Estimates of the risks associated with the use of radiologic contrast agents frequently are based on a subjective review of some of the numerous articles that have been published on the subject. We have chosen instead to synthesize the existing evidence in an objective, quantitative way by statistically combining the data from individual studies through meta-analysis. Although meta-analysis is subject to whatever biases may exist in the underlying data, and may inappropriately pool data from studies with significant differences, the estimates draw proportionately from each original study and, in addition, gain considerable precision owing to the increased amounts of data considered. Using this technique, we analyzed the data from all available original reports appearing since 1980. Crude rates derived from all reports were used to estimate risks with high-osmolality media. Weighted rate differences based on comparative studies only were used in the estimation of the reductions in risk obtainable with low-osmolality media. The risk of death with high-osmolality media was 0.9 per 100,000 uses (95% confidence interval, 0.3-2.6 per 100,000). The difference in risk produced by using low-osmolality media was 0 (95% confidence interval, -1.1 to 1.1 per 100,000). The risk of severe reactions associated with high-osmolality media was 157 per 100,000 uses (95% confidence interval, 144-172 per 100,000). The reduction in risk that can be obtained by the use of low-osmolality media was estimated to be 126 per 100,000 (95% confidence interval, 110-142 per 100,000). This meta-analysis shows that the risk of death is very low with either type of contrast material and that there is no difference between them. Severe nonfatal reactions with high-osmolality media, although more frequent, are still rare; approximately 80% of them can be prevented by using low-osmolality media.