Comparability and repeatability of corneal astigmatism measurements using different measurement technologies

PURPOSE: To determine the comparability and repeatability of corneal astigmatism measurements obtained with different devices and determine the interobserver variability of a new automated keratonneter. SETTING: University Eye Clinic Maastricht, the Netherlands. DESIGN: Prospective cohort study. METHODS: The right eye of healthy subjects was examined with the following 6 devices: IOLMaster (automated keratometry), Lenstar (automated keratometry), SMI Reference Unit 3 (automated keratometry), Javal (manual keratometry), KR-1W (corneal topography), and Pentacam (Scheimpflug imaging). An experienced operator obtained 3 repeated measurements. An inexperienced operator obtained additional measurements with the SMI Reference Unit 3. Astigmatism vector analysis was used to determine the comparability, repeatability, and interobserver variability. RESULTS: Corneal astigmatism vectors measured by automated, manual, or simulated keratometry were comparable except for the Pentacam equivalent keratometry (K) (P <.001, repeated-measures analysis of variance [ANOVA]). The mean difference between the equivalent K and other K values was 0.18 to 0.29 diopter (D) (P <.05, Hotel ling trace multivariate ANOVA). The mean differences between automated, manual, and simulated keratometry were small (<= 0.12 D). The withinsubject standard deviation ranged from 0.05 D @ 21 degrees (KR-1W) to 0.18 D @ 23 degrees (Lenstar). The SMI Reference Unit showed small mean differences and comparable repeatability between the experienced operator and the inexperienced operator. CONCLUSIONS: Vector analysis showed comparable corneal astigmatism measurements using automated, manual, and simulated keratometry. Pentacam equivalent K values were not comparable with those of the other keratometers. The repeatability of astigmatism magnitudes was acceptable; however, the repeatability of astigmatism meridians was moderate. The SMI Reference Unit showed good interobserver variability.