The association between alcohol consumption and osteoarthritis: a meta-analysis and meta-regression of observational studies

To K, Mak C, Zhang C, Zhou Y, Filbay S, Khan W · Rheumatology International

A worked example of an association dissolving under adjustment. Pooled across all studies, any alcohol consumption was associated with less osteoarthritis (OR 0.79, 95% CI 0.68-0.93), and weekly-or-more with the same (0.79, 0.65-0.97), holding for radiographic (0.83), hand (0.80) and knee OA (0.85). But in unadjusted data the OR was 0.70 (0.55-0.89), and in studies adjusted for any covariate it was 0.93 (0.78-1.10) — no association; by design, the cohort studies alone gave 0.90 (0.72-1.12) and the cross-sectional studies 0.75 (0.59-0.94), so the pooled signal came from cross-sectional data. The authors' own conclusion is that the analysis dispels the notion that alcohol protects against osteoarthritis.

Population

29 observational studies — 16 cross-sectional, 10 cohort, 3 case-control — including 25,192 people with osteoarthritis

Intervention

Any alcohol consumption, and weekly or more frequent consumption

Comparator

Non-drinkers

Limitations

Observational studies only, mostly cross-sectional, with the exposure self-reported and the reference group of non-drinkers contaminated by people who stopped drinking because they were unwell. Risk of bias assessed on the Newcastle-Ottawa Scale; publication bias examined by funnel and Galbraith plots, with Egger's test pointing to a bias toward studies reporting protective associations. Funded by Versus Arthritis; no conflicts.