The idea that low-volume alcohol intake protects against all-cause mortality in general populations has been controversial,1 even though many studies appear to support this hypothesis. In fact, a closer look at these studies shows many inconsistent findings.
Our group has carefully conducted several systematic reviews of all relevant published studies and meta-analyses2,3 of their pooled results, to examine the relationship between alcohol use and all-cause mortality risk. These analyses indicated that the so-called protective effect (i.e., “health benefits”) might be due to some errors in many of the identified studies that could bias towards the false appearance of such benefits, and that the observed association between alcohol consumption is highly variable depending on study design and quality.
Light and moderate drinkers, for example, are systematically healthier than current abstainers on a range of health indicators unlikely to be related to alcohol use e.g. dental hygiene, exercise routines, diet, weight, income.4 In fact, even lifetime abstainers appear to be systematically biased towards poorer health so they make all drinkers appear healthier than they should by comparison.5
Furthermore, most studies create further bias in the abstainer reference group by improperly classifying “sick quitters” or former drinkers, many of whom cut down or stop for health reasons, as non-drinkers or never drinkers.2 This bias will get more pronounced in studies of older people, as does the selection bias inherent in studying those who have survived to relatively older ages while maintaining their consumption at ‘moderate’ levels. Most studies have non-representative samples leading to an over-representation of older, white males. Adjustment of study population samples to make them more representative has been shown to eliminate apparent protective associations.6
We previously conducted two studies to investigate the relationship between alcohol use and all-cause mortality using systematic review and meta-analysis methods.2,3 These studies did not discover significantly reduced mortality risk after adjusting for potential confounding effects of study-level factors, suggesting the protective effects of moderate alcohol consumption may be spurious.
One of our other studies also used a systematic review and meta-analysis to examine the association between alcohol consumption and risk of death from coronary heart disease (CHD).7 As predicted, we found that protection from CHD mortality risk for ‘moderate’ drinkers (up to three 8g UK units per day) was only observed for older cohorts and not for studies of younger adults followed up to an age that they might experience heart disease. This is consistent with the theory we have outlined8 that former drinker bias increases over the life course.
What do new studies show?
With the passage of time, many more studies have been published since we last conducted these previous meta-analyses and we decided to update them to investigate this hypothesis more closely.3
This time we identified 107 cohort studies published up to 2021 (20 additional studies) that included the outcome of all-cause mortality. In our Open Science Forum pre-registered protocol,9 we predicted that controlling for former drinker bias and other variables in these studies would, once more, dilute the commonly observed protective effect for moderate drinking. In addition, we also tested the hypothesis that the appearance of protective effects would be enhanced for older cohorts and much weaker for younger cohorts. With more studies we were also able to make separate estimates for males and females.
In this new study recently published in JAMA Network Open, we found no significant reductions in risk of all-cause mortality for all male and female drinkers who drank less than about 3 UK units of alcohol per day compared to lifetime nondrinkers after controlling for key study characteristics, such as drinker biases, median age and sex of study cohorts and other factors of the studies. Before we adjusted for these study characteristics, there was also clear evidence that the apparent protective effects were more pronounced for cohorts of people recruited over the age of 55 years, as predicted.
Much discussion of these findings in the media has focused on our findings regarding the level of alcohol use at which mortality risk increased significantly. At face value, our findings suggest that all-cause mortality risk only increased among male and female drinkers above three units per day, much higher than various low risk guidelines,10-12 with some evidence that risk increased more rapidly for women as they increase daily consumption.
We caution, however, that all-cause mortality studies are not the best way to base drinking guidelines for the public. The great majority of causes of death are not related to alcohol use and so these studies find associations that are due to confounding lifestyle factors in addition to the biases described above regarding definitions of who is an abstainer.
Following recent practice in the UK, Australia, France and Canada, we recommend guidelines development based on alcohol-caused mortality estimates derived from aggregated, weighted, cause-specific risk estimates from conditions that are clearly causally related to alcohol consumption. These suggest much lower levels of consumption are associated with increased mortality risk e.g. the new Canadian guidance is that mortality risk significantly increases from just two drinks per week.13
We conclude with the idea that there are many reasons why people might choose to drink or not drink, but that it is wishful thinking that moderate alcohol use actually confers real health benefits, a fact acknowledged by the World Heart Federation last year.14
Written by Dr Jinhui Zhao, Professor Tim Stockwell and Professor Tim Naimi.
All IAS Blogposts are published with the permission of the author. The views expressed are solely the author’s own and do not necessarily represent the views of the Institute of Alcohol Studies.