Abstract
OBJECTIVE: To document the mammographic changes after neoadjuvant chemotherapy with histopathological correlation, to calculate the accuracy of mammography (MG) in predicting residual tumour size and to measure the interobserver agreement in reading mammograms. METHODS: In 446 consecutive cases, the pre- and post-chemotherapy mammograms were retrospectively evaluated by two blinded observers, and consensus findings were compared with reference standard of surgical specimen. The accuracy of MG in predicting residual tumour size was calculated. Kappa statistics were calculated for measuring the interobserver agreement for reading mammograms. The sensitivity, specificity, positive-predictive value and negative-predictive value for the prediction of residual disease were calculated. RESULTS: The most common primary abnormalities were mass lesions without and with microcalcifications. After chemotherapy, there was decrease in size of most (95.1%) of the measurable masses, with decrease in the mean tumour size from 4.1 to 2.5 cm. The density of the tumour decreased in 66.6% (241/362) cases with residual disease. There was almost perfect interobserver agreement for describing the primary abnormality in the pre- as well as post-chemotherapy mammograms (k = 0.87 and 0.81, respectively) with substantial agreement for measurement of the mass lesions before and after chemotherapy (k = 0.69 and 0.68, respectively). MG showed accuracy of 60.0%, sensitivity of 94.4%, specificity of 50.0%, positive-predictive value of 91.3% and negative-predictive value of 61.8%. CONCLUSION: MG remains a highly sensitive and reproducible investigation for the assessment of residual disease after chemotherapy. ADVANCES IN KNOWLEDGE: There is substantial interobserver agreement in characterizing and measuring breast tumours on mammograms.