A Validated Radiomics T-cell Score Predicts Response to Multi-site SBRT Combined with Pembrolizumab

Abstract

The clinical utility of immune checkpoint blockade (ICB) is rapidly advancing. Combining ICB with local cytoreductive therapy, specifically with stereotactic body radiotherapy (SBRT), remains an active area of research. However, no prognostic markers exist to identify patients likely to benefit from combined therapy. The degree of T-cell-mediated immunity is a potential biomarker for predicting response to immunotherapy and prior work demonstrates a potential role for radiomics to identify T-cell infiltration from standard computed tomography (CT) imaging. Herein we test whether a validated radiomics score (RS) can predict response and survival in the setting of multi-site SBRT with pembrolizumab (pembro). The RS was determined for patients with metastatic treatment-refractory adult solid tumors that received SBRT (30-50 Gy, 3-5 fractions) to 2-4 metastases (mets). Pembro 200 mg IV Q3W began ≤7 days following SBRT. RS for mets was calculated from CT scans using a previously validated formula and the software for radiomic feature extraction. Eight variables (5 radiomic features) calculated the pre-treatment RS. RS was assessed for its ability to predict local response, progression free survival (PFS), and overall survival (OS) after SBRT and pembro. RS was analyzed as a dichotomous variable at both a metastasis level and from the average RS score of all irradiated metastases within patients. The bottom 25th percentile of RS or lower were coded as low-RS while > 25th percentile was deemed as high-RS. Log-rank test and Cox regression were used to compare differences in PFS and OS based on dichotomized RS. Data were analyzed on statistical software and a p-value ≤ 0.05 was considered significant. A total of 68 patients and 139 tumors (53 lung mets) were analyzed. Compared to high-RS tumors, mets with low-RS had a 5 times higher rate of local tumor failure (Shared-frailty Cox hazard ratio [HR]=5.72; 95% CI 1.35-24.35; p=0.018). Patients with low-RS were significantly less likely to be SBRT responders (Fisher’s exact test, p=0.012). Furthermore, low-RS patients had inferior PFS (median: 2.76 vs. 3.38 months, log-rank p=0.01; HR=2.15, p=0.01) and OS (median: 6.21 vs. 13.3 months, log-rank p=0.0036; HR=2.57, p=0.005) compared to patients with high-RS. High RS values correlated with transcriptional activation of innate and adaptive immune signatures in biopsies from irradiated tumors (n=35). The RS predicts local irradiated tumor control and OS for patients treated with combined SBRT and pembro. Transcriptional analysis from a subset of patients is concordant with immune activation in high RS tumors. The RS may allow clinicians to utilize imaging at the time of pre-treatment CT to predict which patients will not benefit from combined therapy and avoid associated toxicity.Abstract 151; Table 1End point9 month Low-RS9 month High-RSTreated metastasis control78.7%95.8%Progression-free survival7.8%22.0%Overall survival26.1%62.2% Open table in a new tab