0761 Adaptability of the Treating Obstructive Sleep Apnea Using Targeted Hypoglossal Nerve Stimulation (OSPREY) Trial

Abstract

Abstract Introduction With few exceptions, clinical trials of hypoglossal nerve stimulation (HGNS) for obstructive sleep apnea (OSA) are single-arm, open-label studies sometimes followed by short-term, unblinded, randomized withdrawal. By contrast, the THN3 study was a parallel-arm, randomized, controlled trial (RCT) of targeted HGNS (THN) in moderate to severe OSA, and provided higher-level evidence of HGNS safety and efficacy.Despite generating strong evidence, conventional RCTs are risky due to their inherent inflexible designs. We therefore launched a confirmatory THN RCT (OSPREY) with an adaptive, Bayesian “Goldilocks” design that optimize its sample size dynamically, yet achieve high-confidence results. Methods Four scenarios were simulated within the OSPREY design framework (randomized 2:1 Treatment:Control) for the primary endpoint of apnea-hypopnea index (AHI) response rate (RR): nominal with results equal to those of THN3 (Treatment AHI RR 52%/Control AHI RR 20%), improved Treatment RR (63%/20%), worsened Treatment RR (41%/20%) and null [Treatment RR=Control RR] (20%/20%). Each scenario was simulated 10 times with 10,000 simulations of each interim analysis. Subject outcomes were determined by randomly drawing from a binomial distribution with the relevant AHI RR.Interim analyses in OSPREY begin at 50 randomized subjects and repeat every 20 additional subjects to the maximum sample size of 150, with opportunities for early success and futility at each milestone to generate high-confidence results from an optimal sample size. OSPREY assesses secondary endpoints including quality of life inventories (Epworth Sleepiness Scale; Functional Outcomes of Sleep Questionnaire; EQ-5D, SF-6D and PROMIS sleep questionnaires) and oximetry metrics (Oxygen Desaturation Index, %sleep time below 90% oxygen saturation). Previous results suggest secondary endpoints will be adequately powered at the final sample size determined by AHI RR. Results Simulations produced the following outcomes formatted as [scenario: randomized sample size, overall success rate, probability of early success, mean success probability]: null: 150, 0%, 0%, 2.47%; nominal: 130-150, 100%, 80%, 95.3%; improved: 90-130, 100%, 100%, 98.9%; worsened: 150, 100%, 0%, 68.6%. Conclusion OSPREY is uniquely able to adapt to various Treatment/Control response scenarios and should provide high-confidence confirmation of the safety and efficacy of THN therapy in moderate to severe OSA. Support (If Any) LivaNova