Obstructive sleep apnoea and glaucoma: a systematic review and meta-analysis

Abstract

Introduction

Obstructive sleep apnoea (OSA) has been thought to be associated with glaucoma, however there are many conflicting studies on this topic. With many new studies having been published since the previous meta-analysis, we believe it is important to clarify this association. Hence, in this study we meta-analyse the recent literature regarding the association between OSA and glaucoma.

Methods

Pubmed, Embase, Scopus and Cochrane Library were searched from inception till the 28th February 2022 for observational as well as cross-sectional studies examining the association between OSA and glaucoma. Two reviewers selected studies, extracted data, graded the quality of included non-randomized studies using the Newcastle-Ottawa scale. The overall quality of evidence was assessed using GRADE. Random-effects models were used to meta-analyse the maximally covariate- adjusted associations.

Results

48 studies were included in our systematic review, with 46 suitable for meta-analysis. Total study population was 4,566,984 patients. OSA was associated with a higher risk of glaucoma (OR 3.66, 95% CI 1.70 to 7.90, I² = 98%, p < 0.01). After adjustment for various important confounders including age, gender and patient comorbidities such as hyperlipidaemia, hypertension, cardiovascular diseases and diabetes, patients with OSA had up to 40% higher odds of glaucoma. Substantial heterogeneity was eliminated through subgroup and sensitivity analyses after consideration of glaucoma subtype, OSA severity and adjustment for confounders.

Conclusions

In this meta-analysis, OSA was associated with higher risk of glaucoma, as well as more severe ocular findings characteristic of the glaucomatous disease process. We suggest more clinical studies looking into the effects of OSA treatment on the progression of glaucoma to help clinical decision making for patients.

Introduction

Glaucoma, the leading cause of blindness worldwide [1], has been acknowledged as multifactorial with vascular, genetic anatomical and immune contributing factors. Currently, it is acknowledged that intraocular pressure (IOP) is the only modifiable risk factor for development and progression of glaucoma. However, studies have shown that one-third to half of all patients with glaucoma may have a normal IOP [2]. Thus, it is important to identify other potential risk factors which may aid early detection and management of patients at risk of glaucoma.

Obstructive sleep apnoea (OSA) is a highly prevalent disease burdening close to one billion people worldwide [3]. In recent years, there have been acknowledgement that OSA is associated with incidence of glaucoma. One predominant theory is that OSA contributes to hypoxic and vascular changes leading to damage to the optic nerve head as well as retinal ganglion cells and their axons, thereby leading to glaucoma [4]. Another theory that has been proposed involves nocturnal IOP relating to a supine position and raised episcleral venous flow [5].

Previous meta-analyses suggest that the evidence available supports an association between OSA and glaucoma. Liu et al. [6] and Shi et al. [7] showed relatively similar results, with a strong positive association between the two conditions. On the other hand, there have been studies previously conducted that suggest a low level of evidence. For example, Girkin et al. [8], a nested case-control study which was included in both abovementioned meta-analyses, showed that the overall evidence for the association was low as no association was found after adjusting for confounders including patients’ metabolic and cardiovascular comorbidities. Similarly, Roberts et al. [9] found no significant association between the oxygen desaturation index (ODI) and glaucoma incidence.

Hence, it is important to interpret the findings of these meta-analyses with due consideration of their limitations. Few studies were included in both meta-analyses – Liu et al. had a total of 3 studies in each of its two analyses while Shi et al. studied 6 cohort studies in their meta-analysis. Neither analysis pooled data from studies which had been adjusted for important confounders. In addition, subgroup analysis studying the effect of OSA severity on glaucoma was not performed. While Liu et al. noted significant heterogeneity in studies containing a mix of glaucoma subtypes, no comparison was made between the varying subtypes.

Furthermore, it may seem intuitive that treatment of OSA results in improvement of glaucoma if the two conditions are truly associated. However, some studies such as Alvarez-Sala et al. [10] suggest that treatment with continuous positive airway pressure conversely raises the IOP.

With many new studies having been published since the previous meta-analysis, we believe it is important to clarify this association. Recent studies have also included adjusted data in their analysis which allows us to evaluate the effects of OSA on glaucoma incidence after adjustment of important confounders. Hence, in this study we aim to perform an updated meta-analysis regarding the association between OSA and glaucoma.

Methods

Our meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [11]. The PRISMA checklist for this review is reported in Supplemental Table 1. This review has been previously registered on PROSPERO (CRD42022333879).

Search strategy

Pubmed, Embase, Scopus and Cochrane Library were searched from inception till the 28^th February 2022. Our free text search strategy was: (“sleep apnea” OR “nocturnal hypoxia” OR “nocturnal hypoxaemia” OR “OSA” OR “obstructive sleep apnea” OR “obstructive sleep apnea syndrome” OR “syndrome, obstructive sleep apnea” OR “obstructive sleep apneas” OR “sleep apnea, obstructive”) AND (“optic neuropathy” OR “ischemic optic neuropathies” OR “anterior optic neuritides” OR “optic neuritis” OR “eye disease” OR “eye diseases” OR “glaucoma” OR “retinal artery” OR “retinal vein” OR “retinopathy” OR “optic” OR “ocular”) AND (“trial” OR “cohort” OR “case control” OR “observational” OR “longitudinal” OR “study” OR “cross sectional”). In addition, bibliographies of reviews and included studies were hand-searched for relevant data, however, no additional relevant records were obtained. As some of the studies may have reported glaucoma incidence as well as relevant ophthalmic parameters within their results as either baseline parameters or secondary outcomes, our initial search strategy was intentionally not limited to glaucoma alone. All potentially relevant full-text articles relating OSA with ophthalmic conditions were assessed for relevant data.

Study selection

Relevant articles were independently reviewed by SKXW and CYW, using the data management software Rayyan QCRI [12]. Titles and abstracts were used for the first stage of study selection; subsequently full texts were screened for final inclusion. Observational studies, cross-sectional studies as well as any randomized controlled trials were included; case reports, case series, reviews, letters and non-English publications without available translations were excluded. Study populations included were individuals 18 years and older, with diagnoses of OSA, compared against a control group without OSA. Outcomes included glaucoma rates in OSA populations and populations without OSA, as well as various ocular parameters including the IOP, retinal nerve fibre layer (RNFL) thickness, visual field mean deviation (MD) and visual field pattern standard deviation (PSD), and cup-disc ratios. Prior diagnosis of glaucoma was not a prerequisite for inclusion. The presence and severity of OSA was determined using the apnoea-hypopnea index (AHI), respiratory disturbance index (RDI), or clinical diagnoses of OSA such as the International Classification of Diseases (ICD) codes versions 9 and 10.

Subgroup analyses were performed based on glaucoma subtypes reported. The main subtypes reported were primary open angle glaucoma and normal tension glaucoma. Only one study, Friedlander 2018, reported on open angle glaucoma suspects. Where data provided was robust, further subgroup analyses based on OSA severity as quantified by the AHI classification were performed. Patients with AHI of 5–15 were considered to have mild OSA; 15–30 was moderate OSA and patients with AHI > 30 were considered to have severe OSA.

Quality of included studies

No randomized-controlled trials met the inclusion criteria – all our included studies were observational. The quality of each individual study was evaluated with the Newcastle Ottawa Scale (NOS), and is reported in Supplemental Table 2 [13]. Studies with three stars or less were of low quality, studies with 4-6 stars were of moderate quality and studies with 7 or more stars were identified as high quality. Two studies [14, 15] included in our systematic review had low quality and their data was excluded from the meta-analysis; the 47 studies in our meta-analysis were of moderate or high quality.

Data extraction

SKXW and CYW extracted the following data from each data into a standardized extraction template: title and year of publication, first author, study design, country of study, sample sizes of various study groups, percentage of males, mean/median age, relevant interventions, exposures and outcomes, as well as key findings.

Statistical analysis

Our meta-analysis looked at the association between baseline OSA and various measures of glaucoma. This included glaucoma rates in OSA populations and populations without OSA, as well as various ocular parameters including the IOP, retinal nerve fibre layer (RNFL) thickness, visual field mean deviation (MD) and visual field pattern standard deviation (PSD), and cup-disc ratios.

Generic inverse variance method was used for our analysis. We favoured maximally covariate-adjusted estimates were available. Random-effects models were used to account for heterogeneity in the observational estimates [16]. Between-study heterogeneity is reported as the I² value in our analysis. An I² of <30% was treated as indicative of low heterogeneity between studies; 30–60% as indicative of moderate heterogeneity and >60% as indicative of substantial heterogeneity [17]. Knapp-Hartung adjustments [18] were used in the calculation of confidence intervals around the pooled effect. Pre-specified subgroup analyses were performed to limit heterogeneity and better characterize the relationship between OSA and glaucoma. These included the severity of OSA and glaucoma subtype, where reported. Further sensitivity analyses were performed in some subgroups to limit included studies to non-glaucomatous eyes. In studies where data of individuals with glaucoma and individuals without glaucoma were reported separately, we only analysed data from individuals without glaucoma to reduce residual confounding. Where treatment data was present comparing pre-treatment eyes and post-treatment eyes, values before treatment were used for analysis. For studies which reported ocular parameters by left and right eyes or by OSA severity, we used the mean value reported with its standard deviation where available. Otherwise, a fixed effects model was utilized for each study to calculate the overall effect size. We used R (version 4.1.1) for our analysis, complying with the statistical approaches recommended in the Cochrane handbook, and considered a two-sided P value of less than 0.05 to be statistically significant.

Results

Our study selection process is shown in Supplemental Fig. 1. A search of the abovementioned four database retrieved a total of 1755 results, of which 207 duplicates were removed. Title and abstract screening excluded 1326 studies, and full text screening excluded a further 176 studies. A total of 46 studies were included in the meta-analysis.

Baseline characteristics of study population

The baseline characteristics of the included studies are reported in Table 1. The 48 studies had a combined study population of 3,316,034 patients. Kadyan 2010 [19] diagnosed OSA via an Oxygen Desaturation Index (ODI) of more than 30 on overnight oximetry; Chen 2014 [20] and Keenan 2016 [21] used the version 9 and 10 International Classification of Diseases (ICD) codes respectively; the remaining 46 studies diagnosed OSA via polysomnography. Among the studies, 15 were prospective [14, 15, 22–34], 5 were retrospective [20, 21, 35–37] and the remaining 28 were cross-sectional studies [5, 19, 38–63]. The total sample size of each study as well as the number of OSA patients within each study are also reported in Table 1. Relevant interventions, exposures, outcomes, and key findings of each study are reported in Table 2.

Table 1.

Characteristics of included studies.

First author and year published	Reference no. in manuscript	Study design	Country of study	Diagnosis of OSA	Mean age (years)	Male (%)	Sample size	OSA population size	Newcastle Ottawa Scale (NOS)
Aptel 2014	22	Prospective cohort	France	PSG	63.25	68	6933	6754	7
Asker 2013	24	Prospective case-control	Turkey	PSG	50.86	72	82	41	6
Bagabas 2019	38	Cross-sectional	Saudi Arabia	PSG	44.08	66	86	45	7
Batum 2020	23	Prospective cohort	Turkey	PSG	47.9	62	60	30	6
Carnero 2020	42	Cross-sectional	USA	PSG	53.6	80	36	18	6
Casas 2013	25	Prospective case-control	Spain	PSG	50.18	72	100	50	6
Casas 2018	26	Prospective case-control	Spain	PSG	51.7	NR	126	63	6
Çekiç 2018	27	Prospective cohort	Turkey	PSG	48.22	81	60	30	7
Chambe 2010	34	Prospective cohort	NIL	PSG	NR	NR	106	53	6
Chen 2014	20	Retrospective cohort	Taiwan	ICD9	45.1	78	12640	2528	8
Chirapapaisan 2016	43	Cross-sectional	Thailand	PSG	51.2	56.1	64	32	6
Cristescu 2020	33	Prospective cohort	Romania	PSG	NR	NR	132	65	6
Dikmen 2022	14	Prospective	Turkey	PSG	47.6	52	25	25	1
Ekinci 2013	44	Cross-sectional	Turkey	PSG	49.3	53.06	214	107	5
Ferrandez 2014	45	Cross-sectional	Spain	PSG	50.0	NR	160	80	6
Ferrandez 2016	46	Cross-sectional	Spain	PSG	42.6	NIL	85	40	6
Friedlander 2018	37	Retrospective cohort	USA	PSG	58.6	96.4	225	225	6
Gharraf 2016	47	Cross-sectional	Egypt	PSG	53.7	36.4	67	40	7
Gutiérrez-Díaz 2012	48	Cross-sectional	Spain	PSG	60.1	63.3	30	20	6
Han 2021	35	Retrospective cohort	Australia	PSG	57.0	45.6	535760	9182	9
Huseyinoglu 2013	49	Cross-sectional	Turkey	PSG	51.9	54.6	202	101	6
Kadyan 2010	19	Cross-sectional	UK	ODI	55.8	83.2	176	89	6
Karakucuk 2007	41	Cross-sectional	Turkey	PSG	51.3	75.0	58	31	6
Keenan 2016	21	Retrospective cohort	UK	ICD10	NR	73.0	2751917	67786	5
Küçük 2019	50	Cross-sectional	Turkey	PSG	50.0	48.9	90	45	5
Lee 2019	28	Prospective cohort	Australia	PSG	NR	66.2	356	178	7
Lee 2020	29	Prospective cohort	Australia	PSG	Mild OSA - 59.6 (median) Moderate OSA - 61 (median) Severe OSA - 60.3 (median)	55.0	820	411	8
Lin 2010	51	Cross-sectional	Taiwan	PSG	Mild OSA - 43.3 Moderate OSA - 44.5 Severe OSA - 44.8	86.7	210	105	5
Lin 2011	52	Cross-sectional	Taiwan	PSG	Mild OSA - 41.6 Moderate OSA - 44.1 Severe OSA - 43.7	88.0	406	209	5
Lin 2013	36	Retrospective cohort	Taiwan	PSG	NR	60.2	2138	1012	9
Lin 2019	15	Prospective cohort	Taiwan	PSG	47.8	90.6	32	32	1
Moghimi 2013	53	Cross-sectional	Iran	PSG	47	70.6	100	51	5
Mojon 1999	54	Cross-sectional	Switzerland	PSG	Mild OSA - 50.7 Moderate OSA - 54.4 Severe OSA - 52.7	Mild SAS - 94.44 Mod SAS - 85.71 Severe SAS - 87.10	135	70	4
Morsy 2019	55	Cross-sectional	Egypt	PSG	NR	55	136	80	5
Muniesa 2014	5	Cross-sectional	Spain	PSG	58.1	79.5	238	127	6
Ngoo 2021	56	Cross-sectional	Malaysia	PSG	45.69	32.4	108	54	7
Ozge 2016	30	Prospective case-control	Turkey	PSG	41.3	NR	84	42	5
Ozkurt 2019	58	Cross-sectional	Turkey	PSG	48.1	77.1	72	36	5
Pedrotti 2017	39	Cross-sectional	Italy	PSG	62.2	77	592	592	5
Sagiv 2013	31	Prospective case-control	Isreal	PSG	51.9	70.4	216	108	6
Salzgeber 2014	59	Cross-sectional	Switzerland	PSG	52.7	77.1	70	35	5
Sergi 2007	60	Cross-sectional	Italy	PSG	64	74.5	99	51	5
Ucak 2020	61	Cross-sectional	Turkey	PSG	Mild OSA - 52.17 Mod OSA - 55.91 Severe OSA - 55.42	57.1	154	77	4
Uslu 2020	32	Prospective case-control	Turkey	PSG	46.96	76.6	94	47	5
Xin 2015	62	Cross-sectional	China	PSG	Mild OSA - 54.5 Mod OSA - 57.3 Severe OSA - 50.8	71	119	91	5
Yazgan 2017	63	Cross-sectional	Turkey	PSG	50.88	72.2	212	106	7
Zengin 2012	57	Cross-sectional	Turkey	PSG	52.2	30.7	121	64	5
Zengin 2014	40	Cross-sectional	Turkey	PSG	52.6	61	88	44	5

PSG Polysomnography.

ICD International Classification of Diseases.

NR Not reported.

OSA Obstructive sleep apnea.

Table 2.

Additional Characteristics of included studies.

First author and year published	Relevant Interventions (CPAP)	Exposures	Outcomes	Covariates (where adjustment was performed)	Key findings
Aptel 2014	NIL	OSA, > 50 years old	Glaucoma prevalence rate (with adjusted OR)	Age, BMI, gender, hypertension, high triglyceride levels, thyroid dysfunction	OSA diagnosis did not significantly influence the risk of glaucoma in univariate analysis (odds ratio [OR], 1.13; 95% confidence interval [CI], 0.87–1.47).
Asker 2013	NIL	OSA	IOP	Unadjusted	No statistical difference was found between the groups of patients with OSA and healthy control subjects in IOP.
Bagabas 2019	NIL	OSA	CDR, IOP, adjusted odds of glaucoma	Age, BMI, gender, diabetes, hypertension, family history of glaucoma, smoking status	The mean CDR among subjects with OSA were similar to the ratios of participants without OSA: 0.24 vs. 0.25 in the right eye and 0.25 vs. 0.24 in the left eye, and the p-values were not statistically significant. Individuals with OSA had a significantly higher IOP in the left eye (p < 0.05) and a p-value close to significance for the IOP level in the right eye (p = 0.051). ORs associated with having a diagnosis of glaucoma for OSA patients compared with OSA-free individuals were 2.21 (crude OR) and 1.71 (adjusted OR), which were not statistically significant.
Batum 2020	CPAP treatment (6 months)	Severe OSA	IOP, RNFL thickness (before and after CPAP)	Unadjusted	No significant difference was found in IOP in severe OSAS group before treatment and control group (p = 0.732). No significant difference was found in IOP before and after treatment in severe OSAS group (p = 0.654). In the comparison between the severe OSA (before treatment) and control groups, thinning was found in pRNFL (average, nasal, inferior) and in the macular layers (external and internal superior quadrants) (p < 0.05).
Carnero 2020	NIL	OSA suspects	IOP	Unadjusted	The IOP curves exhibited a nocturnal acrophase followed by plateau times in acrophase (PTs) of varying extents at which the IOP remained higher than daytime measurement with small variations.
Casas 2013	NIL	OSA	CDR. IOP, RNFL thickness, visual field	Unadjusted	There was no significant difference in CDR between OSA and the control group. IOP values were higher in OSA group than those in controls (p < 0.001, Student’s t-test). OSA patients showed a significant reduction of the nasal quadrant RNFL thickness (74.7 ± 15.8 μm) compared with those values observed in control patients (81.1 ± 16.6 μm, p = 0.047, Student’s t-test). No differences in peri-papillary RNFL thickness were observed when dividing the OSA group in accordance with disease severity. Humphrey visual field results, such as visual field index (VFI), mean deviation (MD), and pattern standard deviation (PSD), were significantly altered in patients with OSA patients as compared with the control group (p = 0.009, p = 0.002, p = 0.014 respectively by Student’s t-test).
Casas 2018	NIL	OSA	IOP, RNFL thickness, visual field	Unadjusted	There was no significant difference in IOP between the OSA and controls. When dividing the OSA group in accordance with the severity of the disease, nasal peripapillary RNFL thickness was significantly lower in severe OSA than that in controls and mild-moderate cases (p = 0.031 and p = 0.016 respectively, Mann–Whitney U test). OSA patients showed a significant reduction of the sensitivity for superior visual field division (p = 0.034, t-student test).
Çekiç 2018	NIL	OSA	IOP	Unadjusted	The IOP values were significantly higher in the severe OSA group (p < 0.05).
Chambe 2010	NIL	OSA	Glaucoma incidence	Unadjusted	No significant association was found between glaucoma and OSA.
Chen 2014	Surgery only, CPAP only, Multiple modality treatment	OSA	Glaucoma incidence (without treatment, with surgery, with CPAP), adjusted HR of glaucoma	Age, gender, diabetes, hypertension, coronary artery disease, hyperlipidemia	The adjusted HR of glaucoma for OSA patients was 1.88 (95% CI: 1.46–2.42), compared with controls. For patients without treatment, the adjusted HR was 2.15 (95% CI: 1.60–2.88). For patients with treatments, the adjusted HRs of glaucoma were not significantly different from controls, except for those with CPAP (adjusted HR = 1.65, 95% CI = 1.09–2.49).
Chirapapaisan 2016	NIL	OSA	CDR, IOP, RNFL thickness	Unadjusted	No significant differences in age, BMI, and CDR were found among the four groups (=0.92, 0.07, and 0.13, respectively, when compared with control subjects). IOP was elevated but within the normal range in the severe group only (p = 0.04, one-way ANOVA). The average and mean RNFL thickness in most of the four quadrants of the severe OSA group trended toward being less than those in the comparable quadrants of the other groups in both the morning and evening. In the moderate OSA group, the average RNFL thickness and temporal and superior quadrant thickness in the morning were significantly higher than in the evening (=0.01, =0.01, and =0.03, respectively). In the severe OSA group, the inferior quadrant thickness in the morning was significantly higher than in the evening (=0.03).
Cristescu 2020	CPAP	OSA	Glaucoma prevalence rate, IOP (before and after CPAP)	Unadjusted	3.09% of the patients (2/ 65) had glaucoma in the OSA group, and 10.26% (4/ 39) in the control. The difference was not significant (Fischer’s exact test: p-value = 0.1940). The IOP was similar in both groups (p = 0,9786). In the OSA group, IOP increased with a higher AHI, but not significantly (p = 0,057). CPAP treatment had no effect on IOP, as initial and final medium IOP were 15.06 and 14.89 respectively (p = 0,8327).
Dikmen 2022	CPAP (3 months)	OSA	IOP, RNFL thickness (before and after CPAP)	Unadjusted	Compared to the pre-treatment measurements at the time of diagnosis, post-treatment measurements showed insignificant decrease in IOP, significant increase in mean macular thickness and significant thinning in superior nasal RNFL. There was no correlation found between pre-treatment measurements including IOP, OCT, CCT, BMI, and sleep parameters. The effect of intermittent hypoxia and hypercapnia on the IOP and macula may be reversible in severe OSA patients receiving 3 months of PAP therapy, but the reversibility of the neurodegenerative effects of OSA on RNFL with this treatment seems controversial.
Ekinci 2013	NIL	OSA	IOP	Unadjusted	Ophthalmologic examination findings including Goldmann applanation tonometry was 18.17 + /- 1.58, 18.16 + /- 1.47 in the study and control group, respectively.
Ferrandez 2014	NIL	OSA	IOP, RNFL thickness, visual field	Unadjusted	IOP was lower in the OSA group. RNFL thickness measured with OCT did not differ significantly between groups. Visual field index (VFI) was lower in the OSA patients (99.52 ± 0.7 versus 97.51 ± 1.1; p < 0.001).
Ferrandez 2016	NIL	OSA	IOP, RNFL thickness	Unadjusted	IOP was significantly lower in the OSA group (17.42 ± 2.6 mmHg in healthy subjects and 14.23 ± 2.6 mmHg in OSA patients; p < 0.001) RNFL thickness measured with OCT was similar between groups.
Friedlander 2018	CPAP	OSA	Glaucoma prevalence rate	Unadjusted	The glaucoma prevalence rate among medical center’s general population was 2.5% which was significantly less (P < .00001) than among those with comorbid OSA.
Gharraf 2016	CPAP	OSA	Glaucoma prevalence rate, CDR, IOP	Unadjusted	Glaucoma was diagnosed in 14 out of 40 patients with OSA yielding an estimate prevalence of 35% among patients with OSA compared to only one case of glaucoma in obese patients without OSA that is to say 6.7% with statistically significant difference between the two groups. CDR in glaucoma patients (CDR = 0.40 (0.20–0.80)) was significantly higher (p = 0.015) than that in the control group (CDR = 0.30 (0.20–0.60)). Mean ± SD value of IOP were 16.73 ± 2.39 mmHg, and 15.53 ± 1.92 mmHg in both patients and controls respectively with no statistically significant difference between the two groups.
Gutiérrez-Díaz 2012	NIL	OSA	IOP, RNFL thickness, Visual field	Unadjusted	There were no significant differences between groups in regard to the IOP.The average RNFL thickness was significantly reduced in the glaucoma eyes compared to the control database and the non-glaucoma eyes. On the Humphrey Visual Field total deviation analysis, all the glaucoma eyes showed significant clusters of abnormal points but none was detected in the non-glaucoma group eyes.
Han 2021	NIL	Middle-age and older participants from the UK	Adjusted odds and hazards ratios (HR) of glaucoma	Age, waist-to-hip ratio, gender, ethnicity, smoking status, systolic blood pressure, diabetes, cardiovascular disease, HDL cholesterol, total cholesterol, Townsend deprivation index	During the 8-year follow-up in the UK Biobank, glaucoma incidence rates per 1000 person-years were 2.46 and 1.59 for participants with and without sleep apnoea. Multivariable adjusted hazard ratios of glaucoma risk for sleep apnoea was 1.33 (95% confidence interval [CI] 1.10–1.60, P = 0.003) relative to participants without sleep apnoea. In the Canadian Longitudinal Study on Aging (CLSA) cohort, disease information was collected through in-person interview questionnaires. During the 3-year follow-up, glaucoma incidence rates per 1000 person-years for those with and without sleep apnoea were 9.31 and 6.97. In the CLSA, similar associations were identified, with glaucoma odds ratios of 1.43 (95% CI 1.13–1.79) in participants with sleep apnoea compared to those without sleep apnoea (P < 0.001).
Huseyinoglu 2013	NIL	OSA	CDR, IOP, RNFL thickness, Visual field	Unadjusted	No significant difference was found among all of the groups in terms of CDR. IOP values were significantly higher in all of the OSA subgroups compared with the control group. There was a significant decrease in RNFL thickness only in the patients with severe OSA compared with the other group. VF parameters were significantly worsened in all the OSA subgroups compared to the control group.
Kadyan 2010	CPAP	OSA	Glaucoma prevalence rate, IOP	Unadjusted	Glaucoma prevalence in OSA patients (3/89, 3.4%) was similar to the controls (1/26, 3.8%) (P = 0.92). There was no significant difference in IOP between OSA and controls.
Karakucuk 2007	NIL	OSA	Glaucoma prevalence rate, IOP	Unadjusted	The prevalence of glaucoma in the group of patients with OSA was 12.9% (4/31); all of these 4 patients with glaucoma were in the “severe” OSA group. No statistically significant difference was found between IOP of patients and controls (p > 0.05). There was a positive correlation between IOP and the AHI (p = 0.001).
Keenan 2016	NIL	OSA, > 55 years old	Glaucoma prevalence rate	Unadjusted	Rate ratio for glaucoma was 1.01 (95% CI 0.85 to 1.19).
Küçük 2019	NIL	OSA	RNFL thickness, IOP	Unadjusted	Evaluation of the study parameters revealed that there were no significant differences between the OSA patients and controls regarding IOP and RNFL thickness.
Lee 2019	NIL	None	IOP, RNFL thickness	Unadjusted	There was no difference in IOP between those with and without OSA or among OSA severity groups (P > 0.05). Participants with OSA on average showed thinner peripapillary RNFL at the inferotemporal (P = 0.026) and superotemporal (P 1⁄4 0.008) segments compared with those without OSA. Additionally, higher AHI results were associated with thinner RNFL superotemporally (P = 0.007).
Lee 2020	NIL	None	RNFL thickness	Unadjusted	Participants with severe OSA had thinner RNFL superotemporally than those without OSA or with mild OSA (P < 0.001 and 0.001, respectively).
Lin 2010	NIL	OSA	CDR, IOP, RNFL thickness, visual field	Unadjusted	No significant difference in CDR between OSA and control (p = 0.343). The IOP was higher in the severe OSA group compared with normal controls (p = 0.016). RNFL thickness was significantly lower for the severe OSA group than for the control and mild OSA groups in the average (p < 0.0001) and in the superior quadrant (p = 0.0007). When subjects without OSA or with mild disease (AHI < 15) were grouped together and compared with patients with moderate/severe OSA (AHI ≧ 15), RNFL thickness measurements for the latter group were significantly lower in the average (p < 0.0001), and in the superior (p = 0.001), inferior (p = 0.029), and temporal (p = 0.007) quadrants. Positive correlations were identified between lowest oxygenation saturation on PSG and RNFL thickness in the average (r = 0.260), superior (r = 0.200) and nasal (r = 0.156) quadrants. No significant differences in spherical equivalence (SE) and mean defect (MD) of the VF were found among the four groups.
Lin 2011	NIL	OSA	Glaucoma prevalence rate, IOP, visual field	Unadjusted	The prevalence of glaucoma in moderate/severe OSA patients was 7.1%, significantly higher (P = 0.033) than that in normal/mild OSA patients. There was no significant difference in IOP and visual field across the different groups.
Lin 2013	NIL	OSA	Adjusted HR of glaucoma	Number of visits for ophthalmologic care during the follow-up period, monthly income, diabetes, hypertension, coronary heart disease, obesity, hyperlipidemia, renal disease, hypothyroidism, and migraine	Hazard ratio for glaucoma within the 5-year period for subjects with OSA was 1.67 (95% CI, 1.30e2.17; P < 0.001) that of comparison subjects.
Lin 2019	CPAP (3 months)	OSA	RNFL thickness (before and after CPAP)	Unadjusted	Inferior quadrant and nasal-inferior sector of RNFL thickness significantly improved after treatment (p = 0.025 and 0.004, respectively).
Moghimi 2013	NIL	OSA	Visual field, IOP, visual field, RNFL thickness	Unadjusted	IOP was significantly higher in the OSA group before (p < 0.001) and after (p < 0.001) correcting for CCT. There was a significant positive correlation between AHI and IOP. No significant correlation was detected between AHI and visual field indices MD (p = 0.41) and PSD (p = 0.44). In RNFL analysis, DISK average, superior average, and NFI were significantly worse in the OSA group compared with the control group. In this study, the only RNFL parameter that correlated with AHI was NFI.
Mojon 1999	NIL	Ethnically white participants with suspected OSA	Glaucoma prevalence rate	Unadjusted	The observed prevalence of glaucoma in patients with OSA (5 of 69, 7.2%) was significantly higher than expected in a white population (2%) (P = 0.01).
Morsy 2019	NIL	OSA	Glaucoma prevalence rate, CDR, IOP, visual field, RNFL thickness	Unadjusted	OSA patients n = 24 (30%) had glaucoma. Among patients who had NTG the mean IOP was 16.14 ± 1.32 mmHg, while the mean IOP in controls was 15.45 ± (1.41) (P-value = 0.325). The mean Cup-Disc (C/D) ratio was 0.671 ± 0.21 and 0.243 ± 0.213 in glaucoma and controlled groups respectively. In the visual fields of glaucomatous patients, the mean deviation value was -6.12 ± 2.01decibels, while the mean pattern standard deviation value was 6.51 ± 1.44 decibels. Parapapillary RNFL thickness was 69.34 ± 11.33 in patients with NTG and 115 ± 12.52 in the control group (p = 0.001).
Muniesa 2014	NIL	OSA and floppy eyelid syndrome (FES)	Glaucoma prevalence rate	Unadjusted	The difference in the prevalence of glaucoma between OSA patients without FES (5.3%) and OSA patients with FES (23.07%) was statistically significant (p = 0.004).
Ngoo 2021	CPAP	OSA	CDR, IOP, RNFL thickness	Unadjusted	There was no significant difference (p = 0.884) in average CDR between OSA (0.567 ± 0.108) and the control group (0.563 ± 0.163). Both groups had normal mean IOP at 15.3 for OSA patients and 14.7 for control patients. RNFL thickness fairly correlate with severity of the disease (AHI), superior RNFL with R = 0.293, R2 = 0.087, p = 0.030 (p < 0.05), and nasal RNFL R = 0.292, R2 = 0.085, p = 0.032.
Ozge 2016	NIL	OSA	RNFL thickness	Unadjusted	When compared with controls, the OSA group had significantly thinner nasal RNFL thickness in the right eye (p = 0.01) and thinner mean RNFL thickness in both eyes (p < 0.001). Other RNFL thickness measurements were similar between groups (p > 0.05).
Ozkurt 2019	NIL	OSA	RNFL thickness	Unadjusted	RNFL segment values showed no difference between OSAS and control groups. RNFL superior segment and inferonasal segments of the mild OSA subgroup were found to be thinner than the moderate and severe OSA subgroups (p = 0.034, p = 0.025, respectively). RNFL point values also showed no difference between the two groups. The RNFL inferior point of the mild OSAS subgroup was found to be thinner than the moderate and severe subgroups (p = 0.036).
Pedrotti 2017	NIL	OSA	Adjusted odds of glaucoma	Age, BMI, gender, hypertension	Severe OSA was significantly associated with glaucoma (OR, 95% CI 1.05 to 5.93, p = 0.037).
Sagiv 2013	NIL	OSA	RNFL thickness	Unadjusted	RNFL was thinner for patients with OSA than that of normal controls by an average of 4.20 mm (P < 0.003), in the superior quadrant by 4.83 mm (P = 0.028) and in the inferior quadrant by 5.19 mm (p = 0.016).
Salzgeber 2014	NIL	OSA	IOP, CDR, RNFL thickness	Unadjusted	A difference was found in mean IOP, although in a fully overlapping range, sleep apnea group: 15.2 ± 3.1, range 8–22 mmHg, controls: 13.6 ± 2.3, range 9–18 mmHg; p < 0.01. None of RNFL parameters and CDR showed a significant difference between the groups.
Sergi 2007	NIL	Ethnically white participants with OSA	Glaucoma prevalence, IOP, CDR, RNFL thickness, visual field	Unadjusted	Three of 51 OSA patients (5.9%) had glaucoma. No patient in the control group had OSA or glaucoma. CDR did not differ between the study patients and the control group. IOP, mean deviation (MD) in the visual field and the reduction of mean retinal nerve fiber layer (RNFL) thickness were significantly increased in OSAS patients compared with controls subjects (P < 0.0001 and P < 0.05, respectively).
Ucak 2020	NIL	OSA	RNFL thickness, IOP	Unadjusted	The IOP values were significantly higher in the severe OSA group than the control cases (p = 0.001). The average retinal nerve fiber layer (RNFL) thickness and the RNFL thickness of the temporal and inferior quadrants were significantly lower in the severe OSA group compared with the control cases (p < 0.05).
Uslu 2020	NIL	OSA	IOP, RNFL thickness, CDR	Unadjusted	There was no significant difference in IOP between the OSA and control groups. The average peripapillary RNFL thickness was lower in obstructive sleep apnea syndrome patients than in controls (90.74 ± 7.98 vs 94.93 ± 6.13 µm; p = 0.008). The average CDR and the vertical CDR in patients with OSA were significantly higher compared to the controls (both p < 0.001).
Xin 2015	NIL	OSA	Glaucoma prevalence rate, CDR. RNFL thickness, IOP, visual field	Unadjusted	There was no significant difference in CDR between the groups. Both moderate and severe groups showed significantly higher IOP than the control group (p = 0.022 and 0.001). Significant differences were found in nasal RNFL and inferior RNFL among the four groups (p = 0.013, p = 0.004). Nasal RNFL correlated with the ODI (p = 0.048). Totally, the prevalence of primary open angle glaucoma was 5.49 %. A significant negative correlation between the AHI and MD (r = −0.206, p = 0.024) and a significant positive correlation between the PSD and AHI (r = 0.323, p < 0.001) were found.
Yazgan 2017	NIL	OSA	IOP, RNFL thickness		No significant difference was found between the two groups regarding IOP. There was no difference between the control and mild groups regarding the RNFL thicknesses except nasal RNFL and inferotemporal RNFL which are thinner in the mild group. The RNFL thicknesses of all quadrants were significantly thicker in the control group compared with moderate and severe subgroups (p < 0.05).
Zengin 2012	NIL	OSA	Glaucoma prevalence rate, IOP, RNFL thickness	Unadjusted	There was a statistically significant correlation among AHI, mean deviation and retinal nerve fiber layer thickness (p < 0.05). Glaucoma was diagnosed in seven patients with OSAS (10.9%). There was no significant difference in IOP between OSA patients and the control group (p = 0.233).
Zengin 2014	CPAP	OSA	IOP, RNFL thickness	Unadjusted	The mean IOP of the OSA patients and controls showed no significant difference at baseline and last control. There was a statistically significant correlation between AHI and RNFL thickness (<0.05)

CPAP Continuous positive-airway pressure.

OSA Obstructive sleep apnea.

BMI Body-mass-index.

AHI Apnea-hypopnea index.

CDR Cup-disc ratio.

RNFL Retinal nerve fibre layer.

IOP Intra-ocular pressure.

OCT Optical coherence tomography.

CCT Central corneal thickness.

Glaucoma diagnoses

19 studies were included in this analysis, and the pooled unadjusted outcomes are presented in Fig. 1. Glaucoma was positively associated with OSA (OR 3.66, 95% CI 1.70 to 7.90, I² = 98%, p < 0.01).

Fig. 1. Unadjusted odds ratio for the risk of glaucoma in patients with OSA.

OSA Obstructive Sleep Apnea, CI confidence interval.

Subgroup analysis based on glaucoma subtype – open angle glaucoma suspects (OAGS), normal tension glaucoma (NTG) and high pressure primary open angle glaucoma (POAG) – was performed and is presented in Fig. 2. We identified Keenan et al. [21] as a source of heterogeneity in our analysis of open angle glaucoma and performed further sensitivity analysis after excluding these two studies – Keenan et al’s patient population were not actual patients but were constructed from linked English hospital episode statistics from 1999 to 2011. Nine studies reported on NTG and 10 reported on POAG. Patients with OSA were 6.72 times as likely to develop NTG (OR 2.78, 95%CI 1.52 to 5.10, I² = 0%, p < 0.01), and 2 times as likely to develop POAG (OR 1.93, 95% CI 1.49 to 2.50, I² = 0%) compared to patients without OSA.

Fig. 2. Unadjusted odds ratios for the risk of glaucoma in patients with OSA, stratified by glaucoma subtype.

OSA Obstructive Sleep Apnea, CI confidence interval.

Further analysis was performed on the adjusted odds ratios (OR) and hazard ratios (HR) of glaucoma, comparing patients with OSA against those without OSA. Three studies reported on adjusted HR [20, 35, 36] and four reported on adjusted OR [22, 35, 38, 39]. All studies adjusted for age, sex, and various comorbidities including hypertension, hyperglycaemia, coronary heart disease, diabetes and obesity. Bagabas et al. also controlled for family history of glaucoma and ocular comorbidities respectively. The pooled outcomes are presented in Figs. 3 and 4. Both adjusted OR (adjusted OR 1.40, 95% CI 1.05 to 1.86, I² = 46.4%, p = 0.02) and HR (adjusted HR 1.59, 95% CI 1.29 to 1.96, I² = 59.7% p < 0.001) demonstrated a significantly increased risk of glaucoma in patients with OSA compared to those without OSA.

Fig. 3. Adjusted odds ratios for the risk of glaucoma in patients with OSA.

RE: random-effects; adjustment performed for age, gender, body-mass-index (BMI) and patient comorbidities.

Fig. 4. Adjusted hazards ratios for the risk of glaucoma in patients with OSA.

Adjustment performed for age, gender, body-mass-index (BMI) and patient comorbidities.

Cup-Disc Ratio (CDR)

Twelve studies reported on the average CDR comparing patients with OSA against those without OSA. The results are shown in Fig. 5. Patients with OSA had a CDR comparable to that of patients without OSA (mean difference (MD) 0.02, 95% CI −0.01 to 0.06, I² = 62%, p = 0.13).

Fig. 5. Mean difference in cup-disc ratio.

Comparing patients with OSA against patients without OSA. OSA Obstructive Sleep Apnea, CI confidence interval.

Visual field defects

Visual field defects can be quantified by both the MD and PSD. The MD of the visual field is a statistical index representing the average difference in visual field sensitivity comparing the patient against the mean sensitivity of a normal individual of the same age – MD is utilized to monitor the overall visual field changes over time as well [64]. Those who require brighter stimuli will have more negative MD values. PSD quantifies the irregularities in visual field loss, which is commonly found to be non-uniform in glaucomatous visual defect.

Nine studies reported on the visual field MD. The results are shown in Fig. 6. Patients with OSA had more significant defects compared to patients without OSA (MD −0.90, 95% CI −1.60 to −0.20, I² = 78%, p = 0.02). Eight studies reported on the visual field PSD. The results are shown in Fig. 7. Patients with OSA had more significant irregularities compared to patients without OSA (MD 0.35, 95% CI 0.04 to 0.65, I² = 72%, p = 0.03).

Fig. 6. Mean difference in visual field MD.

Comparing patients with OSA against patients without OSA. OSA Obstructive Sleep Apnea, CI confidence interval.

Fig. 7. Mean difference in visual field PSD.

Comparing patients with OSA against patients without OSA. OSA Obstructive Sleep Apnea, CI confidence interval.

Intraocular pressure

31 studies reported on IOP in patients with OSA compared against those without OSA. The results are represented in the forest plot (Fig. 8). Patients with OSA had an increased IOP which was statistically significant (MD 0.78, 95% CI 0.19–1.37, I² = 91%, p = 0.01). Subgroup analysis based on OSA severity were performed, demonstrating a significantly raised IOP in patients with mild, moderate and severe OSA. Nominally, those with severe OSA had the largest increase in IOP compared to those without OSA (MD 0.85, 95% CI 0.16 to 1.54, I² = 67%, p = 0.02), followed by those with moderate OSA (MD 0.72, 95% CI 0.06 to 1.41, I² = 51%, p = 0.03). Those with mild OSA had the nominally smallest mean difference compared to patients without OSA (MD 0.53, 95% CI 0.07 to 0.99, I² = 23%, p = 0.02). These results of this subgroup analysis are shown in Fig. 9.

Fig. 8. Mean difference in intraocular pressure.

Comparing patients with OSA against patients without OSA. OSA Obstructive Sleep Apnea, CI confidence interval.

Fig. 9. Mean difference intraocular pressure stratified by OSA severity.

Comparing patients with OSA against patients without OSA. OSA Obstructive Sleep Apnea; CI confidence interval.

Retinal Nerve Fibre Layer (RNFL) thickness

26 studies reported on RNFL comparing patients with OSA against those without OSA. The results are represented in the forest plot (Fig. 10). Patients with OSA had thinner RNFL which was statistically significant (MD −3.45 µm, 95% CI −4.78 to −2.12, I² = 78%, p < 0.01). Subgroup analysis based on OSA severity were performed, demonstrating a significantly thinner RNFL in patients with mild, moderate and severe OSA. Those with severe OSA had a significantly thinner RNFL compared to those without OSA (MD -4.28 µm, 95% CI −6.71 to −1.85, I² = 54%, p < 0.01). Those with moderate OSA (MD −2.63 µm, 95% CI −5.81 to 0.56, I² = 40%, p = 0.09) and mild OSA(MD −1.34 µm, 95% CI −3.25 to 0.56, I² = 34%, p = 0.14) had thinner RNFL compared to patients without OSA but did not reach statistical significance. The results of this subgroup analysis are shown in Fig. 11.

Fig. 10. Mean difference in retinal nerve fibre layer thickness.

Comparing patients with OSA against patients without OSA. OSA Obstructive Sleep Apnea, CI confidence interval.

Fig. 11. Mean difference retinal nerve fibre layer thickness, stratified by OSA severity.

Comparing patients with OSA against patients without OSA. OSA Obstructive Sleep Apnea, CI confidence interval.

As subgroup heterogeneity was high, we conducted further sensitivity analyses looking at RNFL thickness specifically in non-glaucomatous eyes and stratified them by OSA severity. Similar to previously described results, those with severe OSA had a significantly thinner RNFL compared to those without OSA (MD −3.52 µm, 95% CI −6.01 to −1.03, I² = 49%, p = 0.01). Those with moderate OSA (MD −2.36 µm, 95% CI −5.77 to 1.06, I² = 0%, p = 0.12) and mild OSA(MD −0.99 µm, 95% CI −2.63 to 0.65, I² = 0%, p = 0.17) had thinner RNFL compared to patients without OSA but did not reach statistical significance. The results of this subgroup analysis are shown in Fig. 12.

Fig. 12. Mean difference retinal nerve fibre layer thickness in non-glaucomatous eyes, comparing patients with OSA against patients without OSA, stratified by OSA severity.

OSA Obstructive Sleep Apnea, CI confidence interval.

Effects of Continuous Positive Airway Pressure (CPAP) treatment

Several studies looked at the effect of CPAP treatment on ocular parameters, including IOP and RNFL thickness. Dikmen et al. [14] found that 3 months of CPAP treatment in severe OSA patients reduced IOP and increased RNFL thickness, but these changes were not statistically significant. Lin 2019 [15] also followed patients up after 3 months of CPAP treatment but found no change in the IOP, RNFL, macular thickness or visual field PSD pre- and post-treatment. However, they found a significant difference in the visual field mean deviation.

Batum 2020 [7] followed up 30 patients with severe OSA for a longer duration – after at least 6 months of CPAP treatment, there was a statistically significant increase in peripapillary RNFL thickness (p = 0.005) and macular thickness (p = 0.006).

Zengin 2014 [40] followed up 44 patients with OSA of varying severity for 12 months; all patients were on CPAP therapy. After twelve months, there was a statistically significant thinning of RNFL observed in OSA patients, compared against healthy controls over the same period.

Certainty of evidence

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) system (Supplemental Table 3) was utilized to assess the quality of pooled evidence. The quality of evidence was moderate for the primary outcomes before subgrouping; they were downgraded by two levels for substantial statistical heterogeneity. This was except for the adjusted odds and hazard ratios which had moderate statistical heterogeneity. In our subgroup analyses, as reported above, statistical heterogeneity of each primary outcome was lowered to the low or moderate level.

Discussion

In this systematic review and meta-analysis of 3,316,034 patients, we found that patients with OSA had higher odds of glaucoma compared to patients without OSA (OR 3.66, 95% CI 1.70 to 7.90, I² = 98%, p < 0.01). After adjustment for various important confounders including age, gender, and patient comorbidities such as hyperlipidaemia, hypertension, cardiovascular diseases and diabetes, patients with OSA had up to 40% higher odds of glaucoma.

After subgroup analyses of two common subtypes of chronic glaucoma – NTG and open angle glaucoma – we identified a significantly higher risk of both diseases in patients with OSA with the risk of NTG being nominally higher. We acknowledge that the values of the unadjusted ORs may be exaggerated due to the presence of confounders; however, given the significant findings as well as a heterogeneity of 0% in both subgroups after sensitivity analyses, we suggest our findings should not be easily dismissed.

Apart from the diagnosis of glaucoma itself, we also looked at quantifiable markers of glaucomatous change in the eye including characteristic visual field defects, IOP and RNFL thickness. We prioritized analysis of data reported on non-glaucomatous eyes, as glaucomatous eyes naturally have worse ocular parameters. Patients with OSA had more significant visual field defects as well as the irregularity of the visual field loss, compared to patients without OSA. They also had a significantly higher IOP and a thinner RNFL thickness possibly signalling more advanced disease. Patients with severe OSA had the highest IOP and thinnest RNFLs compared to healthy individuals. While a higher IOP of 0.78 mmHg in patients with OSA may not be clinically meaningful, this finding suggests that the increased odds of glaucoma observed in OSA populations may be driven by an IOP-independent mechanism. This is an interesting finding that warrants further investigation. Interestingly, our analysis showed that patients with OSA had a cup-to-disc ratio comparable to that of healthy controls. Subgroup analyses also revealed that severity of OSA has a significant effect on the severity of these ocular parameters.

The pathophysiology underlying the observed association between OSA and glaucoma has been proposed to be multifactorial, including hypoxic damage, vascular changes as well as mechanical factors [4]. In terms of hypoxic changes, OSA results in Sleep Fragmentation (SF) which raises the sympathetic drive and increases systemic inflammation as well as immune dysregulation [65–67]. Repetitive apnoea-hypopnea episodes may also directly damage the optic nerve head, retinal ganglion cells and its axons, leading to glaucoma [68, 69]. In addition, the raised sympathetic drive causes vasoconstriction and hypertension and has been shown to alter both cerebral and ocular circulation [69]. Another contributing factor may be the hypoxia-inducible factors (HIFs) seen in patients with OSA. In vitro models and animal studies have demonstrated that HIFs promote vascular endothelial dysfunction and degenerative vascular disease via NF-κB-mediated inflammatory pathways, increased C-reactive protein levels, and proinflammatory cytokines [70–72]. OSA may also contribute to mechanical pathways in glaucoma progression by raising the IOP, especially during the night. A supine position at night while patients are sleeping as well as a raised episcleral venous pressure have been suggested as possible explanations for this observation [5, 73]. The effect of OSA treatment on the interplay between glaucoma and OSA is a very important question that remains unanswered. There were insufficient studies in our study to perform meaningful statistical analysis. The studies available seem promising in their suggestion that CPAP treatment indeed improves ocular parameters known to be associated with glaucomatous progression including IOP and RNFL thickness. In particular, Batum et al’s [23] findings suggest that a longer follow-duration of at least 6 months is necessary to identify a significant improvement in certain ocular parameters. However, previous evidence seems to challenge these recent findings.

Alvarez-Sala et al. [10] first described the observation that CPAP significantly increases IOP in patients with coexisting glaucoma; subsequent studies by Kiekens et al. [74] and Pepin et al. [75] suggest that CPAP therapy was associated with additional rises in IOP which was worse at night. This seemingly contradictory finding is hypothesized to be due to an increased intrathoracic pressure associated with the use of CPAP, and the additional IOP rise at night from CPAP use may explain glaucoma progression seen in some patients being treated for OSA. Muniesa et al’s [5] recent study looking at the effect of CPAP therapy on 24 h IOP changes in OSA patients found that in 17 of their 22 patients, the highest IOP readings were recorded at night, suggesting that CPAP may indeed be the culprit.

It is important to clarify the underlying pathophysiology of these changes given that CPAP remains the mainstay of treatment in OSA. Chen et al. [20] found that patients with OSA who underwent any form of surgical treatment, including pharyngeal and nasal surgeries, had a lower risk of glaucoma compared to those treated with CPAP alone. This suggests that patients with concomitant OSA and glaucoma should perhaps opt for alternative treatments of OSA including surgical options. Overall, evidence remains controversial, and more studies need to be done to evaluate the effects of various OSA therapies on glaucoma patients.

Given the large number of OSA cases which go undiagnosed and untreated [3], it is important for clinicians to be cognizant of a possible association between these two conditions. A multidisciplinary approach involving both sleep clinicians and ophthalmologists may be ideal for optimizing care for patients diagnosed with either condition, or both. Patients with OSA, especially those with severe sleep apnoea, may benefit from regular ophthalmic examination to pick up asymptomatic or early cases of glaucomatous disease. Future studies should consider basing their assessment of nocturnal hypoxemia on pulse oximetry nadir and area-under-curve analyses of patients’ oxygen saturation such as the cumulative sleep time percentage with oxyhaemoglobin saturation <90% (CT90), which may be better markers than AHI which is used for OSA detection and severity. Furthermore, given the suggestion that sleep position may affect IOP, it may be worth examining if the eyes of patients with positional OSA are affected more severely than those with non-positional OSA. Further studies into the spectrum of glaucoma, including glaucoma suspects or angle closure glaucoma may be helpful in picking out patients at higher risk.

Strengths and limitations

The strengths of our study lie in the large number of studies and participants, as well as the large number of outcomes we considered including characteristic structural and visual field defects seen in glaucoma, on top of glaucoma prevalence and incidence rates. We have included many new studies that have been published since the last meta-analysis was performed, and considered important nuances including the subtypes of glaucoma, adjusted OR and HR values as well as the severity of OSA. Assessment of studies’ risk of bias was performed, with almost all our studies having a moderate to low risk of bias. Nevertheless, the findings of our studies should be interpreted in due consideration of the limitations.

There was significant heterogeneity in our primary outcomes. Glaucoma is a heterogenous disease with many ways of classifying the disease – be it in terms of its chronicity, aetiology or the IOP. With subgroup analyses based on the severity of OSA and glaucoma subtype, as well as further sensitivity analyses performed on data purely from non-glaucomatous eyes, heterogeneity was lowered substantially, to 0% in certain subgroups. Ideally, sufficient adjustment should be performed for ocular parameters to limit the effect of confounders on our findings. The majority of our studies were cross-sectional studies which can only suggest an association between the two conditions rather than any cause-and-effect relation.

There remain other aspects of glaucoma-related structural changes not reviewed in our study, including the thickness of the choroidal and macular layers. Unfortunately, we did not have sufficient data to pool these outcomes and future primary studies looking at these areas may contribute to a deeper understanding. In the same vein, no meaningful conclusion on the effect of OSA treatment could be summarized based on the studies which have been performed thus far. Nonetheless, future studies with randomized-controlled trials being the gold-standard may prove useful to elucidating the effects of various surgical and non-surgical options on patients with concomitant glaucoma diagnoses.

Conclusions

In this meta-analysis, OSA was associated with higher risk of glaucoma, as well as more severe ocular findings characteristic of the glaucomatous disease process. We suggest more clinical studies looking into the effects of OSA treatment on the progression of glaucoma to help clinical decision making for patients.

Author contributions

AJYC, SKXW, CYW, TA and STT designed the study, study protocol and tools. AJYC, SKXW, CYW, KHY, KJYN and FWXX conducted the data collection. AJYC, UA, ACWN, AS, SRHL, TA and STT were responsible for analysing and interpreting the data and wrote the manuscript. All authors read and approved the final manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41433-023-02471-6.