D97-09 False Positive Responses in Hypoglossal Nerve Stimulation Clinical Trials

Abstract Rationale Hypoglossal nerve stimulation (HGNS) trials have largely assessed therapeutic efficacy based on response rates (RRs) in single-arm observational cohorts rather than randomized controlled trials (RCTs). RRs by Sher criteria (≥50% reduction in apnea-hypopnea index (AHI) to AHI20/hr) and ≥25% reduction in oxygen desaturation index (ODI) are commonly reported; however, the absence of control participants and the lack of covariate adjustment have precluded systematic evaluation of false-positive responses. We hypothesized that relevant covariates would explain the false-positive responses among controls from the THN3 and OSPREY RCTs. Methods Control group outcomes for AHI and ODI from HGNS RCTs, along with covariates including baseline OSA severity, the proportion of time spent in REM vs non-REM sleep, and the proportion of time spent in supine vs non-supine position, were evaluated. Summary statistics and response rates were computed. Univariable and multivariable logistic and linear regression models were used to assess the significance of covariates on binary and continuous response outcomes. Results Control group RRs were analyzed from the THN3 (N = 46) and OSPREY (N = 37) RCTs at the month 4 and 7 follow-up visits, respectively. Pooled data showed some control participants meeting definitions of response (AHI RR, 14/83; 16.9% 9.5%, 26.7%; ODI RR, 35/83; 42.2% 31.4%, 53.5%) despite AHI point estimates remaining unchanged from baseline (mean±SD, 35.7±10.3/hr; medianQ1, Q3, 34.5 28.3,41.7/hr) and follow-up (34.6±19.3/hr; 33.3 18.4,46.2/hr). Lower baseline AHI and decreases in supine sleep predicted AHI response, while decreases in supine sleep and increases in REM sleep predicted AHI reductions (all P .05). Conclusions RRs alone do not accurately characterize therapeutic efficacy without adjustment for relevant covariates. Failure to account for influential covariates, including baseline sleep apnea severity and night-to-night differences in sleep stage and body position, significantly limits inferences drawn from previously uncontrolled single-arm HGNS cohort studies. Assessing continuous outcomes in actively treated and control participants with covariate adjustments in the THN3 and OSPREY RCTs more accurately characterizes HGNS therapeutic efficacy than uncontrolled treatment trials that solely rely on dichotomous RR outcomes. This abstract is funded by: LivaNova PLC