Optimized Decision Support For Selection of Transoral Robotic Surgery or (Chemo)radiation: Quantified Pre-therapy Risk Stratification for Patient-reported and Clinician-graded Swallowing Impairment and Toxicity

Abstract

Purpose: To develop a decision-making tool to choose the optimal treatment for oropharyngeal squamous cell cancer (OPSCC) patients who are eligible for primary transoral robotic surgery (TORS) and primary (chemo)radiation therapy with comparable locoregional control and survival. Methods: Decision tree models were constructed to study two decision-making scenarios, 1) TORS vs. definitive radiation therapy (RT), and 2) TORS vs. definitive chemo(radiation) therapy (CRT) based on well-established objective and subjective swallowing-function instruments, MD Anderson Dysphagia Inventory (MDADI), MD Anderson Symptom Inventory-Head and Neck Module (MDASI-HN), and clinician-rated Dynamic Imaging Grade of Swallowing Toxicity (DIGEST) that measure swallowing-related toxicity pre-therapy, 3-6 months (short-term), and 18-24 months (long-term) after therapy for five treatment cohorts (RT, CRT, TORS, TORS with adjuvant RT, and TORS with adjuvant CRT). The optimal treatment was sought as a function of postoperative extranodal extension (ENE) and/or positive margin (PM) that can trigger the adjuvant therapy. 2D heatmaps were constructed indicating the thresholds of postoperative events likelihoods required for TORS or definitive therapy to become the optimal treatment. Additionally, a risk calculation model was developed to quantify the risk associated with TORS in the settings that estimation of postoperative complications likelihoods may not be available. Results: Under the first scenario and for short-term measures, MDADI and MDASI instruments indicate the superiority of definitive therapy to TORS at all times, while DIGEST required a maximum of 40% likelihood for both ENE and PM events to indicate TORS as the optimal treatment. For 18-24 months measures, MDASI indicated TORS as the optimal treatment; however, MDADI- and DIGEST-based long-term measures indicated threshold likelihoods of 90% and 25%, respectively, for TORS to remain the optimal treatment. For short-term outcomes, TORS resulted in higher toxicity even when the likelihood associated with postoperative tumor resection margin (TM) are extremely low. For higher probability of postoperative TM, all instruments indicated high risk associated with TORS (>83%). For long-term swallowing-related toxicity, TORS remained the optimal therapy independent from the probability of postoperative TM based on MDASI instrument. However, MDADI-based measure assigned a high risk to TORS (>86%) when postoperative TM is extremely likely. DIGEST-based measures indicated a very high risk associated with TORS independent from the postoperative TM likelihood (>91%). Under the second scenario (TORS vs. definitive CRT), Both MDASI- and MDADI-based short-term measures indicated TORS as the optimal therapy independent from postoperative ENE/PM events. However, according to the DIGEST-based measure, definitive RT remained the optimal therapy when the probability of postoperative ENE and PM events exceed 80%. In this case, TORS was the optimal therapy if both events were not relatively likely (<55%). The same result was observed for MDADI and MDASI instruments for long-term measures indicating TORS as the optimal therapy independent from postoperative PM or ENE events. However, the DIGEST measure indicated that TORS is the optimal therapy only if the likelihood of both events are very low (<20%). When the postoperative TM is very unlikely (<10%), all instruments indicated TORS as the optimal therapy based on short-term outcomes. However, when the postoperative TM is extremely likely (>90%), DIGEST-based measures demonstrated the superiority of definitive CRT once the likelihood of postoperative ENE or PM is, at least, 60%. Long-term measures indicated higher sensitivity to postoperative TM likelihood with both MDADI- and DIGEST-based measures indicating a moderately high risk (> 60%) with TORS causing higher clinician-rated swallowing toxicity compared to definitive CRT when postoperative TM is extremely likely. Conclusion: The current study using decision modeling shows proof of concept that in the absence of reliable estimation of postoperative ENE/PM events concurrent with significant postoperative positive margins (i.e., more than 2mm) that can trigger adjuvant therapy, the overall toxicity level incurred by OPSCC patients undergoing TORS may become more severe compared to patients receiving non-surgical treatments thus advocating definitive (C)RT protocols. The results further demonstrated that, when available, the likelihoods of postoperative events triggering postoperative adjuvant therapy must be incorporated when choosing the optimal treatment plan for eligible patients.