Seven studies with a total of 392,015 participants (range n=5,016 to 222,135) were included in the review.
Three studies (n=347,324) reported sufficient data to calculate sensitivity and specificity. Study quality was considered moderate. Pooled estimates of sensitivity and specificity were 86% (95% CI 84.2% to 87.6%) and 88.2% (95% CI 88.1% to 88.3%); heterogeneity was high for both parameters.
Five studies (n=51,162) were used to both assess incremental cancer detection and determine false-positive rates. Study quality was considered moderate.
Incremental cancer detection: The pooled estimate for incremental cancer detection rate in women screened with computer-aided detection over those screened with single-read mammography alone was 50 (95% CI 30 to 80) per 100,000 women screened. Approximately 4.1% (95% CI 2.7% to 6.3%) of women recalled for additional testing based on computer-aided detection findings were finally diagnosed with cancer. No significant heterogeneity was detected.
False-positive, incremental recall: 96% (95% CI 93.9% to 97.3%) of women recalled due to computer-aided detection findings did not have cancer, which made the additional recall rate for healthy women 1,190 (95% CI 1,090 to 1,290) per 100,000 screened. Significant heterogeneity was detected for this outcome.
False-positive, incremental biopsy: 65% (95% CI 52.3% to 76%) of women biopsied due to computer-aided detection findings did not have cancer, which made the additional biopsy rate for healthy women 80 (95% CI 60 to 110) per 100,000 screened. No significant heterogeneity was detected.