137P Adherence to and optimization of guidelines for risk of recurrence/Prosigna testing using a machine learning model: A Swedish multicenter study

Gene expression profiles are used for decision making in the adjuvant setting of hormone receptor-positive, HER2-negative (HR+/HER2-) breast cancer. Previous studies have reported algorithms to optimize the use of RS/Oncotype Dx. However, no such efforts have focused on ROR/Prosigna. In addition, there is no data on adherence to testing guidelines. Postmenopausal women with resected HR+/HER2- and node-negative breast cancer that had undergone testing with ROR/Prosigna in four Swedish regions between March 2020 and March 2022 were included (n=348). In addition, we identified all patients that had an indication for gene expression profiling, regardless of whether it was performed. Using the ROR/Prosigna recommendation for chemotherapy as ground truth, we compared the performance of four risk classifications in terms of over- and undertreatment, and patients at intermediate risk. We then developed and validated, by using a 0.7/0.3 splitting rule in the cohort, a machine learning model that comprised simple prognostic factors (size, PgR expression, grade and Ki67) for prediction of ROR/Prosigna outcome. Adherence to guidelines reached 66.3%, with non-tested patients being older and having more comorbidities (p<0.001). Previous risk classifications led to excessive undertreatments (CTS5: 21.8%, MINDACT/TailorX risk definitions: 28.1%) or large intermediate groups that would need to be tested with gene expression profiling (Ki67 10%/40% cut-offs: 86.5%). The model achieved AUC under ROC for predicting ROR/Prosigna result of 0.77 in the training and 0.83 in the validation cohort. By setting and validating upper and lower cut-offs in the model, we could improve correct risk stratification while decreasing the proportion of patients needing testing with gene expression profiles compared to current management. We show the feasibility of machine learning algorithms to improve patient selection for gene expression profiling. Further validation in external cohorts is needed.