Retinal Biomarkers of Cognition in Schizophrenia: A Systematic Review and Meta-Analysis

Individuals with schizophrenia consistently show thinning retinal layers, which can potentially serve as neural biomarkers. The association between retinal morphology and core symptom domains like cognitive functioning, however, has not been systematically synthesized. We hypothesize that deficits in retinal structure are associated with cognitive dysfunction in schizophrenia. This systematic review and meta-analysis looks at studies available from inception to June 9, 2026. Outcome measures include correlation coefficients between cognition scores and retinal thickness values (both macular and peripapillary-total and sector-wise) assessed using optical coherence tomography in patients with schizophrenia and related disorders. We assessed risk of bias with the Joanna Briggs Institute critical appraisal tool. Meta-analysis was conducted using the restricted maximum-likelihood method and Fisher's r-to-z transformation. Data from 16 studies (n=1,475 eyes; 847 patients) were included in a series of 88 pooled analyses. Overall cognitive ability correlated with thickness of the macular-retinal nerve fiber layer (mRNFL) (pooled-r=0.180; p=.002; k=7) and the superior-quadrant of the peripapillary RNFL (pRNFL) (pooled-r=0.214; p=.026; k=7). Attention correlated with mRNFL (pooled-r=0.210; p<.001; k=6), macular-inner plexiform layer (mIPL) (pooled-r=0.142; p=.012; k=8), and pRNFL (pooled-r=0.207; p<.001; k=9) thicknesses. Processing speed correlated with foveal thickness (FT) (pooled-r=0.206; p=.008; k=7), mIPL (pooled-r=0.194; p=.001; k=6), and pRNFL (pooled-r=0.211; p<.001; k=5) thicknesses. Executive function correlated with macular thickness (pooled-r=0.099; p=.018; k=10) and mRNFL (pooled-r=0.159; p=.009; k=6). Age, sex, cardiometabolic comorbidities, and smoking moderated these correlations. We provide evidence suggesting that retinal layer thickness is associated with poorer cognitive functioning in schizophrenia, supporting retinal changes as indicators of disease-related neural processes. The small number of available studies limits these conclusions.