Meta-analytic Cross-Sectional Epidemiological Study

Raw proportions bounded near 0 or 1 have non-normal sampling distributions and heteroscedastic variances, which violate the assumptions of standard inverse-variance meta-analysis. The Freeman-Tukey double arcsine transformation stabilises the variance across the full range of proportions. Skipping it when many primary-study proportions are extreme (below 10% or above 90%) can produce biased pooled estimates and misleading confidence intervals. For proportions comfortably between 20% and 80%, logit transformation is a reasonable alternative.

For prevalence meta-analyses across different populations, settings, and time periods, a random-effects model is almost always appropriate because true prevalence is expected to vary across contexts. A fixed-effect model assumes all studies estimate exactly the same underlying prevalence, which is rarely plausible in epidemiology. The DerSimonian-Laird estimator is the most common random-effects approach, though restricted maximum likelihood (REML) is preferred when the number of studies is large.

High I-squared does not automatically invalidate the meta-analysis, but it does mean the pooled estimate should be interpreted cautiously. Conduct pre-specified subgroup analyses (by region, diagnostic criterion, age group, study quality) to explore sources of heterogeneity. Meta-regression can test whether study-level covariates explain between-study variance. If heterogeneity remains unexplained and is very large, presenting subgroup-specific estimates rather than a single pooled figure is more informative and honest.

No. Pooling cross-sectional data provides frequency estimates, not causal evidence. Cross-sectional studies measure exposure and outcome simultaneously, so they cannot establish which came first. For causal or associational questions, a meta-analysis of cohort, case-control, or randomised studies is required. The meta-analytic cross-sectional design answers 'How common is X?' rather than 'Does A cause B?'

The Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Prevalence Studies and the AHRQ quality assessment tool for observational studies are both widely used and specifically adapted for cross-sectional designs. The Newcastle-Ottawa Scale adapted for cross-sectional studies is another option. Whichever tool is chosen, quality scores should inform sensitivity analyses rather than be used as a simple exclusion threshold.