ObjectiveTo explore features associated with impaired retinal neurovascular coupling among individuals diagnosed with Alzheimer’s disease (AD) and its association with alterations in plasma metabolomics, and to establish an integrated diagnostic model for evaluating its possible significance regarding early-stage identification of AD.MethodsWe prospectively enrolled 85 AD patients, 82 subjects with mild cognitive impairment (MCI), and 83 cognitively normal controls. Retinal vessel density, foveal avascular zone (FAZ) parameters, choriocapillaris flow deficits (CCFDs), ganglion cell layer thickness, and neurovascular coupling index (NVCI) were quantified using optical coherence tomography angiography (OCTA). Untargeted plasma metabolomics was performed via ultra-high-performance liquid chromatography–quadrupole time-of-flight mass spectrometry. Differential metabolites and pathways were identified, followed by correlation, regression, and receiver operating characteristic (ROC) analyses.ResultsCompared with controls, the AD group showed reduced superficial vascular plexus perfusion density (−7.1%, P < 0.001), enlarged FAZ area (+16.1%, P < 0.001), increased CCFDs (+28.3%, P < 0.001), decreased ganglion cell layer thickness (−18.3%, P < 0.001), and lower NVCI (−12.7%, P < 0.001). Values in the MCI group were intermediate. In metabolomic profiling, AD patients had lower leucine (−17.0%, P < 0.001) and phosphatidylcholine PC(16:0/20:4) (−19.4%, P < 0.001), but higher ceramide d18:1/24:1 (+33.6%, P < 0.001) and ceramide/sphingomyelin ratio (+22.2%, P < 0.001). NVCI positively correlated with leucine (r = 0.205, q = 0.0011) and negatively with ceramide d18:1/24:1 (r = −0.139, q = 0.0280). An integrated model combining FAZ area, CCFDs, NVCI, leucine, ceramide d18:1/24:1, and PC(16:0/20:4) discriminated AD from controls with an AUC of 0.908 (95% CI: 0.865–0.952), 83.5% sensitivity, and 88.0% specificity. For predicting MCI-to-AD progression, the AUC was 0.744 (95% CI: 0.669-0.819), with 57.6% sensitivity and 85.4% specificity.ConclusionRetinal neurovascular coupling dysfunction was significantly associated with systemic metabolic disturbances (branched-chain amino acid deficiency and ceramide accumulation) in patients with AD. The OCTA-metabolomics integrated model offers excellent diagnostic performance, supporting its potential for non-invasive early AD detection and risk assessment.