Predicting Chemotherapy Benefit across Different Races in Early-Stage Breast Cancer Patients Using the Oncotype DX Score

Simple Summary The Oncotype DX recurrence score is widely used to help clinicians treat patients with early-stage breast cancer. However, the threshold at which the Oncotype DX score is used to guide chemotherapy treatment may vary for different patient groups, and the present study describes a machine learning method to quantitatively determine the optimal chemotherapy sensitivity threshold. Utilizing publicly available breast cancer survival data, we demonstrated that 2.05-2.72x more lymph-node-negative and 2.08-5.02x more lymph-node-positive patients may benefit from receiving chemotherapy at a lower Oncotype DX score than current guidelines (RS > 25 or RS > 30) suggest. Additionally, our models indicate a racial difference in chemotherapy benefit that may help physicians provide tailored treatment to specific patients. Specifically, white, black, and Asian women with early-stage ER+/HER2-/LN- tumors may benefit from chemotherapy when their Oncotype DX scores are greater than 19.9, 37.2, and 18.0, respectively. Background: Oncotype DX assay, a multigene molecular test, has been widely used to stratify relapse risk and guide chemotherapy treatment in breast cancer. However, the optimal threshold of the Oncotype DX score in predicting chemotherapy benefit and its racial variation has not been investigated. Methods: In this study, we apply a random forest survival model to the SEER-Oncotype cohort data (Surveillance, Epidemiology, and End Results with Oncotype DX test information for breast cancer patients) and determine chemotherapy benefit thresholds in early-stage, estrogen-receptor-positive (ER+), and HER2-negative (HER2-) patients of different races. Results: Our results indicate that early-stage ER+, HER2-, and LN-/LN+ patients may benefit from receiving chemotherapy at a lower Oncotype DX score than current guidelines (Recurrence Score, RS > 25 or RS > 30) suggest. According to the estimated chemotherapy sensitivity thresholds from our models, 2.05-2.72-fold more lymph-node-negative (LN-) and 2.08-5.02-fold more lymph-node-positive (LN+) patients who may not currently be recommended for chemotherapy by their Oncotype DX test result may actually have the potential to benefit from chemotherapy. Furthermore, our models indicate a racial difference in chemotherapy benefit: white, black, and Asian women with early-stage ER+/LN- tumors benefit from chemotherapy when their Oncotype DX scores are greater than 19.9, 37.2, and 18.0, respectively. Conclusions: Our study provides a method for calibrating multigene molecular tests to help guide treatment decisions in racially and ethnically diverse patients with cancer. Specifically, we identify key chemotherapy sensitivity thresholds for the Oncotype DX recurrence score test in breast cancer patients and provide evidence that certain patients may benefit from receiving chemotherapy at a lower threshold than the current clinical guidelines suggest.

Automated summary

In this study, a machine learning method is applied to determine the optimal threshold for using the Oncotype DX score to guide chemotherapy treatment. The results show that more lymph-node-negative and lymph-node-positive patients may benefit from chemotherapy at a lower Oncotype DX score than current guidelines suggest. Additionally, the study reveals a racial difference in chemotherapy benefit, providing evidence for tailored treatment for specific patients.