Study suggests lifetime selection biases skew positive health claims in alcohol research

In a recent study published in the Journal of Studies on Alcohol and Drugs, a group of researchers identified and tested study characteristics that may bias estimates of all-cause mortality risk associated with low-volume alcohol consumption using theory and evidence.

Background 

An increasing number of studies have challenged the idea that low-level alcohol consumption protects against serious diseases such as ischemic stroke (A stroke caused by a blockage in an artery supplying blood to the brain), ischemic heart disease (IHD) (A condition where narrowed heart arteries reduce blood flow to the heart), and type 2 diabetes (A chronic condition affecting how the body processes blood sugar). While many observational studies suggest moderate drinkers live longer and are healthier than abstainers, these comparisons often suffer from selection bias.

Assumptions about alcohol’s health benefits significantly impact global disease burden estimates and national drinking guidelines. Recent analyses emphasize the critical role of study design, especially the criteria for defining reference groups, in evaluating alcohol’s health effects. High-quality studies are essential to assess the impact of low-volume alcohol use on mortality accurately. Further research is needed to determine the true health impact of low-volume alcohol consumption, free from selection biases and methodological flaws.

About the study 

The present study analyzed 107 cohort studies to explore the relationship between alcohol consumption and all-cause mortality risk, identified in a systematic review and meta-analysis. Further meta-analyses were conducted on subgroups categorized as “lower” or “higher” quality based on potential biases in reference groups classified as abstainers. Focus was placed on individuals with low-volume alcohol consumption, defined as one to two drinks per day (1.3 g to 24 g of ethanol).

Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, studies published up to July 31, 2021, were identified through Web of Science and PubMed searches and previous meta-analyses. Inclusion criteria were English language, original research, prospective cohort studies, and general populations. Exclusions included non-original studies and specific illness-based participant selection. Data extraction was independently coded by three reviewers, focusing on outcomes, alcohol consumption measures, study characteristics, and controlled covariates.

Potential biases included contamination of nondrinking reference groups with former or occasional drinkers, baseline illness, median cohort age, smoking control, alcohol measurement, and socioeconomic status (SES) control. Meta-analytic methods involved converting relative risk (RR) estimates to natural log formats, assessing between-study heterogeneity, and performing mixed-effects regression analyses.

Significant heterogeneity was managed through mixed regression analysis weighted by inverse variance. Statistical analyses used SAS Version 9.4, with results converted to exponential scales and significance tests assuming two-tailed p-values or a 95{e60f258f32f4d0090826105a8a8e4487cca35cebb3251bd7e4de0ff6f7e40497} confidence interval.

Study results 

Studies using younger cohorts, excluding participants with current or prior ill health, assessing alcohol use over less than 30 days, avoiding abstainer bias in the reference group, and not controlling for smoking status or SES reported significantly higher mean RR estimates of all-cause mortality for low-volume drinkers. In total, 99 studies provided results for low-volume drinkers, with some reporting for multiple cohorts.

Further exploratory analysis was conducted on studies stratified by smoking status due to significant differences observed. Among six studies reporting results for nonsmokers, the RR for low-volume drinkers was 1.16 (95{e60f258f32f4d0090826105a8a8e4487cca35cebb3251bd7e4de0ff6f7e40497} CI [0.91, 1.41]), compared to 0.93 (95{e60f258f32f4d0090826105a8a8e4487cca35cebb3251bd7e4de0ff6f7e40497} CI [0.71, 1.16]) for seven studies reporting results for smokers.

Outcomes from six studies assessed as having a lower risk of lifetime selection bias defined by the absence of abstainer bias in the reference group and younger cohorts followed by older showed an RR of 0.98 (95{e60f258f32f4d0090826105a8a8e4487cca35cebb3251bd7e4de0ff6f7e40497} CI [0.87, 1.11]), close to 1.0.

Adjusted RR estimates for low-volume drinkers showed significantly lower all-cause mortality risk in the entire pooled set of 107 studies and models meeting combinations of lower study quality criteria (older cohorts, misclassification of former and/or occasional drinkers, and poor alcohol use measures).

Models meeting only one quality criterion (sampling younger cohorts or avoiding abstainer bias) still indicated a lower risk of all-cause mortality (RR = 0.88). However, incorporating both quality criteria resulted in an RR of 0.98, which is not significantly different from 1.0. Adding a higher-quality alcohol measure increased the RR to above 1.0 (RR = 1.04).

Graphical comparisons of all-cause mortality risk estimates for low-volume alcohol consumption revealed significantly lower mortality risk in studies prone to selection bias, an effect not seen in higher-quality studies. Higher-quality studies meeting all three quality criteria demonstrated higher mortality risk at each drinking level compared to lower-quality studies, indicating the importance of rigorous study design in accurately assessing the health impact of low-volume alcohol consumption.

Conclusions

To summarize, the study analyzed 107 cohort studies on alcohol use and all-cause mortality, identifying significant variation in support of the health benefit hypothesis.

Characteristics of studies, such as younger cohorts, exclusion of ill participants, and avoiding abstainer bias, were associated with higher RR estimates for low-volume drinkers. Smoking and SES also influenced outcomes. High-quality studies showed no significant reduction in mortality risk, suggesting lifetime selection biases create false health benefits for moderate drinking.

The findings highlight the need for rigorous study designs to assess the impact of low-volume alcohol use on mortality accurately.