Association and causal inference between obstructive sleep apnea and retinal nerve fiber layer thickness: a meta analysis and bidirectional Mendelian randomization study

Abstract

Obstructive sleep apnea (OSA) is a common sleep-related breathing disorder that may contribute to retinal neurodegeneration through intermittent hypoxia and vascular dysregulation. However, the association between OSA and retinal nerve fiber layer (RNFL) thickness, as well as their potential causal relationship, remains unclear. Observational studies comparing RNFL thickness between patients with OSA and controls, or across different OSA severity groups, were systematically retrieved from eight primary databases from database inception to February 2026, with supplementary searches conducted in Google Scholar, ClinicalTrials.gov, and ChiCTR. Meta-analysis was performed using RevMan 5.4 and Stata 15.1, with subgroup, sensitivity, and publication bias analyses conducted. In the bidirectional two-sample MR analysis, GWAS data for RNFL were obtained from the GWAS Catalog, and OSA data were derived from FinnGen R13. The IVW method was used as the primary analytical approach, with the weighted median, MR-Egger, simple mode, and weighted mode methods used as complementary analyses. A total of 62 studies were ultimately included in the meta-analysis. The results showed that the mean RNFL thickness in patients with OSA was significantly lower than that in healthy controls (MD = -3.56, 95% CI: -4.98 to -2.15, P < 0.00001). Quadrant-based analysis demonstrated significantly reduced RNFL thickness in the temporal (T), nasal (N), superior (S), and inferior (I) quadrants, with MD values of -1.57, -3.04, -3.53, and − 4.66, respectively (all P < 0.001). Stratification by disease severity showed that, in mild OSA, only the T quadrant RNFL thickness was lower than that in healthy controls (MD = -1.09, 95% CI: -2.03 to -0.15, P = 0.02). In contrast, patients with moderate and severe OSA showed significant reductions in average RNFL thickness and in all four quadrants compared with healthy controls. Formal tests for subgroup differences by geographic region showed a significant between-region difference only for S RNFL ( P = 0.02), whereas no significant between-region differences were observed for average, T, N, or I RNFL. In the MR analysis, the forward analysis did not identify a significant causal association between genetically predicted RNFL thickness and OSA risk (IVW: OR = 1.002, 95% CI: 0.988 to 1.017, P = 0.749). In the reverse analysis, the primary IVW estimate was not statistically significant (β = 0.159, SE = 0.237, P = 0.503), whereas the weighted median and weighted mode methods yielded nominally significant positive estimates, indicating inconsistency across MR estimators. OSA was observationally associated with RNFL thinning, and RNFL thinning differed across OSA severity categories, with the T quadrant being the only quadrant showing a significant reduction in mild OSA. Although bidirectional MR analysis did not provide consistent genetic evidence supporting a causal relationship, RNFL thickness may warrant further evaluation as a candidate adjunctive marker for OSA screening and severity assessment. Not applicable.