Reader variability and preditive positive value for mammography BI-RADS TRADE MARK category 4 / Variação interobservador e valor preditivo positivo da categoria 4 do BI-RADS MARCA REGISTRADA para mamografia




Purpose: To determine the predictive positive value of BI-RADS 4th edition categories 4A, 4B and 4C and determine the inter-observer variability for lesion on which biopsies had been performed. Methods: This cross-sectional study was conducted at the National Institute of Cancer (INCA), Rio de Janeiro, Brazil. Cases for analysis were selected from the medical records among women who were submitted to breast needle localization and diagnostic surgical biopsy, from January 1999 to December 2002, and which had been previously categorized as BIRADS 4. The mammograms were reviewed by three senior breast imaging radiologists. Readers provided a BI-RADS final assessment category to indicate the probability of malignancy in category 4. For statistical calculations, a ?consensus? variable was created using the outputs from the three assessors. Interobserver variability for each subcategory was determined using the weighted kappa statistic. Results: Of 775 lesions, ?consensual? assessment rendered 235 (30.3%) mammograms classified as 4A, 338 (43.6%) as 4B and the remainder 161 (20.8%) as 4C. Radiologists did not reach a consensus in 41 (5.3%) exams. Overall agreement between raters was moderate (?=0.44), ranging from ?=0.38 between raters two and three to ?=0.54 between radiologists one and two. Agreement between raters ranged from slight (category 4B, k=0.09) to fair, considering categories 4A and 4C. In BI-RADS 4B and 4C groups, the proportion of each abnormality in the whole group was similar to that in 4A, but the PPV were higher. Conclusions: The PPV for subcategories 4A, 4B and 4C was progressive and the agreement between readers was moderate. Key words: Breast cancer, mammography, BI-RADS


mammography mamas - cancer breast mamografia mamas - radiografia

Documentos Relacionados