Predictive gene signature for the efficacy of pazopanib in solitary fibrous tumor: A Spanish Group for Research in Sarcoma (GEIS) study.

Abstract

11540 Background: Pazopanib was shown to be active in advanced Solitary Fibrous Tumor (SFT), with 58% and 51% of the typical and malignant/ dedifferentiated SFT patients, achieving an objective response by Choi criteria, in an international phase II clinical trial led by our team. Nonetheless, predictive biomarkers of pazopanib efficacy represent a clinical unmet need, to support the rational selection of this drug in this histology. We presented here a predictive transcriptomic-based signature for the efficacy of pazopanib in SFT. Methods: Patients enrolled in the GEIS 32 trial (ClinicalTrials.go ID: NCT02066285), testing pazopanib in two separate cohorts of SFT (typical and malignant/ dedifferentiated SFT), and with available tumor samples, were included in this study. Gene expression was assessed by direct transcriptomics, using the HTG EdgeSeq Oncology Biomarkers Panel (HTG Molecular Diagnostics, Inc.; Tucson, AZ, USA), according to manufacturers’ instructions. Raw counts were normalized by variance stabilizing transformation (VST) using DESeq2. Univariate Cox regression analysis was performed to identify the genes significantly associated with progression-free survival (PFS; p < 0.01). These remaining genes were used as input to build a gene expression signature, using a multivariate Cox regression applying a Lasso penalty (10-fold cross-validation). Risk scores were calculated by multiplying the expression of every gene with its corresponding Cox regression coefficient. Results: A series of 40 patients was included for data analyses, with a median age of 64 years old, 62.5% being females, and a median follow-up from pazopanib treatment of 18 months. A total of 24 (60%), 14 (25%), and 2 (5%) patients were diagnosed with malignant, typical, or dedifferentiated SFT, respectively. The predictive signature of pazopanib efficacy was built with 18 genes, identified as significant in the univariate analysis, applying the Lasso penalty. This signature included 13 and 5 genes associated with resistance or sensitivity to pazopanib, respectively. Genes overexpressed and associated with low PFS of pazopanib included CKS2 , FANCA, KPNA2, and CXL14 , among others. Patients in the high‐risk gene signature group (N = 23) showed a significantly worse PFS for pazopanib treatment, compared with patients in the low‐risk group (N = 17): [5.6 months (95% CI 3.7-10.0) vs. 10.0 months (95% CI 6.5-NR), p < = 0.012; HR = 1.25 (95% CI 1.1-1.4, p < 0.001]. The cut-off calculated by MAXSTAT was 42.515. Conclusions: Our study identified a novel 18-gene-based signature that significantly predicts the efficacy of pazopanib in SFT patients. Future studies will focus on the prospective validation of this predictive gene signature.