Harnessing cell-free orphan non-coding RNAs as a predictive measure of long-term survival in neoadjuvant breast cancer therapy.

Abstract

585 Background: Pathologic complete response (pCR) is a valuable metric for predicting survival following neoadjuvant therapy (NAT) in breast cancer; however, blood-based biomarkers may refine the prediction of treatment outcomes and inform follow-up treatment decisions. Orphan non-coding RNAs (oncRNAs) are a novel category of small RNAs that are present in tumors and largely absent in healthy tissue. Here, we explore the feasibility of a tumor-naive oncRNA-based liquid biopsy assay for predicting distant recurrence-free survival (DRFS) in breast cancer patients using a standard blood draw. Methods: Our cohort included 538 high-risk breast cancer patients from the multicenter I-SPY2 trial having serum available after NAT and before surgery (timepoint T3). Of the 538 patients (mean age: 48.8 ± 10.3), 191 (36%) achieved pCR. We isolated small RNA from 0.8 mL of patient serum and sequenced it at an average depth of 50 million 100-bp single-end reads. We used our catalog of tumor-derived oncRNAs, discovered in The Cancer Genome Atlas, to develop an AI model for predicting tumor burden based on tumor size. Our AI model was trained on oncRNA profiles from an independent cohort of 719 serum samples collected at diagnosis from individuals with cancer. We assessed tumor burden by oncRNA in the I-SPY2 cohort at T3 and associated this measure with DRFS using robust cutpoints established for each of the two pCR groups: high- ( n=70) and low-burden ( n=468) cutpoints were verified through cross-validation and almost all folds converged on a single value, which we used in our analysis. Results: The continuous oncRNA-derived tumor burden at T3 had a hazards ratio (HR) of 2.4 (95% CI: 1.1–5.1) for predicting DRFS. This measure remained significant in multivariate Cox models that included pCR (HR = 2.7 [1.3–5.6]) or residual cancer burden (RCB II/III versus 0/I; HR = 3.4 [1.6–7.1]). Patients with pCR had a 3-year DRFS of 94%; patients with pCR having low ( n=180) versus high ( n=11) oncRNA burden resulted in 95% and 73% 3-year DRFS, respectively (HR = 7.4 [2.3–24]). Those without pCR had a 3-year DRFS of 78% and stratification by oncRNA burden resulted in 85% and 71% 3-year DRFS for the low ( n=288) and high ( n=59) burden groups, respectively (HR=2.4 [1.5–3.9]). Importantly, we saw no restricted mean survival time difference (RMST) at 3 years between those with and without pCR among the high oncRNA burden group (0.1-month difference, p=0.97). Conclusions: We demonstrate the utility of a tumor-naive, oncRNA-derived tumor burden AI model as a predictive measure of treatment outcome and as a prognostic biomarker that can be used in addition to pCR to further stratify patients as low or high risk of recurrence. Integration of our oncRNA assay with existing biomarkers and clinicopathologic variables, has the potential to add value for predicting outcomes and prioritizing patients that would benefit from additional care.