The association between alcohol consumption and all-cause mortality: An umbrella review of systematic reviews using lifetime abstainers or low-volume drinkers as a reference group.
Level 2 - randomized trial
Umbrella review of systematic reviews of prospective cohort studies
PubMed 38465993 · doi:10.1111/add.16446
What was done
Authors conducted an umbrella systematic review of systematic reviews evaluating alcohol consumption and all-cause mortality in prospective cohort studies. They searched PubMed/Medline, Embase, PsycINFO, and Cochrane Library through March 2022. Inclusion was restricted to systematic reviews using lifetime abstainers or low-volume/occasional drinkers as the reference group (excluding former drinkers). Risk of bias was formally assessed for each included review.
What was found
From 2,149 retrieved articles, 25 systematic reviews were identified, but only 5 excluded former drinkers from their reference group. Four of the five included reviews had a high risk of bias. Three reviews reported a J-shaped relationship with significant decreased risk for low-volume drinking (RR range 0.84 to 0.95), whereas two did not. The single review at low risk of bias reported a monotonically increased mortality risk with greater consumption compared to occasional drinking: RR 1.02 for low-, 1.13 for medium-, 1.33 for high-, and 1.52 for higher-volume drinking. All five reviews reported significantly increased mortality risk at higher alcohol intake levels (RR range 1.28 to 3.70). Conversely, 17 of the 20 excluded reviews that failed to separate former drinkers reported decreased mortality risk for low-volume drinking.
Why it matters
Much of the apparent protective effect of light drinking (the J-shaped curve) appears to stem from reference-group bias (the 'sick-quitter effect'), where former drinkers with existing health problems are pooled into abstainer groups. When systematic reviews control for this bias and maintain low risk of bias, light drinking shows no clear survival advantage.
Limits
Only five systematic reviews met the reference group criteria, and four of them suffered from high risk of bias. Subgroup analyses for age and sex had substantial evidence gaps for multiple relevant factors. All findings ultimately depend on observational cohort data susceptible to misclassification of self-reported alcohol intake and residual confounding.