MicroRNA-based biomarkers of outcome in soft tissue sarcoma treated with hypofractionated preoperative radiation therapy.

Abstract

11538 Background: Soft tissue sarcomas (STS) are rare, aggressive malignancies with high variability in response to radiation therapy (RT). Our previous research in a small cohort who received preoperative hypofractionated RT (SBRT) on a phase II trial suggested that germline microRNA-based single nucleotide polymorphisms (mirSNPs) could identify patients at increased risk of major wound toxicity (MWT). This study explores the potential of this class of biomarker to predict MWT and additional outcomes to preoperative SBRT in STS. These findings could guide personalized treatment strategies and possibly the choice of RT regimen in the future. Methods: We analyzed 110 patients with high-risk extremity or trunk STS treated with five-day preoperative SBRT (30 Gy in five fractions). Over 100 mirSNPs were evaluated for their ability to predict RT outcomes, including late toxicity, MWT, distant metastases, and pathological response. Pathological response was defined as a necrosis score >= 70, late toxicity as grade >=2 at 2 years, and MWT as grade >= 3. Preliminary genetic models were developed using elastic net, random forest, and boosted tree algorithms and evaluated using leave-one-out cross-validation (LOOCV) performance metrics. mirSNPs were pre-filtered using Fisher or Jonckheere-Terpstra p-values (<0.2) for relevance to outcomes. Results: We developed preliminary genetic signatures to accurately predict 4 different outcomes in sarcoma patients undergoing preoperative SBRT: late toxicity (AUC=0.830), distant failure (AUC=0.775), pathological response (AUC=0.765), and major wound toxicity (AUC=0.736). Our pathological response genetic model has balanced sensitivity (0.750) and specificity (0.781), suggesting it could reliably predict which patients respond well to SBRT. Conclusions: Our study highlights the promise of mirSNP-based models in predicting STS outcomes to preoperative SBRT. By identifying patients with favorable versus unfavorable responses, these models could help identify patients who could be considered for preoperative SBRT versus standard fractionated radiation, as there may be fractionation-dependent radiation toxicity in sarcoma, as has been identified in other malignancies. Future directions include investigating associations among identified outcomes, inclusion of additional clinical variables, and comparison to toxicity with preoperative standard fractionated radiation. These efforts are significant steps towards paving the way for more personalized sarcoma care. Clinical trial information: NCT02701153 . LOOCV performance metrics in sarcoma data (n=110). Outcome Sensitivity Specificity PPV NPV F1 Score AUC Any Late Toxicity (Grade 2 at 2 Years) 0.786 0.875 0.478 0.966 0.595 0.830 Distant Failure 0.680 0.871 0.607 0.902 0.642 0.775 Path Response (Necrosis Score >= 70) 0.750 0.781 0.600 0.877 0.667 0.765 Major Wound Toxicity 0.667 0.805 0.595 0.849 0.629 0.736