Study selection was performed as described in Fig. 1. A total of 817 records were initially identified from the database search from inception to 9 July 2024, from which 36 full-text articles were retrieved and assessed for eligibility after removing duplicates and screening. Of these, 21 articles were excluded due to missing outcomes of interest (n = 4), not using CPAP (n = 6), different comparisons (n = 1), secondary analyses (n = 8), pooled analysis (n = 1), and open-label study (n = 1). One additional study was identified following an updated database search from 10 July 2024 to 9 January 2025. Thus, 14 RCTs (total N = 2969), reported in 16 publications, were ultimately considered eligible for qualitative and quantitative analyses.

All study participants were adults (age ≥ 18 years) who had OSA with residual sleepiness after CPAP treatment. Study characteristics are described in Table 1 [23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38]. Criteria for defining OSA, residual sleepiness, and effective CPAP therapy [use \(\ge\) 4 h per night, for at least 70% of nights] were similar between most trials. Participant demographics were generally similar across the trials included, i.e., mean age of 50 years, male predominance, and BMI of greater than 30. However, the Inoue study included Asian participants with lower BMI (27.57 versus \(\ge\) 32 kg/m2) than other studies [28]. The majority of studies predominantly included Caucasian participants, however, ethnicity was not reported for some studies. ESS scores at baseline were similar across studies. All of the trials had placebo as a comparator. The study by Herring et al. included two arms of WPAs, i.e., modafinil and mk-0249; the latter treatment is a compound that has not been clinically approved and so this arm was excluded from our analysis [29].

PRISMA flow chart for study selection of wake-promoting agents for the treatment of residual sleepiness in obstructive sleep apnea despite continuous positive airway pressure. *As two articles were follow-up analyses for other randomized controlled trials (RCTs), 14 RCTs were included in meta-analysis but only **13 RCTs included AE data that could be analyzed. AE, adverse event; CGI-C, clinical global impression of change; EQ-5D, Euro-QoL 5-dimension scale; ESS, Epworth sleepiness scale; FOSQ, functional outcomes of sleep questionnaire; MWT, maintenance of wakefulness test; SF-36, 36-item short-form health survey

Overall, six, four, two, and two RCTs compared modafinil, armodafinil, solriamfetol, and pitolisant with placebo, respectively. Modafinil dosages were 200 mg/day (two RCTs) [28, 29], 300 mg/day (one RCT) [27], and 400 mg/day (three RCTs) [23,24,25,26]. Armodafinil dosages were 150 mg/day (one RCT) [31], 200 mg/day (one RCT) [33], and 150 and 250 mg/day (two RCTs) [30, 32]. Solriamfetol dosages varied from 37.5 to 300 mg/days (two RCTs) [34,35,36] whereas pitolisant dosages varied from 5 to 40 mg/day (two RCTs) [37, 38]. Most RCTs (11 RCTs) [23,24,25,26,27,28, 30,31,32,33, 35,36,37,38] were parallel-arm designs, while 2 were crossover designs [29], and another was a withdrawal design [34]. All RCTs evaluated outcomes at 2–12 weeks, which was categorized for further analyses as < 4 weeks and 4–12 weeks (Table 1).

All studies showed low risk of bias. The funnel plots showed no evidence of publication bias (Supplementary Appendix 2-Fig. 1, Appendix 3-Fig. 1, Appendix 4-Figs. 1 and 2, Appendix 5-Figs. 1–7, and Appendix 6-Fig. 1).

All 14 RCTs [23, 24, 26,27,28,29,30,31,32,33,34,35, 37, 38] assessed the effects of the four WPAs on ESS scores; 6 [23, 24, 26,27,28,29] and 4 [30,31,32,33] of these RCTs compared modafinil and armodafinil with placebo. A DMA suggested that modafinil significantly reduced ESS score (compared with placebo) within 4 weeks by −2.44 points (95% CI −3.61, −1.27) (Supplementary Appendix 2-Table 1); likewise, armodafinil also significantly reduced ESS scores by −2.38 points (−3.21, −1.55) at \(\le\) 4 weeks and by −2.88 points (−3.85, −1.91) at 4–12 weeks after treatment.

A NMA of ESS measured at \(\le\) 4 weeks included four treatments (N = 2,634) [23, 24, 26,27,28,29,30,31,32,33,34,35, 37, 38] (Fig. 2a). Compared with placebo, ESS was significantly reduced by solriamfetol, modafinil, and armodafinil, but not pitolisant, with pooled MDs of −3.84 (−5.60, −2.07), −2.44 (−3.38, −1.49), −2.41 (−3.60 to −1.21), and −0.86 (−2.36, 0.63), respectively (Table 2). Comparing active WPAs, only solriamfetol significantly reduced ESSs when compared to pitolisant with a pooled MD of −2.98 (−5.29, −0.66).

Network of eligible comparisons for efficacy at \(\le\) 4 weeks and at >4-12 weeks and safety at study endpoint. Efficacy (ESS, MWT, and CGI-C) was analyzed at \(\le\) 4 weeks and at >4-12 weeks. Duration in CPAP use and patients experiencing serious adverse events, adverse events leading to discontinuation, and specified adverse events were analyzed at study endpoint. Size of nodes was weighted by number of subjects. Size of edges was weighted by number of studies in each comparison. CGI-C, clinical global impression of change; CPAP, continuous positive airway pressure; ESS, Epworth sleepiness scale; MWT, maintenance of wakefulness test

A network map of the four WPAs on ESS at 4–12 weeks was constructed (N = 1935) [26, 30,31,32, 35, 37, 38] (Fig. 2d). ESS scores at 4–12 weeks were significantly reduced with pooled MDs (95% CI) of −4.11 (−6.14, −2.08), −2.88 (−3.85, −1.91), −2.70 (−3.66, −1.73), and −2.46 (−3.68, −1.24) for solriamfetol, armodafinil, pitolisant, and modafinil relative to placebo, respectively (Table 2).

Five [23, 26,27,28,29] and three [30,31,32] RCTs compared the effect of modafinil and armodafinil relative to placebo on MWT (Supplementary Appendix 3-Table 1). A DMA indicated both treatments significantly increased MWT within 4 weeks with corresponding pooled MDs (95% CI) of 3.62 (2.48, 4.76) and 2.52 (1.27, 3.76).

A NMA was performed using data from 10 RCTs [23, 26,27,28,29,30,31,32, 34, 35]; 3 treatments including solriamfetol, modafinil, and armodafinil were included for assessing MWT \(\le\) 4 weeks (N = 1764) [23,