Smyth A, Teo KK, Rangarajan S, O’Donnell M, Zhang X, Rana P, Leong DP, et al. Alcohol consumption and cardiovascular disease, cancer, injury, admission to hospital, and mortality: a prospective cohort study. Lancet 2015. Pre-publication. http://dx.doi.org/10.1016/ S0140-6736(15)00235-4. Online/Comment http://dx.doi.org/10.1016/ S0140-6736(15)00236-6.
Background: Alcohol consumption is proposed to be the third most important modifiable risk factor for death and disability. However, alcohol consumption has been associated with both benefits and harms, and previous studies were mostly done in high-income countries. We investigated associations between alcohol consumption and outcomes in a prospective cohort of countries at different economic levels in five continents.
Methods: We included information from 12 countries participating in the Prospective Urban Rural Epidemiological (PURE) study, a prospective cohort study of individuals aged 35-70 years. We used Cox proportional hazards regression to study associations with mortality (n=2723), cardiovascular disease (n=2742), myocardial infarction (n=979), stroke (n=817), alcohol-related cancer (n=764), injury (n=824), admission to hospital (n=8786), and for a composite of these outcomes (n=11,963).
Findings: We included 114,970 adults, of whom 12,904 (11%) were from high-income countries (HICs), 24,408 (21%) were from upper-middle-income countries (UMICs), 48,845 (43%) were from lower-middle-income countries (LMICs), and 28,813 (25%) were from low-income countries (LICs). Median follow-up was 4.3 years (IQR 3.0-6.0). Current drinking was reported by 36,030 (31%) individuals, and was associated with reduced myocardial infarction (hazard ratio [HR] 0.76 [95% CI 0.63-0.93]), but increased alcohol-related cancers (HR 1.51 [1.22-1.89]) and injury (HR 1.29 [1.04-1.61]). High intake was associated with increased mortality (HR 1.31 [1.04-1.66]). Compared with never drinkers, we identified significantly reduced hazards for the composite outcome for current drinkers in HICs and UMICs (HR 0.84 [0.77-0.92]), but not in LMICs and LICs, for which we identified no reductions in this outcome (HR 1.07 [0.95-1.21]; pinteraction < 0.0001).
Interpretation: Current alcohol consumption had differing associations by clinical outcome, and differing associations by income region. However, we identified sufficient commonalities to support global health strategies and national initiatives to reduce harmful alcohol use.
Most prospective epidemiologic studies that have evaluated the association of alcohol consumption with health and disease have been carried out in more developed countries, with such information from the developing world quite limited. The present publication is from the Prospective Urban Rural Epidemiological (PURE) study that focuses especially on middle-income and lower-income countries. Included in this analysis are data from 12,904 subjects from high-income countries (HICs: Sweden and Canada); 24,408 from upper-middle-income countries (UMICs: Argentina, Brazil, Chile, Poland, South Africa and Turkey); 48,845 from lower-middle-income countries (LMICs: China and Columbia); and 28,813 (25%) from low-income countries (LICs: India and Zimbabwe).
Problems in combining data from diverse populations: It has long been recognized that, when evaluating the health effects of drinking, it is especially important to evaluate alcohol intake within a particular population, with its specific genetic, lifestyle, environmental, and cultural factors, and not study alcohol in isolation. Forum member Van Velden has emphasized that we must consider alcohol consumption “ . . . within the context of other important lifestyle factors such as weight management, exercise, no smoking, effective stress management, sense of purpose in life (spiritual health?), and an overall healthy diet. Alcohol must never be seen in isolation; this is often the problem in science: the reductionalistic view.”
Forum members considered it a misapplication of epidemiologic principles for the authors of this paper to attempt to answer overall questions about alcohol and health by combining data from such diverse populations. Even though the study includes information from many countries, the data do not permit the investigators to determine the net health effects of alcohol consumption that are applicable to people everywhere. Their analytic results cannot be used, as they suggest, to “support global health strategies,” although they may be useful for “national initiatives to reduce harmful alcohol use” in particular countries.
Reviewer Ellison stated: “Attempting to use the results presented in this paper to develop a ‘global’ guide to alcohol and health would, in my opinion, not be proper. What possible message can be derived by combining data from higher-income countries where only 6% of men had little or no education and 15% were current smokers with data from low-income countries where 41% of men had little or no education and 60% were current smokers? There are too many factors associated with health outcomes to attempt to determine the specific alcohol effects when considered in isolation.”
He continued: “The reporting of data from the less developed countries provides important information from areas of the world for which previous results are sparse, and can provide guidance for developing measures to control abusive alcohol use in these areas. However, the authors should have stopped with a description of such data, and not attempted to provide a ‘pure’ estimate of the effects of alcohol on a number of health conditions by compiling data from these markedly diverse populations into a single analysis.”
Forum member Waterhouse agreed: “It seems to me that the authors are trying to make universal health/nutrition recommendations for widely and wildly diverse populations. Perhaps some alternate examples could help show the futility of this exercise. Certainly the issue of very different causes of mortality in two populations, a very common situation, would of necessity lead to very different approaches to improving longevity. From a nutritional perspective, the issue of iron deficiency might be a useful example. If everyone took an iron supplement because one or two countries had deficiencies, there would actually be toxic effects in some populations.”
Forum member Thelle had additional comments. “The PURE study is well-planned and executed with standardized methods (as far as that is possible for multi-centre studies) both with regard to exposure variables and end-points. When assessing the results we should ask whether the study design and its execution might induce systematic errors, and to what extent the subsequent analyses have taken confounders into account. The prospective design should protect against reversed causation, even if I am unsure whether they actually did sensitivity analyses by excluding end-points during the first 1 or 2 years after the baseline examination. Otherwise I see no obvious faults in the design.
“The authors have tried to adjust for confounders as far as that was possible. Still, one cannot exclude further confounding especially as subjects from low income countries contain a larger proportion of smokers. The authors also point at differences between countries not fully explained by drinking pattern (or other factors); this suggests that further confounding exists. Lumping all types of stroke into one category may be hiding associations, especially as low income countries might have a higher frequency of cerebral hemorrhages. Similarly, using a single grouping for all cancers may be inadequate.
“We can look at the end-point categories used in this analysis as final effects of exposure and non-exposure to multiple causal factors, be it smoking, alcohol, treatment, SES, health policies, etc. The disease event is the result of a complex of factors, each needed to establish a sufficient cause leading to an end-point. The question we can raise is whether abstention from alcohol (or at least reduction) would result in reduced incidence rates in this heterogeneous cohort. The authors obviously believe so, but we might ask for more evidence, i.e. longer follow-up, larger studies and more specified hypotheses before we feel sure that there is no net benefit from a moderate alcohol intake in middle-aged and elderly individuals. In the young ones I can’t see much benefit.”
Stated reviewer Mattivi: “The data collected are interesting and may provide guidance for developing measures to control alcohol abuse, especially in under-developed, low income countries; this is an important aim which deserve unconditioned support. That said, the discussion is incautious: to compare apples to oranges in the quest of a theory which must fit all mankind seems an exercise of over-extrapolation.” Reviewer Stockley stated that “It would have been good for the authors to have teased out more the difference in observations between low and high income countries.”
Other limitations to the study: The authors recognized many limitations to their study. The accompanying Comment by Connor and Hall is directed at using the results of this study for determining general population approaches for alcohol control. These authors point out other problems, such as short follow-up time (4.3 years), the low number of cases of disease (for example, for studying myocardial infarction among moderate drinkers, there were only 59 cases.), and the classification in many analyses of all “drinkers” as one group, without indication of the amount consumed or the pattern of drinking. While the data could help in developing regional approaches for controlling abuse, they cannot necessarily be used to develop guidelines that would be applicable globally.
Consistency of these new data with previous research: Many of the results in this paper are consistent with much previously reported data. These include, for example, the differences noted by the authors related to the effects of previous alcohol use among current abstainers (ex-drinkers almost always have higher risk of disease than lifetime abstainers); pattern of drinking (better results from regular, moderate intake of alcohol with food versus from binge drinking); type of beverage (generally greater health outcomes are seen for consumers of wine, which may be due both to the beverage and the drinking patterns of subjects); and a significant lowering of risk of myocardial infarction, as seen in almost all epidemiologic studies.
The present publication is from the very large Prospective Urban Rural Epidemiological (PURE) study that focuses especially on middle-income and lower-income countries, for which previous data on the association of alcohol consumption with health outcomes are sparse. Included in this analysis are data from high-income countries (HICs: Sweden and Canada); upper-middle-income countries (UMICs: Argentina, Brazil, Chile, Poland, South Africa and Turkey); lower-middle-income countries (LMICs: China and Columbia); and low-income countries (LICs: India and Zimbabwe). The large majority of the subjects were from lower income countries.
It has long been recognized that, when evaluating the health effects of drinking, it is especially important to evaluate alcohol intake within a particular population, with its specific genetic, lifestyle, environmental, and cultural factors, and not study alcohol in isolation. While the PURE study is a very well-done study, Forum members were concerned that the authors have attempted to answer overall questions about alcohol and health by combining data from such diverse populations. For example, the issue of very different causes of mortality in different populations would lead to very different approaches to improving longevity. There are too many factors associated with health outcomes to attempt to determine the specific alcohol effects when considered in isolation.
The authors tried to adjust for confounders as far as that was possible, but one cannot exclude further confounding, especially as subjects from lower-income countries contained a much larger proportion of smokers than there were in higher-income countries. The authors also point at differences between countries not fully explained by drinking pattern (or other factors); this suggests that further confounding exists. Further, lumping all types of stroke into one category may be hiding associations, especially as lower-income countries might have a higher frequency of cerebral hemorrhages. Similarly, using a single grouping for all cancers may be inadequate. Other limitations of this study include the short duration of follow-up time (4.3 years) and the low number of cases of each outcome.
The Forum concluded that even though the study includes information from many countries, it is not possible from such data to determine net health effects of alcohol consumption that are applicable to people everywhere. Their analytic results cannot be used, as the authors suggest, to “support global health strategies,” although they may surely be useful for developing “national initiatives to reduce harmful alcohol use,” especially for lower-income countries.
* * * * * * * * * *
Contributions to this critique by the International Scientific Forum on Alcohol Research have been provided by the following members:
Andrew L. Waterhouse, PhD, Department of Viticulture and Enology, University of California, Davis, USA
David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa
Dag S. Thelle, MD, PhD, Senior Professor of Cardiovascular Epidemiology and Prevention, University of Gothenburg, Sweden; Senior Professor of Quantitative Medicine at the University of Oslo, Norway
Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway
Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia
Fulvio Mattivi, PhD, Head of the Department of Food Quality and Nutrition, Research and Innovation Centre, Fondazione Edmund Mach, in San Michele all’Adige, Italy
Dominique Lanzmann-Petithory,MD, PhD, Nutrition/Cardiology, Praticien Hospitalier Hôpital Emile Roux, Paris, France
Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA
R. Curtis Ellison, MD. Section of Preventive Medicine & Epidemiology, Department of Medicine, Boston University School of Medicine, Boston, MA, USA
Elizabeth Barrett-Connor, MD, Distinguished Professor, Division of Epidemiology, Department of Family Medicine and Public Health and Department of Medicine, University of California, San Diego, La Jolla, CA USA