Developing a nomogram prediction model to enhance diagnostic accuracy of supplemental ultrasound post-negative mammography

The effectiveness of mammography in women with dense breasts is compromised by a high rate of false-negative results. While supplemental ultrasound increases sensitivity, its low positive predictive value (PPV) leads to more unnecessary biopsies. This study aims to develop a nomogram model to predict the malignancy of breast masses that are additionally identified as suspicious by supplemental ultrasound after an initial negative screening mammography. The goal is to improve the PPV of supplemental ultrasound and potentially reduce unnecessary biopsies. In this study, eligible data were collected retrospectively and then randomized into training and validation sets. The Least Absolute Shrinkage and Selection Operator was used to identify the most important predictive variables in the training set. The maximum Youden index determined the optimal model threshold, and model performance was evaluated using receiver operating characteristic curves, calibration curves, decision curve analyses, and metrics such as sensitivity, specificity, PPV, and negative predictive value. The study included 425 breast masses, 345 benign and 80 malignant. These were divided into 298 for the training set and 127 for the validation set. Least Absolute Shrinkage and Selection Operator identified the 5 most important predictive variables for the construction of the model. The model showed strong discrimination with area under the curve values of 0.91 (0.87-0.95) for the training set and 0.88 (0.81-0.96) for the validation set. Hosmer-Lemeshow tests indicated a good model fit, with P-values of 0.78 and 0.12 for the training and validation sets, respectively. In addition, decision curve analyses highlighted the clinical utility of the model. The model also showed commendable diagnostic performance in terms of sensitivity, specificity, PPV, and negative predictive value. The nomogram model significantly increased the PPV of supplemental ultrasound from 0.18 to 0.56 in the training set and from 0.21 to 0.56 in the validation set. This study successfully developed a nomogram model to predict the malignancy of suspicious breast masses additionally identified by supplemental ultrasound. The model shows robust performance and significantly improves the PPV of supplemental ultrasound, suggesting a promising way to reduce unnecessary biopsies in such cases.