzour, PhD, Dept. of Food and Nutritional Sciences, The University of Reading, UK.