Putting Together the Pieces: Development and Validation of a Risk Assessment Model for Nipple-Sparing Mastectomy
Jordan D. Frey, M.D., Ara A. Salibian, M.D., Mihye Choi, M.D., Nolan S. Karp, M.D..
NYU Langone Health, New York, NY, USA.
Background: Optimizing outcomes and assessing appropriate candidates for breast reconstruction after nipple-sparing mastectomy (NSM) is an ongoing goal for plastic surgeons.
Methods: All patients undergoing NSM from 2006 to June 2018 were reviewed and randomly divided into test and validation groups. A logistic regression model calculating the odds ratio for any complication from 12 risk factors was derived from the test group while the validation group was used to validate this model.
Results: The test group was comprised of 537 NSMs (50.2%) with an overall complication rate of 27.2% (146 NSMs). The validation group was comprised of 533 NSMs (49.8%) with an overall complication rate of 22.9% (122 NSMs). A logistic regression model predicting overall complications was derived from the test group. NSMs in the test group were divided into deciles based on predicted risk in the model. Risk increased with probability decile; decile 1 was significantly protective while deciles 9 and 10 were significantly predictive for complications (p<0.0001). The relative risk in decile 1 was significantly decreased (0.39; p=0.006); relative risk in deciles 9 and 10 was significantly increased (2.71; p<0.0001). The model developed using the test group was then applied to the validation group utilizing the identical procedures. A histogram plotting probability risk of overall complication based on this test group model against the frequency of NSMs in the validation group demonstrates a similar skew compared to the similar histogram using the test group, indicating good fit of the model on a separate population of NSMs. Deciles 1 as well as combined deciles 9/10 for the validation group of NSMs using the model developed using the test group were then examined to further establish the fitness of the model. The relative risk of any complication in decile 1 of the validation group was decreased at 0.55 (95% Confidence Interval [CI]: 0.28-1.01) (p=0.057). The relative risk of any complication in the combined deciles 9/10 was significantly increased at 1.89 (95% CI: 1.37-2.61) (p<0.0001). Lastly, a receiver operating characteristic curve was performed for the validation group establishing a statistically significant area under the curve of 0.668 (p<0.0001), demonstrating that the model was diagnostically meaningful.
Conclusions: We establish and validate a predictive risk model and calculator for NSM with far-reaching impact for surgeons and patients alike.
Back to 2019 Posters