Critique 267 – Alcohol consumption and risks of more than 200 diseases in Chinese men
Im PK; Wright N; Yang L; Chan KH; Chen Y; Guo Y; Du H; Yang X; Avery D; Wang S; Yu C; Lv J; Clarke R; Chen J; Collins R; Walters RG; Peto R; Li L; Chen Z; Millwood IY & China Kadoorie Biobank Collaborative Group
Nature Medicine, 29:1476–1486; 2023 https://doi.org/10.1038/s41591-023-02383-8
Introduction and Aim Alcohol consumption accounts for ~3 million annual deaths worldwide, but uncertainty persists about its relationships with many diseases.
Method We investigated the associations of alcohol consumption with 207 diseases in the 12-year China Kadoorie Biobank of >512,000 adults (41% men), including 168,050 genotyped for ALDH2-rs671 and ADH1B-rs1229984, with >1.1 million ICD-10 coded hospitalized events. At baseline, 33% of men drank alcohol regularly.
Results Among men, alcohol intake was positively associated with 61 diseases, including 33 not defined by the World Health Organization as alcohol-related, such as cataract (n = 2,028; hazard ratio 1.21; 95% confidence interval 1.09–1.33, per 280 g per week) and gout (n = 402; 1.57, 1.33–1.86). Genotype-predicted mean alcohol intake was positively associated with established (n = 28,564; 1.14, 1.09–1.20) and new alcohol-associated (n = 16,138; 1.06, 1.01–1.12) diseases, and with specific diseases such as liver cirrhosis (n = 499; 2.30, 1.58–3.35), stroke (n = 12,176; 1.38, 1.27–1.49) and gout (n = 338; 2.33, 1.49–3.62), but not ischemic heart disease (n = 8,408; 1.04, 0.94–1.14). Among women, 2% drank alcohol resulting in low power to assess associations of self-reported alcohol intake with disease risks, but genetic findings in women suggested the excess male risks were not due to pleiotropic genotypic effects.
Conclusions Among Chinese men, alcohol consumption increased multiple disease risks, highlighting the need to strengthen preventive measures to reduce alcohol intake.
Background including previous results
The relation between alcohol consumption and health is a complex one. Consuming alcohol in moderation has been associated with decreased overall mortality mainly caused by a reduced risk associated with cardiovascular diseases and other disorders like diabetes type II, dementia and certain cancers. Alcohol abuse and consuming alcohol at levels above moderate drinking guidelines is associated with increased risk for specific diseases.
The study recently published in Nature Medicine (Im et al., 2023) concerns the evaluation of a cohort of Chinese men and women and their risk for more than 200 diseases associated with their alcohol consumption. Contrary to many Western countries, China faces an increase in alcohol consumption. Chinese men reporting to consume alcohol in the last 12 months increased from 59% to 85% and yearly per capita alcohol consumption increased from 7.1 to 11.2 litres of pure alcohol based on WHO data (Manthey et al., 2019).
The motivation for this study is that the associations between alcohol consumption and disease may differ between Western and Chinese populations, because of drinking types of alcohol, patterns of consumption, patterns of diseases and differences in the ability to metabolize alcohol may differ between populations.
The cohort consisted of 512,724 Chinese, 41% of these, viz 210,217, were men with an average age of 52 + 11 years. During the 12.1 years of follow-up 134,641 men experienced at least one reported hospitalization event or death. In total 333,541 events were recorded in these 134,641 men over a 12.1-year period. This means that these Chinese men had on average about 2,5 disease events recorded over a 12-year time period. For comparison: in the US with 336 million inhabitants only 37 million hospitalizations occur, so 1 hospitalization per 10 Americans per year. It seems that this Chinese cohort consisted of relatively unhealthy people. This is also suggested by Table 1, there it is shown that 20 – 30% of these Chinese men and women self-reported previous chronic disease. This is essentially different than most epidemiological studies into the association between alcohol consumption and disease outcome. Usually, cohorts start with apparently healthy persons. The occurrence of new diseases is monitored and after a follow-up period association with specific diseases is assessed. The authors do not report whether the data have been corrected for these previous chronic diseases nor for poor self-reported health.
Also, confounding factors need to be corrected for in epidemiological studies; other life style factors may modify the alcohol consumption – disease relation. In this study the legend to Figure 2 indicates that the Cox models used were stratified by age and study area and adjusted for education and smoking. This is an unfortunate shortcoming of this study, since other potentially confounding factors of which data have been provided in Table 1, have not been corrected for. These factors include intake of fresh fruit, physical activity, blood pressure and BMI. The observations reported may therefore be confounded by various other important life-style factors.
Only 33% of the men in this cohort (44,027) were regular drinker. These drinkers reported an average of 286 g alcohol per week. This average is quite high, namely about 40 g of alcohol on average per day. The authors also indicate that underreporting may even have occurred in these heavy drinkers. The drinking categories used for men are non-drinker, ex-drinker, occasional drinker, less than 140 g/week, 140-280 g/week, 280-419 g/ week and more than 420 g/week. The latter very high drinking category still represents about 20% (6/33) of the drinking population, the higher than moderately drinking men represent another 45% of the investigated group. So, in total 65% of this Chinese cohort represents men that drink more than Western drinking guidelines. The authors also report that they may have missed extreme problematic drinkers, a common problem in population-based cohort studies.
Interesting to see that again, as in most high-quality epidemiological studies, a J-shaped curve was observed between alcohol consumption and overall morbidity (Figure 2). Those drinking 100-300 g per week had a lower Hazard Ratio for the cohort’s WHO alcohol-related diseases, the cohort’s new alcohol-related diseases, all alcohol-related diseases and all morbidity as compared to non-drinkers (20% of the population). Unfortunately, all analyses have been performed comparing all current drinkers, mainly consisting of alcohol abusers (33% of the population), with occasional drinkers (38% of the population) who have a similar Hazard Ratio as those drinking up to 100 g of alcohol per week. When studying the effect of a life-style factor like alcohol consumption it is essential that the comparison is made to the control group consisting of non-drinkers, preferably those that have been abstaining all their life. Using other control groups, as has been done in this study, implies that the effects of moderate alcohol consumption are not evaluated.
Unfortunately, as the authors indicate, drinking pattern was not studied nor analysed in this paper. The huge consumption levels however, suggest that Chinese drinking may not fit into the classically perceived healthy way of alcohol consumption, which is; in moderation, viz. 10-30 g on average per day (for men), preferably with a meal at several days of the week rather than the daily average all at once at one or two weekend nights.
Analyses performed in men were not performed in women. Reason indicated was that women had an extremely low alcohol consumption. This is unfortunate since the larger group of women mainly consisted of non-drinkers (64%), hardly any ex-drinkers (1%) and occasional drinkers (33%). Also, a large number of events (476,986) occurred in these women. It may have been a good population to study the effects of moderate alcohol consumption on disease outcomes in Chinese women.
Comments on Mendelian Randomization
Mendelian randomization analysis (MR) is an emerging research method that uses genetic proxies to test if certain behaviours are linked to health outcomes There has been a recent reliance on MR analysis studies in determining the risk of major causes of death and disability from alcohol consumption. MR analysis was subsequently used in this study to assess the causal relationship between alcohol consumption and disease outcome. The authors believe that this type of analyses better assesses causal relationships between alcohol consumption and disease outcome. Unfortunately, the authors do not discuss this methodology, which has been criticized extensively, specifically for its application in the alcohol-disease association.
MR has been criticized for its application in epidemiological alcohol research times (Mukamal et al., 2020, Mukamal and Beulens, 2022). MR analysis depends on assuming that the genetic variants: (i) are associated with the exposure (the relevance assumption); (ii) have no common cause with the outcome (the independence assumption); and (iii) have effects on the outcome that are solely mediated by the exposure (the exclusion restriction assumption). One of the basic problems with this technique therefore is that the variability in genotypes are not good indicators for alcohol consumption behaviour (assumption i) (Wehby et al., 2008), simply because genotype is not the only factor determining alcohol consumption. Also, ADH and ALDH alleles should not have an association with the outcomes independent from alcohol consumption (assumption ii). However, there seems to be an association between ADH and ALDH alleles and some disease outcomes independent from alcohol consumption (Zhang et al., 2023). Furthermore, the outcome (200 diseases in this case) should only be determined by exposure to alcohol (assumption iii), which also seems highly unlikely. Yet another problem with MR is that only linear associations are to be detected, whereas more complex associations as a J-shaped association is simply not possible to be detected by this technique.
Interestingly, the authors state in their abstract: genetic findings in women suggested the excess male risks were not due to pleiotropic genotypic effects. Does this mean that pleiotropic genetic effects are less relevant for the effects observed and that other factors such as their high alcohol consumption are the most important factors involved?
Furthermore, Frost and Wald (2021) had critiqued a previous study by these authors that also used an MR analysis of the same genetic polymorphism (genes ADH1B and ADLH2) that affects tolerance to alcohol (Millswood et al. 2019). While conventional analyses supported a J-shaped relationship between alcohol consumption and vascular disease, MR did not so the authors concluded that the relationship was actually monotonic, despite 30 years of biological and epidemiological data to the contrary. In this study, MR analysis again did not support a J-shaped relationship between alcohol consumption and vascular disease, although conventional analyses did (Im et al. 2023).
Indeed, there is a fundamental weakness in MR analyses used in these two studies which tends to conceal a true underlying relationship. It arises because when a true relationship between an outcome and an exposure is non-monotonic (such as J—shaped), then the formation of groups around the inflection point may be too coarse to reveal the non-monotonic relationship (Frost and Wald 2021).
Hence, this paper basically shows that alcohol abuse leads into a broad array of diseases in a primarily unhealthy elderly population of Chinese men who consume large amounts of alcohol. Accordingly, the paper should have been entitled “Alcohol abuse and risks of more than 200 diseases in apparently unhealthy Chinese men”. This observation adds little new insight in the already existing knowledge on alcohol consumption and health.
Specific Comments from Forum Members
Forum member Ellison suggest that “the points emphasized are on target, and raise questions about the applicability of the results of this paper to other cultures. I agree that the men in this study appear to be relatively heavy drinkers as they seem to have higher reported amounts of intake that seen in most cohort studies in North America and Europe. Moreover, it is stated that the alcohol consumed was mainly spirits, and many of the men drank spirits every day and apparently in rather large amounts. This prevents the results of this study from being useful for setting guidelines for moderate drinkers.
Thus, in my opinion, there are a number of problems with the study: the first is the lack of reliable information on the drinking pattern and the type of alcohol consumed. We have very reliable data from many previous studies showing that the pattern of drinking is a strong factor, in that more frequent but moderate drinkers tend to have healthier outcomes. Consuming alcohol with or without food makes perhaps an even larger difference in health effects; and it is clear that the best health effects seem to be associated with the regular but moderate consumption of wine, and there were few wine drinkers in this study. Unfortunately, for this group of subjects, it is difficult to identify those who were consuming alcohol in a “healthy” fashion. Further, I have little faith in using MR methodology, especially when they use such data to attempt to support their findings when using self-reported amounts of alcohol.
If these men are mainly relatively heavy spirits drinkers, the findings of an increase (rather than a decrease) in heart disease may be true, but these are not the seen in essentially all well-done population -based studies in the west.
Since we have incomplete information on what beverage is being consumed, in moderation or in binges, or whether or not it is in association with food, it is difficult for me to know what the relevance of this study is for western populations or even how to compare its results with those of other studies.”
Forum Member Skovenborg suggests that “there are also several caveats associated to the studies of Chinese men. One issue is that spirits is the type of alcoholic drink preferred by Chinese men and previous research suggests that male drinkers in China average rather large amounts of alcohol, which they often consume in binges (Li et al., 2011). Further, the two alleles studied have been shown to affect more than just alcohol consumption, limiting their use as an instrumental variable (Chen et al. 2014)”.
Forum Member de Gaetano considers that “the MR approach applied to alcohol epidemiology is questionable for (at least) two main reasons.
Indeed, when the genetic regulation of a phenotype is strong, stable over time and marginally influenced by non-genetic (environmental) factors, the MR approach is more appropriate. But in the case of alcohol, MR investigates the association between a “genetic predisposition” to consume alcohol (at any dose) and the outcome. The polymorphisms that “regulate” its consumption actually have limited impact on the phenotype, which is on the contrary largely influenced by environmental/cultural factors.
Second, the crucial issue is whether drinking in moderation, say a drink a day, is better for the health than not drinking at all. From a MR perspective, this would require a targeted genetic analysis comparing light-drinkers vs abstainers, which has not been carried out so far, to the best of our knowledge. If we compare (any daily dose) drinkers vs non-drinkers, it may well happen that identified-by-polymorphisms drinkers (without any reliable distinction on consumption levels) are at higher outcome risk in comparison with identified by- polymorphisms non-drinkers”.
Forum Member Waterhouse suggests “that there are differential effects considering the type of beverage. In China, the predominant alcohol is by far is distilled Baiju (from fermented sorghum and other grains). They apparently asked their subjects about the type of beverage (see Assessment of alcohol consumption), but I saw no such data reported in Table 1, nor any discussion of wine (or any beverage type). My guess is that the population of wine drinkers was too small to call out, but beer might have been a factor. It seems they collected beverage type data simply to quantify alcohol consumption, but it seems it was an oversight to not distinguish the various types, especially since they do mention that wine might be cardioprotective due to the non-alcoholic components.”
Concluding comments from Forum Members
Unfortunately, the evidence-base around the accuracy of MR analysis studies, increasingly employed by epidemiologists to identify ‘causal’ relationships between exposures of interest and various endpoints in the absence of actual experimental data, has become increasingly less black and white. For example, genetic variants are used as variables to investigate the causal relationship between potentially modifiable risk factors and health outcomes, such as cardiovascular diseases and more recently cancers, both alcohol-related and non-alcohol-related.
The usefulness of MR analysis studies of alcohol consumption and these diseases is, therefore, limited. Although they can be employed as an additional or supplementary analytical methodology, their underlying assumptions are problematic if two dimensions are to be analysed simultaneously with one instrumental variable, as in the analyses on the impact of alcohol consumption on diseases such as ischaemic heart disease (Holmes et al. 2014, Frick and Rehm 2016). The validity of the results also depend that pleiotropic or other direct causal pathways do not explain the association with the outcome; pleiotropy occurs when one gene influences multiple, seemingly unrelated symptoms or traits. Further, even MR yields results with differing conclusions that are not necessarily reproducible (Han et al., 2013, Holmes et al., 2014).
While they are less likely to be affected by unmeasured confounding or reverse causation than conventional observational or self-reported studies, they depend on underlying assumptions, the plausibility of which must be evaluated and the relevance of the results interpreted in consideration of other sources of evidence including conventional observational studies. This evaluation and consideration does not necessarily happen, however, as exemplified in a MR meta-analysis undertaken by Holmes et al. (2014), and incorrect conclusions can subsequently be drawn. The conclusions of such MR analyses are only sound if their underlying assumptions are sound and integrated and combined with conventional observational studies, and hence can often be controversial.
An excerpt from Ellison et al. (2021)
“There is little question that genetic analyses will progress markedly in the future, probably very quickly within the next few years. And, improved genetic scores will be developed and found to make major contributions to our efforts to identify persons at increased risk of developing CHD, as well as improving therapies. Genetic factors and Mendelian randomization studies represent a topic that will need frequent, ongoing reassessment as research progresses to be able to judge their (surely important) role in our prevention, diagnosis, and treatment of cardiovascular diseases.
However, even though we now have identified many genetic factors that we can include in MR analyses, it is clear that results from a variety of types of studies must be considered when attempting to judge health effects of alcohol. This is especially the case because type of beverage, drinking patterns (e.g., regular moderate versus binge drinking, rate of consumption, with or without food), smoking and other lifestyle habits, diet, and many other environmental factors relate to the effects of alcohol consumption. Thus, the combination of data from observational studies, clinical trials, animal experiments, as well as MR analyses, will be needed to improve our knowledge on the relation of alcohol intake to health and disease; it remains a continuing challenge.”
Chen, C.H., Ferreira, J.C., Gross, E.R. and Mochly-Rosen, D. (2014), Targeting aldehyde dehydrogenase 2: new therapeutic opportunities. Physiological Reviews, 94(1), 1-34, https://doi:10.1152/physrev.00017.2013.
Costanzo, S., de Gaetano, G., Di Castelnuovo, A., Djoussé, L., Poli,A. and van Velden, D.P. (2019), Nutrition, moderate alcohol consumption and lower total mortality risk: Justified doubts or established facts? Metabolism & Cardiovascular Diseases, 29, 1003-1008.
Ellison, R.C., Grønbæk, M. and Skovenborg, E. (2021) Using Mendelian Randomization to evaluate the effects of alcohol consumption on the risk of coronary heart disease. Drugs & Alcohol Today, 21:84-95. https://doi.org/10.1108/DAT-09-2020-0061
Frick, U., Rehm, J. (2016) Can we establish causality with statistical analyses? The example of epidemiology. In: Wiedermann W., von Eye A., editors. Statistics and Causality: Methods for Applied Empirical Research. Hoboken, NJ: Wiley; pp. 407–432.
Frost, C., and Wald, N. (2021) A limitation of genetic epidemiological analysis when associations are genuinely J-shaped illustrated using a prospective study of alcohol consumption and vascular disease. International Journal of Epidemiology, 50(6), 1757–1760. https://doi.org/10.1093/ije/dyab162
Han, H., Wang, H., Yin, Z., Jiang, H., Fang, M. and Han, J. (2013). Association of genetic polymorphisms in ADH and ALDH2 with risk of coronary artery disease and myocardial infarction: a meta-analysis. Gene, 526, 134–141. doi:10.1016/j.gene.2013.05.002
Holmes MV, Dale CE, Zuccolo L, Silverwood RJ, Guo Y, Ye Z, et al… InterAct Consortium. (2014) Association between alcohol and cardiovascular disease: Mendelian randomisation analysis based on individual participant data. British Medical Journal, 10(49) :g4164. doi: 10.1136/bmj.g4164. PMID: 25011450; PMCID: PMC4091648.
Im, P. K., Wright, N., Yang, L., Chan, K. H., Chen, Y., Guo, Y., et al., … China Kadoorie Biobank Collaborative Group. (2023). Alcohol consumption and risks of more than 200 diseases in Chinese men. Nature Medicine, 29(6). https://doi.org/10.1038/S41591-023-02383-8.
Li, Y., Jiang, Y., Zhang, M., Yin, P., Wu, F. and Zhao, W. (2011). Drinking behaviour among men and women in China: the 2007 China chronic disease and risk factor surveillance, Addiction, 106(11), 1946-1956, https://doi:10.1111/j.1360-0443.2011.03514.x.
Manthey, J., Shield, K. D., Rylett, M., Hasan, O. S. M., Probst, C., and Rehm, J. (2019). Global alcohol exposure between 1990 and 2017 and forecasts until 2030: a modelling study. Lancet 393(10190), 2493–2502. https://doi.org/10.1016/S0140-6736(18)32744-2
Millwood, I.Y., Walters, R.G., Mei, X.W., Guo, Y., Yang, L., Bian, Z., et al; China Kadoorie Biobank Collaborative Group (2019). Conventional and genetic evidence on alcohol and vascular disease aetiology: a prospective study of 500 000 men and women in China. The Lancet. 393(10183):1831-1842.
Mukamal, K. J., and Beulens, J. W. J. (2022). Limited alcohol consumption and lower risk of diabetes: can we believe our own eyes? American Journal of Clinical Nutrition, 116(6), 1460–1461. https://doi.org/10.1093/AJCN/NQAC258
Mukamal, K. J., Stampfer, M. J., and Rimm, E. B. (2020). Genetic instrumental variable analysis: time to call mendelian randomization what it is. The example of alcohol and cardiovascular disease. European Journal of Epidemiology, 35(2), 93–97. https://doi.org/10.1007/S10654-019-00578-3
Wehby, G. L., Ohsfeldt, R. L., and Murray, J. C. (2008). Mendelian randomization. equals instrumental variable analysis with genetic instruments. Statistics in Medicine, 27(15), 2745–2749. https://doi.org/10.1002/sim.3255
Zhang, J., Guo, Y., Zhao, X., Pang, J., Pan, C., Wang, J., Wei, S., Yu, X., Zhang, C., Chen, Y., Yin, H., and Xu, F. (2023). The role of aldehyde dehydrogenase 2 in cardiovascular disease. Nature Reviews. Cardiology, 20(7). https://doi.org/10.1038/S41569-023-00839-5
Comments on this critique by the International Scientific Forum on Alcohol Research were provided by the following members:
Henk Hendriks, PhD, Netherlands
Creina Stockley, PhD, MBA, Independent consultant and Adjunct Senior Lecturer in the School of Agriculture, Food and Wine at the University of Adelaide, Australia
R Curtis Ellison, MD, Section of Preventive Medicine/Epidemiology, Boston University School of Medicine, Boston, MA, USA
Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark
Pierre-Louis Teissedre, PhD, Faculty of Oenology–ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France
Fulvio Ursini, MD, Dept. of Biological Chemistry, University of Padova, Padova, Italy
Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy
Andrew Waterhouse, PhD, Department of Viticulture and Enology, University of California, Davis
Arne Silvaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway