Hypnotics on Obstructive Sleep Apnea Severity and Endotypes: A Systematic Review and Meta-Analysis

Abstract

Rationale: Low arousal threshold and poor muscle responsiveness are common determinants of obstructive sleep apnea (OSA). Hypnotics were hypothesized as an alternative OSA treatment via raising the arousal threshold and possibly genioglossus responsiveness. Objectives: To examine the effect of common hypnotics on arousal threshold, OSA severity, and genioglossus responsiveness. Methods: We searched MEDLINE, Embase, CENTRAL, and ClinicalTrials.gov for randomized clinical trials, and we ran meta-analyses to determine the effect of oral hypnotics on arousal threshold, OSA severity, and genioglossus responsiveness. The Grades of Recommendation Assessment, Development and Evaluation was used to rate the quality of evidence (QoE). The association between post-treatment apnea-hypopnea index (AHI) and arousal threshold percentage reductions was explored in individual patient data meta-analyses (overall sample and low arousal threshold subgroups). Measurements and Main Results: On the basis of our analysis (27 studies; 25 for AHI, 11 for arousal threshold, 4 for genioglossus responsiveness), hypnotics minimally raised arousal threshold (mean difference [95% confidence interval], 2.7 [1.5, 3.8] cm H₂O epiglottic pressure swings; moderate QoE) but did not change OSA severity (-1.4 [-3.5, 0.7] events/h; moderate QoE). Individual patient data meta-analysis (N = 114) showed no association between changes in arousal threshold and AHI, independent of arousal threshold subgrouping. However, people with very low arousal threshold or those who exhibited a 0-25% arousal threshold increase from placebo experienced the greatest, yet still modest, post-treatment AHI reductions (∼10%). Hypnotics did not affect genioglossus responsiveness (high QoE). Conclusions: Further research testing or clinical use of hypnotics as OSA alternative treatments should be discouraged, unless in the presence of comorbid insomnia or as part of combination therapy in individuals with very low arousal threshold.

Keywords: Z-drugs; arousal threshold; benzodiazepines; genioglossus responsiveness.

Figure 1.

Flowchart of our literature search. AHI = apnea–hypopnea index.

Figure 2.

Forest plot illustrating the mean effect of different hypnotics on arousal threshold, by drug class subgroups and in the overall study sample. One study tested different drugs in the same protocol (9); data are pooled per agent/dose. The mean difference denotes epiglottic pressure swings; however, data from a number of studies were converted from esophageal pressure (26, 38, 45) or diaphragm electromyography (25) (see text for further details). The risk of bias is represented on the right for each study. Green, yellow, and red correspond to low, unclear, and high risks of bias, respectively. Other bias includes potential carryover effect and potential selection bias in one study (20). CI = confidence interval; IV = inverse variance.

Figure 3.

Forest plot illustrating the mean effect of different hypnotics on obstructive sleep apnea severity in different subgroups separated by drug class and in the overall study sample. Note that several works studied more than one molecule or drug dose in the same protocol (28, 36, 39, 48); data are pooled per agent/dose. The risk of bias is represented on the right for each study. Green, yellow, and red correspond to low, unclear, and high risk of bias, respectively. Other bias includes potential carryover effect in one study that investigated two drugs (36) and in another study (20), as well as potential dose–response effect in one study because two dosages were administered in the same arm without further specification (28), and potential selection bias in three studies (20, 27, 28). CI = confidence interval; DORAs = dual orexin receptor antagonists; IV = inverse variance.

Figure 4.

Percentage changes in arousal threshold (positive values equal positive effect) and apnea–hypopnea index (AHI; negative values correspond to positive effect) binned per category of arousal threshold percentage change from placebo (green, <0%; orange, ≥0% and <25%; blue, ≥25%). Increasing arousal threshold seems to reduce obstructive sleep apnea severity up to a threshold (note the U-shaped relationship between changes in AHI and arousal threshold). AHI weighted percentage change from placebo was calculated as follows: (AHI_hypnotic − AHI_placebo)/(AHI_hypnotic + AHI_placebo). Bars indicate 95% confidence interval.

Figure 5.

Forest plot showing the mean effect of different hypnotics on genioglossus responsiveness. One study tested different drugs in the same protocol (9); data are pooled per agent/dose. The risk of bias is represented on the right for each study. Green and yellow correspond to low and unclear risk of bias, respectively. CI = confidence interval; IV = inverse variance.