Twenty-five studies (n=1,213 participants, with 1,225 breast cancers) were included in the review.
The pooled estimate of sensitivity was 0.63 (95% CI 0.56 to 0.70, I2=63.5%). The pooled estimate of specificity was 0.91 (95% CI 0.89 to 0.92, I2=74.1%). The pooled estimate of diagnostic odds ratio was 17.05 (95% CI 10.59 to 27.19); there was significant between-study heterogeneity for diagnostic odds ratio.
Regression analyses suggested that the rate of complete pathologic remission was the only factor investigated to significantly effect the diagnostic performance of MRI (p=0.02).
Subgroup analyses indicated that the specificity of MRI was lower in studies with a complete pathologic remission rate of at least 20% than in studies with a complete pathologic remission rate less than 20% (p=0.0003).
A funnel plot suggested possible publication bias.