Multi-Reader Multi-Case Controlled Clinical Trial to Assess the Fujifilm FFDM and DBT Reader Training Program

This endpoint was evaluated qualitatively. Reported the number of readers meeting the Pass Criteria on the final FFDM plus DBT assessment case set, which requires adequate performance in cancer cases (detection rate) as well as non-cancer cases (recall rate). Per-subject BI-RADS, POM and recall scores were derived. Credit was only given for identifying a subject with cancer if the reader marked findings in at least one location with cancer. Findings that did not match the location of a malignant lesion were ignored for cancer cases in the per-subject analyses.

The magnitude and direction of differences between performance metrics for the two modalities, and variance components and correlations that would influence sample size and case mix for the pivotal reader study (FMSU2013-004G) comparing performance metrics between the two modalities, obtained using established methods for analysis of MRMC (multi-reader, multi-case) studies. The statistician estimated AUC's for each reader in each review condition based on per-subject POM scores requiring correct lesion localization. Statistician performed MRMC comparison of AUC's between reading conditions using the MRMC analysis of variance (ANOVA) method.