Diagnostic performance and inter-test agreement of optical coherence tomography for posterior vitreous detachment: a systematic review of blinded, within-subject studies

Abstract Background To systematically review the diagnostic accuracy and inter-test agreement of optical coherence tomography (OCT) for detecting posterior vitreous detachment (PVD) in blinded, within-subject studies that use different reference standards. Methods This systematic review followed PRISMA guidelines and was registered in PROSPERO (CRD42024580929). PubMed, Scopus, Embase, ScienceDirect, Web of Science, Google Scholar, and the Cochrane Database were searched from inception through March 2025. Eligible studies were English-language diagnostic accuracy studies evaluating PVD using blinded, within-subject designs, in which any OCT modality served as the index test and was compared with either a non-OCT reference standard or an alternative OCT protocol that differed in modality, scan protocol, diagnostic criteria, or grader type. Index and reference tests were performed on the same eyes within 24 h. Results Five studies (812 patients, 922 eyes) met the inclusion criteria; three evaluated diagnostic performance and two assessed inter-grader agreement. All studies used spectral-domain OCT (SD-OCT). Diagnostic performance varied substantially by imaging protocol and reference standard. High-density imaging protocols that included both the macula and optic nerve head achieved a sensitivity of 96.4%, a specificity of 87.5%, and substantial agreement (κ = 0.84) compared with composite clinical examination. In contrast, sparse macular-only protocols showed poor agreement (κ = 0.14) and markedly lower sensitivity compared with clinical or surgical reference standards. In two inter-grader studies with markedly different designs and populations, high agreement was observed between reader types under the specific imaging conditions evaluated; these findings should not be extrapolated beyond those contexts. Two studies were judged to be at low risk of bias, whereas three were judged to be at high risk of bias due to unclear masking procedures. Conclusion The diagnostic performance of SD-OCT for PVD is strongly influenced by scan density, field coverage, and the reference standard used for comparison. In the single study using a high-density, wide-field protocol, diagnostic accuracy was excellent, whereas sparse macular-only acquisitions may fail to detect residual vitreoretinal adhesions. Although inter-grader agreement was generally high, the available evidence remains limited and heterogeneous. Further studies using standardized diagnostic criteria and contemporary swept-source OCT platforms are warranted.