Performance of screening MRI in high risk patients at initial versus subsequent screen

To compare the characteristics, outcomes, and performance metrics in women undergoing initial breast MRI screening versus subsequent screening.

A retrospective database search identified screening MRIs performed at an academic practice from 2013 to 2015. MRIs were divided into two groups: (1) initial screens and (2) subsequent screens (interpreted with at least one prior MRI for comparison). Benignity was confirmed with pathology or >1-year follow-up. Malignancy was confirmed by pathology. Performance metrics were calculated. Comparisons were made using Binomial and Fisher's exact tests.

We observed a higher rate of abnormal interpretations (52% vs. 19%; p < 0.001) and rate of biopsy (49% vs. 15%; p < 0.001) in the initial versus subsequent screen group. The positive predictive value of biopsy was slightly lower at initial versus subsequent screen (17% vs. 19%, p = 0.99). However, the cancer detection rate was higher at initial than at subsequent screen (85 vs. 29/1000, p = 0.08). Sensitivity was higher at initial (100%) versus subsequent (88%) screen. However, the specificity at initial screen was low (55%) compared to subsequent screen (83%).

The higher rate of abnormal interpretations in the initial versus subsequent screen group in part reflects a prevalence screening. Although we observed more abnormal interpretations in the initial screen, this was likely justified by the significantly higher cancer detection. This evidence may be used to counsel patients and referring providers on the expected higher likelihood of recall from an initial screening MRI balanced with higher detection of malignancies. Findings also highlight the importance of having comparison MRIs to decrease false positives.