Automated Videotopography for Dry Eye Diagnostics: Analytical Performance, Spatial Dynamics, and a Novel Multivariate Model

This study quantified the analytical performance of an automated videotopography system for dry eye disease (DED) and evaluated a novel multivariate composite score to improve diagnostic accuracy. In a prospective, repeated-measures study, 35 adults completed three visits involving automated (Keratograph 5M) and manual (fluorescein break-up time; slit-lamp meniscus height) assessments. Data were analysed using linear mixed-effects models and Bland–Altman plots, while a logistic regression-based Objective Symptom Risk Score (Objective-SRS) was derived to predict symptom status. Results showed that automated meniscus height (NIKTMH) had excellent precision (CV 8.8%) and reliability (ICC 0.727), whereas non-invasive break-up times were highly variable (CV > 40%). Automated and manual measures demonstrated wide limits of agreement and systematic bias, precluding interchangeability. While individual objective tests failed to differentiate symptom groups, the composite Objective-SRS achieved good accuracy (AUC 0.768) and superior net clinical benefit. The study concludes that diagnostically useful information is distributed across multiple signals; thus, automated and manual measures should be used complementarily, with multivariate models offering superior discrimination of DED symptom burden.