Search

Background Ko’s scoring system was developed to predict malignancy upgrades in patients diagnosed with atypical ductal hyperplasia by core needle biopsy. The Ko algorithm was able to identify a subset of patients who were eligible for exclusively clinical follow-up. The current study statistically investigated the patient outcomes to determine whether this scoring system could be translated and used safely in clinical practice. Methods We tested the statistical performance of the Ko scoring system against an external independent multicentre population. One hundred and seven cases of atypical ductal hyperplasia diagnosed by an 11-gauge biopsy needle were available for inclusion in this study. The discrimination, calibration and clinical utility of the scoring system were quantified. In addition, we tested the underestimation rate, sensitivity, specificity, and positive and negative predictive values according to the score threshold. Results The overall underestimation rate was 19% (20/107). The area under the receiver operating characteristic curve for the logistic regression model was 0.51 (95% confidence interval: 0.47–0.53). The model was not well calibrated. The lowest predicted underestimation rate was 11%. The sensitivity, specificity, positive predictive value, and negative predictive values were 90%, 22%, 20%, and 89%, respectively, according to the most accurate threshold proposed in the original study. Conclusion The scoring system was not sufficiently accurate to safely define a subset of patients who would be eligible for follow-up only and no additional treatment. These results demonstrate a lack of reproducibility in an external population. A multidisciplinary approach that correlates clinicopathological and mammographic features should be recommended for the management of these patients.