Abstract P2-08-09: Nomogram update based on TAILORx clinical trial results - Oncotype DX breast cancer recurrence score can be predicted using clinicopathologic data

Abstract

Abstract Oncotype DX (ODX) recurrence score (RS) breast cancer (BC) assay provides prognostic and predictive BC recurrence information for hormone(+)/node(-) patients (pts). Pts with low ODXRS (0-10) can safely forego adjuvant chemotherapy (ACH), while ACH is recommended for high ODXRS (≥26). No ACH recommendations were previously available for intermediate (11-25) ODXRS until 6/3/2018, when the TAILORx clinical trial results were presented and e-published. According to the new data, pts with ODXRS 11-25 can now safely forego ACH, although some benefit of ACH was found in pts ≤50 with ODXRS 16-25. These new data now allow us to categorize ODXRS as a binary variable. Since ODX is a costly assay, we previously developed and published a user-friendly nomogram based on clinicopathologic characteristics (CPC) of ODX tested patients captured by the National Cancer Data Base (NCDB) as a surrogate prediction model for the ODX assay. As intermediate score patients were excluded from our previously created nomogram, the objective of this update is to test the predictive value of CPC variables for forecasting the new TAILORx binary ODXRS stratification using the entire NCDB population of ODX tested pts. Five CPC variables (age, tumor size, grade, progesterone receptor status (PR) and the 4 most frequent BC histologic types) were assessed with logistic regression to predict for a low- or high-risk ODXRS test results using 0-15 or 0-25 and 16-100 or 26-100 for a low- and high-ODXRS, respectively. These ranges were used in the TAILORx trial. A training cohort consisted of 65,754 ODX tested ER+/HER2-/lymph-node-negative pts with 6-50mm tumor size, captured by the NCDB from 2010-2014; 18,585 ODX tested pts in 2015 served as an external validation cohort. The predictive accuracy of the regression model was yielded using a Receiver Operator Characteristic (ROC) analysis. Model fit was analyzed by plotting the predicted probabilities against the actual probabilities. Grade and PR are the most significant predictors for a low- or high-risk ODXRS, followed by tumor size, histologic tumor type and age for any of the tested cut-off values. The ROC curves showed the best agreement between the nomogram prediction and actual observation for 0-25 (low) and 26-100 (high) ODXRS cut-off value (C-index=0.81). Overall, our model correctly predicted for 99.2% of low-risk and 18.6% of high-risk ODX cases. These cut-off values were used for building the updated nomogram model. Predicting for Low-risk (LR) ODXRSTraining cohort N=65,754Points assignedAge (19-90)0-9Tumor Size (6-50mm)31-0G1100G262G30PR+ ≥1%70PR- <1%0IDC0ILC31IDC+ILC21IDC+other6 Total Points 0-241Probability LR ODXRS% accurately predicted LR ODX160.9195132.893112.79298.69024.1595% accurate for High-risk ODXRSIDC=Invasive ductal BC ILC=invasive lobular BC G=grade An updated nomogram is now available, created and validated based on the entire population of ODX tested pts (84,339) captured by the NCDB from 2010-2015. The updated nomogram correctly predicts for 99.2% of a low ODXRS with a 0.81 C-index. This revised calculator will continue serving as a surrogate for BC pts older than 50 for which ODX testing is not necessary, affordable or available. Citation Format: Orucevic A, McNeil ML, Bell JL, McNabb AP, Heidel RE. Nomogram update based on TAILORx clinical trial results - Oncotype DX breast cancer recurrence score can be predicted using clinicopathologic data [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P2-08-09.