Análisis de concordancia en la interpretación de imágenes mamográficas entre médicos radiólogos de Bogotá, D.C.

Introduction: Among all cancers, breast cancer is the primary cause of female mortality in Colombia. Therefore, it has been considered necessary to strengthen mammogram-based early detection programs by implementing quality assurance programs, including the evaluation of the quality of clinical images and analysis in the diagnostic concordance of the BIRADS categorization. Objective: To evaluate the inter-observer variability among radiologists in the interpretation of mammograms categorized by means of Breast Imaging Reporting and Data System (BI-RADS). Methods: This cross-sectional study includes a sample of 323 mammographic examinations in asymptomatic women between 50 and 69 years. The participants were randomly selected and stratified according to the BI-RADS categorization, in 11 mammographic centers in Bogotá, D. C., Colombia. By means of the kappa coefficient, the concordance among the consensus of two expert radiologists of the Colombian International Cancer Institute and radiologists of the mammography centers in the interpretation of mammograms by means of the BI-RADS categorization was determined. Results: After reaching consensus between two expert radiologists from the Colombian National Cancer Institute, the analysis of concordance with the radiologists of all institutions showed an agreement rate of 54,9% and a kappa coefficient (k) of 0,21 (95% CI: 0,14-0,28). When BI-RADS-groups I and II were clustered, both the agreement rate as well as the kappa coefficient increased to 75,3% and k=0,26 (95% CI: 0,17-0,34), respectively. Discussion: The results show a weak concordance of the BI-RADS categorization presented in mammographic reports, suggesting that there is a high inter-observer variability at the time of interpretation of mammographic images.