The results of this study show that oral administration of 15 mg of suvorexant did not result in significant changes in AHI in patients with severe OSA. To the best of our knowledge, this is the first study to assess the effects of an orexin receptor antagonist on patients with severe OSA.

The recommended doses of suvorexant in the United States and Japan are 15 mg for elderly patients and 20 mg for adults. A previous study that examined the effects of a single dose of 40 mg of suvorexant on patients with mild to moderate OSA found no significant difference in mean SpO2 during the total sleep time, and concluded that 40 mg of suvorexant had no clinically significant sleep-disordered breathing effects on patients with mild to moderate OSA [24]. However, because the safety of suvorexant in patients with severe OSA has not been confirmed, in this study, we examined the use of the lowest clinically used dose of 15 mg in Japan.

We did not observe a significant difference between AHI on the first night and that after oral administration of 15 mg suvorexant on the second night (mean difference: 1.873, 95% CI: [− 2.241, 5.987], p = 0.359). Power calculations were performed with reference to a previous study [27]; accordingly, it was calculated that a minimum of 24 cases would be required and that therefore 30 cases would be sufficient for this study. The AHI is widely used for the diagnosis and severity classification of OSA [28] and has been used in many studies examining respiratory safety in patients with OSA [24,28]. As it has been noted that an increase in AHI of 5 events/h increases the risk of hypertension and type 2 diabetes [17,29],this difference is generally used as a clinically meaningful criterion in studies on the safety of sleep medication for patients with OSA [23,24,27]. Therefore, the non-inferiority margin was set at 5 in this study as well, but the upper limit of the 95% CI was found to be 5.98, which is greater than 5, and therefore non-inferiority could not be proven. Compared to the latest study [27] that examined the effect of orexin receptor antagonists on OSA, the standard deviation of AHI in the population of this study was 15.3, which was larger than that of AHI of 8.49 in the population of the previous study [27]. This may have contributed to the larger 95% CI.

Of the 30 patients with severe OSA, 22 (73%) had an AHI change of 5 events/h or more (increase in 12 cases, decrease in 10 cases), and eight (26%) had an AHI change of 10 events/h or more (increase in six cases, decrease in two cases). It is known that there is night-to-night variability in PSG results [30,31]. In particular, the AHI of patients with OSA has been reported to be highly variable across multiple PSG trials [31,32]. Since the AHI varied around 10 events/h in 32–49% of patients with OSA during two consecutive nights of PSG [33,34], the AHI change in this study may have been due to the night-to-night variability that can occur in PSG.

The 3% oxygen desaturation index and SpO₂ < 90% time are also important indicators of respiratory safety [28,35]. In the present study, the mean duration of apnea was significantly prolonged (mean difference: 1.967 s, 95% CI: [0.195, 3.74], p = 0.031). However, all the parameters related to oxygenation, such as the 3% oxygen desaturation index, SpO₂ < 90% time, and lowest SpO₂, showed no significant difference between the first and second nights, although the CI for the 3% oxygen desaturation index tended to be slightly higher (Table 2).

First-night effects on sleep have been reported to decrease total sleep time, increase mid-night awakenings, and disrupt sleep structure (increase N1 sleep, decrease REM sleep, and prolong REM latency) [31]. In this study, total sleep time was significantly prolonged, sleep efficiency increased, and mid-night awakenings were reduced on the second night compared to those on the first night. These results are consistent with the release of the first-night effect. However, although there was no significant difference in the percentage of N1, N2, and N3 sleep stages between the two nights, the increase in the percentage of REM sleep and the shortening of REM latency were thought to be influenced by suvorexant as well as the first-night effect. A previous study on suvorexant reported that total sleep time was prolonged and REM sleep increased after the first night of administration, but that there was no significant difference in the percentage of non-REM sleep [24], which is consistent with this study.

It is known that some OSA patients exhibit an increase in AHI during REM sleep (REM AHI is more than twice the NREM AHI) [36]. In this study, four patients were identified as having REM-dependent OSA on the first night of PSG. Among these four patients, three did not exhibit REM dependency on the second night, and no new patients exhibited REM dependency on the second night.

A paired t-test comparing the AHI on the first and second nights for these four patients showed no significant difference (mean difference: 11.2, 95% CI: [− 16.8, 39.25], p = 0.2922). Similarly, a paired t-test comparing REM-AHI between the two nights also showed no significant difference (mean difference: − 9.78, 95% CI: [− 25.93, 6.38], p = 0.1497). Therefore, our study did not find clear evidence that suvorexant administration increases AHI in patients with REM-dependent OSA. However, it is important to note that the sample size for REM-dependent cases was small, with only four patients on the first night. This may not be sufficient for statistical analysis.

Additionally, we examined the correlation between changes in AHI and changes in the percentage of REM sleep over the two nights. The correlation coefficient was − 0.172, with a 95% confidence interval of − 0.52 to 0.201 and a p-value of 0.363, indicating no significant correlation.

Linear regression analysis using the differences between values of each PSG parameter on the first and second nights as the objective variables and age and BMI as dependent variables showed no significant effect of age or BMI on either respiration-related or sleep-related parameters. Thus, age and BMI did not appear to have a significant effect on any of the parameters in this study population.

In summary, while this study found no significant changes in AHI and other respiration-related parameters while demonstrating some effects of suvorexant, such as increased REM sleep percentage and shortened REM latency, the upper limit of the 95% CI for the difference in AHI was 5.98, which is slightly more than 5, which did not prove non-inferiority. The changes in the other parameters can be interpreted as due to night-to-night variability.

Some limitations of this study must be noted. As this study was based on PSG testing on two different nights, night-to-night variability cannot be ruled out.　Moreover, as a crossover study is desirable when examining safety, the results should be interpreted with caution. The presence of confounding factors other than age and BMI also cannot be ruled out. Additionally, the study participants and technicians who assessed PSG were not blinded, due to which participant and observer bias could not be eliminated, which limits the external validity of the study.