The Effect of Continuous Positive Airway Pressure on Clinical Outcomes in Stroke Patients With Obstructive Sleep Apnea: A Systematic Review and Meta-Analysis

Search Strategy and Study Selection

The literature search was conducted using three electronic databases: PubMed, ScienceDirect, and Taylor & Francis. Access to other major databases such as Embase and the Cochrane Library was not available at the time of the search. Therefore, these databases were not included in the search strategy. Search keywords were adapted using both free-text terms and Medical Subject Headings (MeSH), encompassing population-related terms such as “Sleep Apnea, Obstructive” and “Stroke,” as well as the intervention term “Continuous Positive Airway Pressure.” Boolean operators “OR” and “AND” were applied to combine search terms appropriately, as presented in Table 1. Articles were independently screened by three reviewers, and duplicate records were removed. Titles and abstracts were first reviewed for relevance and availability of full texts. Studies that did not meet the inclusion criteria were excluded. The full texts of the remaining studies were then assessed for eligibility. Studies were included if they met the inclusion criteria: 1) studies involving adult patients diagnosed with stroke and OSA; 2) studies comparing CPAP intervention groups with control groups (no CPAP or sham CPAP); 3) studies using a randomized controlled trial (RCT) design; and 4) full-text articles available. Studies were excluded if they met any of the following exclusion criteria: 1) nonresearch sources such as book chapters, reports, conference papers, or review articles; 2) publications in the form of editorials, commentaries, literature reviews, case reports, systematic reviews, or meta-analyses; and 3) animal studies.

Table 1.

Search strategies

Data Extraction and Quality Assessment

Studies that met the inclusion criteria were further reviewed for data extraction. Data from each study were collected and summarized in tabular form. Extracted information included the authors’ names, year of publication, country of origin, and details of the intervention and control groups. In addition, data on sample size, participant characteristics such as age, sex, and body mass index, stroke type, timing of CPAP initiation after stroke onset, inclusion criteria based on the apnea-hypopnea index (AHI), duration of intervention, mean nightly CPAP use, and follow-up period were recorded. The included studies were then further assessed for methodological quality using the Cochrane Risk of Bias Tool for Randomized Trials, version 2 (RoB 2).

Outcomes

The primary outcomes included CPAP adherence (hours per night), treatment dropout rates, functional status assessed using the Barthel Index (BI), Modified Rankin Scale (mRS), Functional Independence Measure (FIM), and Utrecht Scale for Evaluation of Rehabilitation (USER), as well as neurological function evaluated with the National Institutes of Health Stroke Scale (NIHSS) and Canadian Neurological Scale (CNS). The secondary outcomes consisted of cognitive function measured by the Mini-Mental State Examination (MMSE), functional balance assessed with the Berg Balance Scale (BBS), depression levels evaluated using the Beck Depression Inventory (BDI) and the Hospital Anxiety and Depression Scale–Depression subscale (HADS-D), cardiovascular events and mortality, reduction in AHI, and daytime sleepiness measured by the Epworth Sleepiness Scale (ESS) and the Stanford Sleepiness Scale (SSS).

Data Synthesis and Statistical Analysis

Data were analyzed using Review Manager (RevMan) version 5.4 (Cochrane) and R version 4.5.1 (R Core Team) through RStudio IDE 2025.09.0+387 (Posit) with the meta package. For dichotomous data, the Mantel–Haenszel method was used to calculate the pooled risk ratio (RR) with corresponding 95% confidence intervals (CIs). For continuous data, analyses were performed using either mean difference (MD) or standardized MD (SMD), depending on the consistency of measurement units across studies.

When studies reported outcomes at multiple follow-up time points, the shortest follow-up duration was used to maintain consistency of outcome timing across studies. Between-study heterogeneity was assessed using the I² statistic, with I² values greater than 50% indicating substantial heterogeneity. A fixed-effect model was applied when heterogeneity was low, while a random-effects model was used when heterogeneity was high. A p-value of less than 0.05 was considered statistically significant.

Search Results and Quality Assessment

As of January 24, 2025, a literature search across PubMed, ScienceDirect, and Taylor & Francis yielded 1,485 records. After removing 34 duplicates, 1,451 studies were screened, and 1,442 were excluded for not meeting the eligibility criteria. Of the 49 full-text studies assessed, 28 did not meet the population, intervention, control, and outcome criteria, 11 had inappropriate study designs, and six had incomplete data. Ultimately, six studies met all inclusion criteria and were included in the meta-analysis, as shown in Fig. 1.

Fig. 1.

Literature search flowchart. PICO, population, intervention, control, and outcome.

A total of six RCTs consisting Ryan et al. [16], Parra et al. [17], Aaronson et al. [18], Gupta et al. [19], Kim et al. [20], and Denis et al. [21] were included in the systematic review and metaanalysis. Two studies (33.3%) were categorized as having a high overall risk of bias, while four studies (66.7%) demonstrated a moderate risk of bias. The highest risk of bias was observed in the domains of missing outcome data, deviations from intended interventions, and the randomization process. Detailed results of the risk of bias assessment are presented in Figs. 2 and 3.

Fig. 2.

Risk of bias assessment diagram.

Fig. 3.

Summary of risk of bias assessment.

Study Characteristics

From the six RCTs [16-21] that met the inclusion criteria, a total of 349 participants were included, comprising 180 in the CPAP intervention group and 169 in the control group. Detailed study characteristics are presented in Table 2, while Tables 3 and 4 summarize the outcomes in the CPAP and control groups.

Table 2.

Study characteristics

Table 3.

Comparison of dropout rates and cardiovascular outcomes and mortality between the CPAP and control groups

Table 4.

Comparison of functional status, neurological function, cognitive function, functional balance, depression, AHI reduction, and sleepiness between CPAP and control groups

Primary Outcomes

CPAP adherence

Based on the six studies analyzed, four studies [16,17,19,20] reported an average CPAP usage of more than 4 hours per night, one study [18] reported usage of less than 4 hours per night, and one study [21] did not report the duration of CPAP use, as shown in Table 2. A pooled analysis of four studies [16-19] showed that the average CPAP usage was 4.35 hours per night (95% CI: 3.41 to 5.29). However, the heterogeneity test indicated an I² value of 86.5% with p<0.0001, suggesting a very high level of heterogeneity among the studies (Fig. 4).

Fig. 4.

Forest plot of CPAP use in the CPAP and control groups. SD, standard deviation; CI, confidence interval; CPAP, continuous positive airway pressure.

Dropout rate

Across six studies [16-21], 28 participants (15.56%) in the CPAP group discontinued therapy, compared with 3 participants (1.78%) in the control group (Table 3). Meta-analysis showed that the CPAP group had a significantly higher risk of discontinuation (RR=5.41; 95% CI: 2.19 to 13.39; p=0.0003) and low heterogeneity (I²=6%), as shown in Fig. 5.

Fig. 5.

Forest plot of dropout rate. CI, confidence interval.

Functional status

Fig. 6 presents the meta-analysis of functional status assessed using the BI, mRS, FIM, and USER self-care scales. The pooled results from five studies [16-20] showed no significant difference between the CPAP and control groups (SMD=0.46; 95% CI: -0.08 to 1.00; p=0.10), with substantial heterogeneity (I²=80%), although a trend toward improved function in the CPAP group was observed. Fig. 7 shows the meta-analysis of functional status measured with the BI, based on three studies [17,19,20], which also revealed no significant difference between groups (MD=1.23; 95% CI: -0.67 to 3.13; p=0.21), with low heterogeneity (I²=0%), though BI scores tended to be higher in the CPAP group. Fig. 8 summarizes the meta-analysis of functional status assessed using the mRS across two studies [17,20]. The results were not statistically significant (MD=0.97; 95% CI: -0.10 to 2.05; p=0.08), with very high heterogeneity (I²=96%), but showed a tendency toward functional improvement with CPAP.

Fig. 6.

Forest plot of functional status assessment using the Barthel Index, Modified Rankin Scale, Functional Independence Measure, and Utrecht Scale for Evaluation of Rehabilitation Self-Care scales. SD, standard deviation; CI, confidence interval.

Fig. 7.

Forest plot of functional status assessment using the Barthel Index scale. SD, standard deviation; CI, confidence interval.

Fig. 8.

Forest plot of functional status assessment using the Modified Rankin Scale. SD, standard deviation; CI, confidence interval.

Neurological function

The meta-analysis of four studies [16-18,20] assessing neurological function in stroke patients with OSA showed no statistically significant effect of CPAP compared with controls, although a trend toward improvement was observed (SMD=1.18; 95% CI: -0.35 to 2.70; p=0.13; I²=96%) (Fig. 9). Kim et al. [20] and Aaronson et al. [18] used the NIHSS, while Parra et al. [17] and Ryan et al. [16] used the CNS scale. Fig. 10 shows the NIHSS subgroup analysis of two studies [18,20], which found no significant difference between groups (MD=0.55; 95% CI: -0.24 to 1.35; p=0.17; I²=0%). The CNS subgroup analysis of three studies [16-18] also showed no significant effect of CPAP (MD=0.14; 95% CI: -1.45 to 1.73; p=0.87), with very high heterogeneity (I²=99%) (Fig. 11).

Fig. 9.

Forest plot of neurological function assessment using the National Institutes of Health Stroke Scale and Canadian Neurological Scale. SD, standard deviation; CI, confidence interval.

Fig. 10.

Forest plot of neurological function assessment using the National Institutes of Health Stroke Scale. SD, standard deviation; CI, confidence interval.

Fig. 11.

Forest plot of neurological function assessment using the Canadian Neurological Scale. SD, standard deviation; CI, confidence interval.

Secondary Outcomes

Cognitive function

Fig. 12 presents the meta-analysis results from two studies, Gupta et al. [19] and Kim et al. [20], which evaluated cognitive function in stroke patients with OSA using the MMSE scale. The analysis shows no statistically significant difference between the CPAP and control groups (MD=0.43; 95% CI: -2.35 to 3.21; p=0.76). The level of heterogeneity is high (I²=78%), indicating substantial variation among the included studies.

Fig. 12.

Forest plot of cognitive function assessment. SD, standard deviation; CI, confidence interval.

Functional balance

Fig. 13 presents the meta-analysis results on functional balance assessed using the BBS. Two studies were analyzed [16,20], with a total of 42 patients in both the CPAP and control groups. The results show no significant difference between the two groups, with an MD of -1.27 (95% CI: -4.16 to 1.62; p=0.39). The heterogeneity between studies was low (I²=0%).

Fig. 13.

Forest plot of functional balance assessment. SD, standard deviation; CI, confidence interval.

Depression

Based on Fig. 14, a meta-analysis of two RCTs [16,18] was conducted to evaluate the effect of CPAP therapy on depressive symptoms in stroke patients with OSA. Aaronson et al. [18] assessed depressive symptoms using the HADS-D, while Ryan et al. [16] used the BDI. The analysis showed no statistically significant difference between the CPAP and control groups in reducing depressive symptoms, with an SMD of -0.19 (95% CI: -0.63 to 0.25; p=0.41), and low heterogeneity between studies (I²=19%).

Fig. 14.

Forest plot of depression level assessment. SD, standard deviation; CI, confidence interval.

Cardiovascular events and mortality

The two studies [17,19] included in the meta-analysis showed that 8 participants (8/87; 9.15%) in the intervention group discontinued therapy, compared with 14 participants (14/109; 12.84%) in the control group (Table 3). Using a random-effects model, the meta-analysis yielded a RR of 0.63 (95% CI: 0.15 to 2.75; p=0.54), indicating no significant difference between the CPAP and control groups. Moderate heterogeneity was observed (I²= 47%), suggesting variability between studies (Fig. 15).

Fig. 15.

Forest plot of cardiovascular events and mortality. CI, confidence interval.

Reduction of AHI

The meta-analysis of AHI reduction was based on two studies [16,20] shown in Fig. 16. The results showed that CPAP significantly reduced AHI compared with the control group (MD= -21.31; 95% CI: -33.14 to -9.47; p=0.0004). Due to the high level of heterogeneity between studies (I²=88%), a random-effects model was applied.

Fig. 16.

Forest plot of apnea-hypopnea index reduction. SD, standard deviation; CI, confidence interval.

Sleepiness

Fig. 17 presents the meta-analysis of four studies evaluating daytime sleepiness in stroke patients with OSA using the ESS and SSS scales. Aaronson et al. [18] and Gupta et al. [19] used the ESS, whereas Kim et al. [20] and Ryan et al. [16] used the SSS. The analysis, conducted using the SMD with a random-effects model, showed that CPAP significantly reduced sleepiness compared with the control group (SMD=-1.43; 95% CI: -2.76 to -0.10; p=0.04). However, the high heterogeneity (I²=93%) indicates substantial variability across studies.

Fig. 17.

Forest plot of sleepiness assessment using Epworth Sleepiness Scale and Stanford Sleepiness Scale. SD, standard deviation; CI, confidence interval.

Daytime sleepiness assessed with the ESS was reported in three studies [16,19,20]. A fixed-effect meta-analysis demonstrated that CPAP significantly reduced sleepiness scores compared with the control group (MD=-2.60; 95% CI: -2.93 to -2.27; p<0.00001), with low heterogeneity (I²=0%), indicating consistent findings among the studies (Fig. 18). Sleepiness assessed using the SSS was analyzed in two studies [16,18] as shown in Fig. 19. The fixed-effect model showed a statistically significant reduction in sleepiness in the CPAP group (MD=-0.38; 95% CI: -0.62 to -0.14; p=0.002), with no heterogeneity observed (I²=0%).

Fig. 18.

Forest plot of sleepiness assessment using Epworth Sleepiness Scale. SD, standard deviation; CI, confidence interval.

Fig. 19.

Forest plot of sleepiness assessment using Stanford Sleepiness Scale. SD, standard deviation; CI, confidence interval.

Effect of CPAP in Clinical Outcomes of Stroke Patient with OSA

Combined analysis of four studies [16-19] showed that the average CPAP use was 4.35 hours per night. Additionally, Kim et al. [20] reported adherence of more than 4 hours per night, whereas Denis et al. [21] did not clearly specify the duration of CPAP use. Thus, most studies recorded CPAP use meeting or exceeding the minimum threshold of 4 hours per night. These findings are consistent with the literature showing that CPAP use ≥4 hours per night is associated with reduced risk of recurrent major adverse cardiac or cerebrovascular events [22]. Nevertheless, the considerable variability across studies requires caution in interpreting these results.

Six studies included in the meta-analysis [16-21] reported that 28 participants (28/180; 15.56%) in the intervention group discontinued therapy, compared with 3 participants (3/169; 1.78%) in the control group. The meta-analysis indicated that patients receiving CPAP had a significantly higher risk of treatment discontinuation than those in the control group. This trend was consistent across studies, with varied reasons for discontinuation, such as mechanical discomfort, cognitive impairment, or difficulty using the device. Consistent with these findings, Pavwoski and Shelgikar [23] noted that although CPAP is beneficial, adherence is often hindered by factors such as mask discomfort, claustrophobia, pressure intolerance, and lifestyle or social considerations.

Combined analysis of several studies showed that CPAP use in stroke patients with OSA did not result in statistically significant improvements in neurological recovery. Sub-analyses based on the type of scale used (NIHSS and CNS) showed similar results, with improvement trends that did not reach statistical significance. High heterogeneity was observed in analyses using the CNS scale, whereas no heterogeneity was found for the NIHSS. Regarding functional status, neither the overall analysis nor the sub-analyses based on BI and mRS showed significant differences between CPAP and control groups, although the direction of improvement generally favored the intervention group. High heterogeneity was observed in the mRS analysis, but not in the BI.

In individual analyses, Parra et al. [17] demonstrated significant early benefits of CPAP during the first month of intervention, particularly on the CNS and mRS scales, although these effects did not persist in long-term follow-up. One explanation proposed is the ceiling effect, in which some patients reach the maximum score at baseline or early after intervention, preventing detection of further improvements. Aaronson et al. [18] reported that this occurred in approximately 29%–68% of patients depending on the scale used.

Limitations in the sensitivity of measurement instruments likely influenced these outcomes. The BI is known to be less sensitive in detecting functional changes in patients with mild to moderate stroke due to both floor and ceiling effects [24]. NIHSS has also been reported to demonstrate similar ceiling effects, either because certain items cannot be assessed in patients with very severe stroke and are thus assigned maximum scores [24,25] or because some patients achieve maximum scores during follow-up, making further improvements difficult to detect [26].

The relatively short duration of therapy in several studies may also have influenced the results. Ballester et al. [27] noted that sensitivity to post-stroke therapy decreases gradually but may persist for more than 12 months. This suggests a long-term critical period with enhanced neuroplasticity that allows improvements in body function and structure even during the chronic phase. Additionally, adherence to CPAP has been associated with reductions in insulin resistance without further increases in brain-derived neurotrophic factor, which plays an important role in regulating synaptic plasticity and insulin sensitivity [28]. Therefore, longer-duration CPAP interventions during the chronic phase may still provide meaningful benefit. Consistent with this, Gupta et al. [19] reported that the proportion of patients showing more than a one-point improvement in mRS at 6 and 12 months was significantly higher in the CPAP group than in the control group.

Post-stroke recovery profiles vary widely. Therefore, classification based on stroke severity, location, and type is more appropriate to capture the complex and non-linear nature of recovery. Patients with mild initial deficits generally demonstrate better recovery compared to those with more severe deficits [29]. Thus, future studies would benefit from classifying populations according to stroke type, location, and severity.

Regarding functional balance, the analysis showed that CPAP did not result in significant improvements compared with control. Both Ryan et al. [16] and Kim et al. [20] reported improvements in balance scores within each group after intervention, but the between-group differences were not statistically significant. A possible explanation is the relatively short treatment duration, which may not have been sufficient to produce meaningful changes. Additionally, the instrument used, the BBS, may not be sensitive enough to detect CPAP-induced changes. Although the BBS is known for its high validity and reliability as a balance assessment tool [30], it may be less sensitive to changes produced by interventions like CPAP. This is attributable to balance being influenced by multiple factors depending on an individual’s condition, including muscle weakness, impaired motor coordination, cognitive deficits, and sensory disintegration, all of which may be affected by neurological impairment [31].

The meta-analysis of two studies assessing cognitive function showed no significant difference between CPAP and control groups. Individually, Gupta et al. [19] found no significant differences, whereas Kim et al. [20] reported significant improvement in cognitive function in the CPAP group compared with the control. This meta-analysis demonstrated high heterogeneity. Differences in timing of CPAP initiation may account for this. In Gupta et al. [19], CPAP was initiated at least 6 weeks after stroke, whereas in Kim et al. [20], intervention began much earlier on average 4.6±2.8 days post-stroke. These findings are consistent with the meta-analysis by Yang et al. [32], which showed that early initiation of CPAP provides significant improvement in global cognitive function in stroke patients with OSA.

Depressive symptoms also did not show significant differences between CPAP and control groups in the meta-analysis of two studies. Although both groups experienced slight reductions in depression scores after intervention, the between-group differences were not statistically significant. This finding was consistent across studies despite the use of different scales (BDI and HADS-D). Individually, Ryan et al. [16] reported significant improvement in the affective component of the BDI score in the CPAP group (p=0.006). Ryan et al. [16] suggested that improvements were mainly seen in affective rather than somatic components, as somatic components are influenced by physiological effects of stroke, such as loss of appetite. These findings suggest that the effectiveness of CPAP for depressive symptoms likely depends on the dimension assessed and the sensitivity of the measurement instrument, indicating that CPAP is more appropriate as an adjunct therapy rather than a primary treatment for affective disorders in stroke patients with OSA.

The meta-analysis of cardiovascular events and mortality did not show significant effects of CPAP. Although several studies reported trends toward improvement, such as lower recurrence of vascular events or improved cardiovascular survival in the CPAP group, these differences did not reach statistical significance. These findings align with a meta-analysis in stroke patients with sleep apnea, which reported nonsignificant differences, although recurrent vascular events were numerically lower in the CPAP group [33].

CPAP was found to be significantly effective in reducing AHI severity in stroke patients with OSA compared with controls. This supports evidence that CPAP improves sleep-related breathing. However, high heterogeneity was observed across studies, likely due to differences in study design, participant characteristics, intervention duration, and timing of therapy initiation.

Reduction in sleepiness showed a significant improvement favoring the CPAP group, with lower sleepiness scores compared with controls. To address the high heterogeneity in this analysis, sub-analyses were performed based on the type of scale used. Both ESS and SSS sub-analyses showed significant improvements in the CPAP group, with no heterogeneity detected. These findings align with the meta-analysis [32], which concluded that CPAP consistently reduces subjective sleepiness in stroke patients with OSA. Similarly, Suusgaard et al. [34] reported that high CPAP adherence in patients post-stroke or TIA with OSA was associated with significant reductions in sleepiness.

Considering the insufficient evidence demonstrating the benefit of CPAP on post-stroke clinical outcomes, as well as the relatively low adherence to CPAP therapy, alternative treatment modalities for OSA should be considered. One such alternative is the use of oral appliance therapy. A retrospective study found that oral appliance therapy lowered the AHI by ≥50% or to <10 events per hour in most post-stroke patients with OSA, suggesting its potential role as an alternative treatment to CPAP [35]. Other treatment approaches for OSA, including positional therapy, weight management, and surgical interventions, have also been shown to be beneficial in patients with OSA [36-38]. However, to date, there is still limited evidence specifically evaluating these interventions in stroke patients with OSA.

Limitation

This study has several limitations that should be considered when interpreting the results. The number of studies included in the analysis was relatively limited, with small sample sizes in each study, which may restrict statistical validity and reduce the generalizability of the findings. Additionally, variations in study design, timing of therapy initiation, intervention duration, outcome measurement instruments, and study population characteristics contributed to heterogeneity and inconsistency of results. Patient adherence to CPAP use was also reported to be low in several studies, potentially influencing the therapeutic effectiveness. Consistent with the risk-of-bias assessment, some studies demonstrated weaknesses in blinding procedures, loss of follow-up data, and deviations from intended interventions, which may have affected the internal validity of this meta-analysis.

Another limitation of this study is that the literature search was restricted to three databases. Other important databases, such as Embase and the Cochrane Library, were not included due to lack of access. However, the included databases cover a large proportion of biomedical literature.

Recommendations

1) Future studies should include a larger sample size, longer follow-up duration, earlier initiation of therapy, and study populations classified according to stroke type, location, and severity.

2) Subsequent research should be designed with more rigorous and standardized methodologies to minimize the risk of bias, along with the use of valid and sensitive measurement instruments to improve the accuracy of clinical change detection and the validity of research findings.

3) Further research is required to evaluate treatment modalities other than CPAP for the management of OSA in stroke patients.

Conclusion

Current evidence showed that CPAP therapy is insufficient to prove a benefit in stroke-related clinical outcomes, including neurological function, functional status, cognitive function, functional balance, depression level, cardiovascular events, and mortality. A significant effect of CPAP therapy was consistently observed in outcomes directly related to OSA, namely a reduction in AHI and improvement in daytime sleepiness. On the other hand, the high treatment dropout rate in the CPAP group indicates adherence issues, which may have affected the clinical effectiveness of the intervention.