Observer variation in the grading of screen detected mammographic abnormalities. 1. Assessment of reproducibility

Interpretation of mammograms is a subjective process liable to observer error. This study was performed to examine the level of agreement and reproducibility between radiologists in the identification and assessment of mammographic abnormalities. Data were collected prospectively utilizing mammograms from women attending breast screening. Diagnostic consistency in interpretation of abnormalities classified into four categories (malignant, suspicious of malignancy, suspicious — probably benign and benign) and showed complete agreement between radiologists in 62% of cases (kappa 0.44). Complete intra-observer agreement was 76% and 66% (kappa 0.62 and 0.51) for each of the radiologists. The findings show differences in observer interpretation of mammographic lesions do occur and support the need for double reading of mammograms.