[Validity and reliability of mammographic interpretation by Mexican radiologists, using the BI-RADS system]

Mammographic interpretation using the BI-RADS system determines if a patient will continue in the regular screening program or needs additional studies for breast cancer diagnosis, hence the importance of its evaluation. Sensitivity, specificity, diagnostic accuracy of mammographic interpretation and inter- and intra-observer variation, were assessed in a random sample of 29 radiologists from the Mexican Ministry of Health.Eighty diagnostic and screening mammographic studies were selected from the National Cancer Institute in Mexico City. Under the same conditions, two interpretations using the BI-RADS system were performed by each of the participating radiologists.Sensitivity in the first and second measurement was 72 and 74% respectively, while the specificity was 80% for the first and 82% for second. Multivariate analysis showed a positive association between the last year number of interpreted mammograms, the number of courses taken and diagnostic accuracy (beta=0.00002, 95% CI 0.000004, 0.00003, p=0.02, beta=0.005, CI 95% -0.0004, 0.01, p=0.07, respectively). Inter-observer agreement was low in both measurements (kappa=0.24, kappa=0.25, respectively). However, after combining BI-RADS categories in 2, agreement increased to 0.46 for both measurements. The intra-observer agreement was good (kappa=0.67).The number of mammographic studies interpreted the previous year is a good measure of the experience of the radiologist and is reflected in the diagnostic accuracy of mammographic interpretation. Courses on mammographic interpretation also have a positive impact on performance.