Establishing the ground truth for keratoconus progression: combining repeated measures and adapting precision limits to disease severity in tomography

Purpose:To analyze measurement precision using a contemporary tomography device (MS-39) in a large sample of patients with keratoconus.Setting:Moorfields Eye Hospital, London, United Kingdom.Design:Prospective and retrospective analyses of clinical data.Methods:For a range of key corneal tomographic indices, repeatability and reproducibility were compared in a prospective series of 1248 scans (n = 208 eyes) with 2 observers each acquiring 3 scans per eye. Measurement repeatability for single measures vs the mean of 3 consecutive measures was then compared using the same dataset. 95% repeatability coefficients were mapped at each unit interval through the measurement range in a retrospective analysis of 13 734 scans (n = 4578 eyes) to adapt progression thresholds to disease severity. Finally, keratoconus classification using alternate progression definitions was compared at 1 and 2-year time points.Results:For all corneal tomographic indices examined, no significant differences were found between MS-39 measurement repeatability and reproducibility (P >.05), repeatability was (average) 44% improved using the mean of 3 consecutive measures, and measurement repeatability deteriorated as keratoconus advanced. Progression was classified incorrectly in over 50% of cases at both time points where adaptive thresholds and repeated measures were not used.Conclusions:MS-39 measurement precision was not significantly influenced by the operator. Consecutive repeated measures and adaptive thresholds for observed change in key tomographic parameters based on device-specific precision limits tailored to disease severity are necessary for accurate classification of progression.