Eleven studies (n = 4,983) were included in the review. The ROC curves did not suggest the presence of a significant cut-off effect. Significant heterogeneity was present in most analyses. The diagnostic OR (a measure of overall accuracy) for MRI and mammography was 124.8 (95% CI: 36.4, 427.4) using BI-RADS 4 or 5 as positive and 45.9 (95% CI: 17.5, 120.9) using BI-RADS 3 or more as positive [the last statistic is taken from table, which conflicted slightly with the text figures in the review].
Using BI-RADS 4 or 5 as positive, the negative LR (decrease in odds of disease after a negative result) for MRI plus mammography was 0.14 (95% CI: 0.05, 0.42) compared with 0.7 (95% CI: 0.59, 0.82) for mammography alone. The probability of disease after a negative test result was 0.3 per cent (95% CI: 0.1%, 0.8%) for the combination compared with 1.4 per cent (95% CI: 1.2%, 1.6%) for mammography alone.
The positive LR (increase in odds of disease after a positive result) and probability of disease after a positive test result were 16.4 (95% CI: 11.1, 24.1) and 25 per cent (95% CI: 18.4%, 33%), respectively, for MRI plus mammography and 24.8 (95% CI: 11.6, 53) and 33.6 per cent (95% CI: 10.1%, 51.9%), respectively, for mammography alone. Results for other outcomes and BI-RADS cut-off values were reported.