Upstaging Prediction Model to Guide the Application of Sentinel Lymph Node Biopsy in Patients With Ductal Carcinoma In Situ: A Retrospective Comparative Study.

Abstract

Background and objectives: Indication for sentinel lymph node (SLN) biopsy in ductal carcinoma in situ (DCIS) patients with high-upstaging risk remains inconsistent. Our previous systematic review and meta-analysis had reported five variables that were significantly higher in the upstaging group. We developed the "high-risk upstaging model" and investigated its predictivity and accuracy.

Methods: The study included patients initially diagnosed with DCIS in a medical center between 2011 and 2020. Patients' clinicopathological data were obtained through web-based surgical medical record database. Two prediction models were built, in which patients who met at least one (Model A) or two (Model B) of the predictors would be predicted to upstage in the final pathology. We compared the accuracy of our models with National Comprehensive Cancer Network (NCCN) guideline and original data.

Results: The analyses included 249 patients, of which 67 DCIS patients upstaged in final pathology. The excess treatment in Model A (70%) was lower than the original data (80.2%). The incomplete treatment in Model A (3%) was lower than the NCCN guideline model (38.8%) and the original data (7.5%). Both Model A and Model B yielded a higher receiver operating characteristic (AUC) curve compared with original data.

Conclusions: Our Model A derived from the systematic review of the real-world data reduced the incomplete treatment rate of SLNB. Our Model B also showed the highest predictive value. With the two models, we provided a clearer indication for surgeons to perform SLNB in DCIS patients and demonstrated proof of concept, allowing ready input of patient data.