Abstract 71P
Background
Predictive biomarkers that guide rational selection of therapeutic agents represent an unmet clinical need in patients with advanced soft-tissue sarcomas (STS). While signatures like GemPred and GemCore have been developed for gemcitabine efficacy in Pancreatic Adenocarcinomas, no such tool exists for STS. To address this need, we developed GemSarc, the first predictive transcriptomic-based signature for gemcitabine efficacy in STS.
Methods
Participants with STS enrolled in the prospective Electronic Health Record (EHR)-Based Comprehensive Bone and Soft Tumor Registry (NCT02677961) at The Ohio State James Cancer Center who received at least one cycle of gemcitabine-based chemotherapy and had transcriptomic data were included (n= 25). Additionally, transcriptomic data from 11 STS cell lines from the Cancer Cell Line Encyclopedia (CCLE) was included. Genes correlated with Gemcitabine duration of therapy and IC50, in vitro, were identified using Random Forest Machine Learning. An 18-gene signature was built using multivariate Cox Regression with a lasso penalty. This signature was validated in an independent cohort of 27 STS patients from the Oncology Research Information Exchange Network (ORIEN).
Results
GemSarc included genes related to nuclear transport, nucleotide metabolism and cell cycle progression. Patients in the low-risk gene signature group had a significantly longer time on gemcitabine therapy compared to patients in the high-risk group: (11.39 months vs 2.67 months, HR 0.2341, 95% CI: 0.082-0.66, p<0.01). Validation in the ORIEN cohort demonstrated similar findings as low risk conferred significantly longer time on therapy compared to high-risk: (7.98 months vs 1.83 months HR 0.2987, 95% CI: 0.128-0.694, p<0.01). Importantly, GemSarc did not predict overall survival (p = 0.146) nor correlate with histologically observed mitoses (R2 = 0.0058). Furthermore, GemSarc was specific to STS as it did not predict gemcitabine efficacy in non-STS, including Non-Small Cell Lung (p=0.386), Bladder (p = 0.742), and Ovarian cancer (p = 0.831).
Conclusions
Our study generated the first genetic signature that predicts gemcitabine duration of therapy in STS patients. Prospective validation of this predictive signature is warranted.
Clinical trial identification
NCT02677961.
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
The Ohio State University.
Disclosure
D. Leibner: Financial Interests, Personal, Advisory Board: AADI Biosciences. J. Chen: Financial Interests, Personal, Full or part-time Employment: Tempus Genomics Inc. G. Tinoco: Financial Interests, Personal, Advisory Board: ESMO Open. All other authors have declared no conflicts of interest.