Objective This study was to aggregate the prevalence and risks of epiretinal membranes (ERMs) and determine the possible causes of the varied estimates.

Design Systematic review and meta-analysis.

Data sources The search strategy was designed prospectively. We searched PubMed, Embase and Web of Science databases from inception to July 2016. Reference lists of the included literatures were reviewed as well.

Study selection Surveys published in English language from any population were included if they had a population-based design and reported the prevalence of ERM from retinal photography with or without optical coherence tomography. Eligibility and quality evaluation was conducted independently by two investigators.

Data extraction The literature search generated 2144 records, and 13 population-based studies comprising 49 697 subjects were finally included. The prevalence of ERM and the ORs of potential risk factors (age, sex, myopia, hypertension and so on) were extracted.

Results The pooled age-standardised prevalence estimates of earlier ERM (cellophane macular reflex (CMR)), advanced ERM (preretinal macular fibrosis (PMF)) and any ERM were 6.5% (95% CI 4.2% to 8.9%), 2.6% (95% CI 1.8% to 3.4%) and 9.1% (95% CI 6.0% to 12.2%), respectively. In the subgroup analysis, race and photography modality contributed to the variation in the prevalence estimates of PMF, while the WHO regions and image reading methods were associated with the varied prevalence of CMR and any ERM. Meta-analysis showed that only greater age and female significantly conferred a higher risk of ERMs.

Conclusions Our findings suggest that ERMs are relatively common among aged population. Race, image taking and reading methodology may play important roles in influencing the large variability of ERM prevalence estimates.