Presenter(s)
K. McGreevy1, D. Telesca2, R. K. Chin3, J. Deng3, R. R. Savjani3, D. I. Rosenthal4, A. K. Singh5, T. J. Galloway6, R. M. Lanning7, F. Alite Jr8, S. N. Ashton9, M. J. Awan10, J. J. Caudell11, K. Attwood12, Q. T. Le13, S. S. Yom14, and J. B. Weidhaas15; 1MiraDx, Los Angeles, CA, 2UCLA, Los Angeles, CA, 3Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA, 4Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 5Department of Radiation Medicine, Roswell Park Comprehensive Cancer Center, Buffalo, NY, 6Fox Chase Cancer Center, Philadelphia, PA, 7Department of Radiation Oncology, University of Colorado School of Medicine, Aurora, CO, 8Department of Radiation Oncology, Geisinger Health Systems, Danville, PA, 9Intermountain Healthcare, Cedar City, UT, 10Department of Radiation Oncology, Medical College of Wisconsin, Milwaukee, WI, 11Moffitt Cancer Center, Tampa, FL, 12American College of Radiology, Philadelphia, PA, 13Stanford University, Stanford, CA, 14University of California, San Francisco, San Francisco, CA, 15Department of Radiation Oncology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA
Purpose/Objective(s):
Radiation therapy and concurrent chemotherapy (CRT) is integral to the treatment of Head and Neck Squamous Cell Carcinoma (HNSCC). Despite technological advances, patients still experience severe CRT adverse events (AEs), with HNSCC having one of the highest AE burdens among cancers. Although there are some identified risk factors associated with AEs, it has not been possible to identify patients at highest risk, although evidence supports that differences are likely due to inherent radiosensitivity. In this analysis, we hypothesized that germline genetic variants, especially microRNA single nucleotide polymorphisms (mirSNPs), could better predict severe toxicity as well as survival in HNSCC patients treated with CRT. Here we developed genetic-based risk models and compare those to clinical risk models using data from a prospective trial.
Materials/Methods:
We evaluated 171 patients from RTOG 0522 who received =35 RT fractions and cisplatin alone. We develop genetic-based risk models for acute toxicity (Grade >3), late toxicity (Grade >4), and overall survival at 4 years. We filtered candidate mirSNPs and trained various machine-learning models to predict the three outcomes. Clinical variables included age, gender, dose, tumor/nodal- stage, cisplatin cycles, primary site, and HPV status. We applied leave-one-out cross-validation to evaluate performance. These genetic models were compared to models built using clinical variables only.
Results:
The integrated genetic-clinical models outperformed clinical-only models in predicting acute toxicity, late toxicity, and four-year overall survival. The acute toxicity signature has 82% sensitivity, 75% specificity, and overall AUC of 0.783. The late toxicity prediction improved from an AUC of 0.680 (clinical only) to 0.812 with mirSNPs. The overall survival model has an AUC of 0.743 and patients predicted to be high-risk had a median survival of 2.25 years compared to 4.46 years in the predicted low-risk group. The most important mirSNP in the late toxicity signature is in the BRCA2 3’ UTR. Our model estimates the odds of late AEs decrease by approximately 83.2% for every additional BRCA2 SNP after controlling for all other SNPs and clinical factors.
Conclusion:
Our findings suggest that integrating mirSNP data with clinical variables significantly enhances the ability to predict CRT-related acute and late toxicity as well as overall survival for HNSCC patients. Validation studies using the most current RT delivery technologies are ongoing and will be important to further the application of these results. However, these findings indicate the potential to identify high-risk patients which could enable best optimization strategies to improve treatment choice and outcome in HNSCC.
