This meta-analysis summarizes the accuracy of magnetic resonance angiography (MRA) for diagnosing residuals in coiled cerebral aneurysms by using the threefold Roy classification (residuals: none, neck, or sac). Four databases were searched from 2000 to June 2013 for eligible studies that compared MRA to digital subtraction angiography (DSA) and reported 3 × 3 count data of threefold Roy classification, or a reduced scheme of 2 × 2 count data. Bivariate and trivariate Bayesian random-effects models were used for meta-analysis. Among 27 included studies (2,119 coiled aneurysms in 1,809 patients) the average prevalence of DSA-confirmed sac residuals was 18.2 % (range 0-43 %). The pooled sensitivity was 88.0 % (95 % CI 81.4-94.0) and specificity was 97.2 % (94.6-99.0 %) for assessing sac residuals by MRA. In the trivariate meta-analysis, a "sac residual" finding at MRA had a high positive likelihood ratio of 28.2 (14.0-79.0). A "neck residual" finding had a moderate negative likelihood ratio of 0.246 (0.111-0.426), and the MRA finding of "no residual" had a good negative likelihood ratio of 0.044 (0.013-0.096). Subgroup analyses identified no significant influence of covariates on diagnostic accuracy (P > 0.05). In conclusion, in coiled cerebral aneurysms MRA with application of the threefold Roy classification is well suited for detecting or excluding sac residuals that might require retreatment.