Background: Numeric grading systems have been validated for the reporting of mammographic abnormalities and for breast cytology. The impact of integrating lesion grades from both disciplines on the accuracy of assessment of screen-detected lesions has not been investigated.
Methods: In the current study, all lesions were prospectively classified using the Tabar radiologic grading system, as well as the National Health Service Breast Screening Programme (NHSBSP) cytologic reporting system. For lesions assessed between January 1996 and January 2003, based on final histology or 12 months of follow-up, positive predictive values (PPV) were calculated for each imaging grade and for each cytologic grouping. After integration of the data, changes in the PPV and the negative predictive values (NPV) of cytology among lesions with varying imaging grades were tracked.
Results: Data were retrieved for 4806 lesions. The differences in the rates of malignancy for lesions in the different imaging grades were significant (Grade 5, 95.8%; Grade 4, 54.6%; Grade 3, 11.6%) (P < 0.001). Similarly, the cytologic categories stratified lesions into groups with significantly different rates of malignancy (positive, 99.5%; suspicious, 89.2%; atypical, 43.4%; benign, 5.7%; inadequate, 33.3%) (P < 0.001). Integration of cytologic results with the imaging grade of lesions led to significant improvements in the PPV and NPV. Positive smears were likely to represent malignant lesions in 99.9% of Grade 5 lesions, 99.2% of Grade 4 lesions, and 95.2% of Grade 3 lesions (P < 0.0001). Similarly, negative cytology corresponded to a nonmalignant lesion in 99% of Grade 3 lesions, 81.6% of Grade 4 lesions, and in only 45.5% of Grade 5 lesions (P < 0.001).
Conclusions: The integration of numeric grading schemes for breast imaging and cytology improves the accuracy of assessment of screen-detected lesions. Because 99.9% of Grade 5 lesions with positive cytology are malignant, core biopsy confirmation may not be required in this group.