Background: Keratometric methodology varies between instruments and the differences may have a clinical impact. We investigated the agreement and reproducibility of six keratometers.Methods: Keratometry was performed on 100 subjects at two separate sessions with IOLMaster 500, Pentacam, OPD scanner, Medmont E300, Javal-Schiøtz and TMS-5. A second observer assessed 30 subjects to determine inter-observer variability. A single individual was assessed on 10 separate sessions to determine intra-observer variability. Data were analysed using coefficient of variation (CV) and intra-class correlation coefficient (ICCC) for intra-observer variation. Inter-observer concordance was evaluated by the ICCC. Bland-Altman plots, Pearson's correlation coefficient and repeated measures analysis of variance were used to assess agreement of data produced by the instruments.Results: OPD scanner and Javal-Schiøtz mean spherical equivalent (MSE) results were systematically different (p < 0.001) from other instruments (flatter and steeper, respectively). J0 /J45 were similar for all instruments (p < 0.05). Bland-Altman comparison plots indicated that Pentacam and IOLMaster demonstrated greatest level of agreement (ICC results MSE = 0.992, J0 = 0.934 and J45 = 0.890). Agreement (ICC) between observers for MSE ranged from 0.955 to 0.995 for all instruments; lower levels of agreement were found for J0 /J45 (0.289 to 0.901). IOLMaster showed greatest correlation and Medmont the lowest. All instruments showed high intra-observer repeatability of MSE (CV 0.1 to 0.3 per cent). The J0 /J45 readings showed greater variability (CV range 8.8 to 57.6 per cent).Conclusion: When considering MSE alone IOLMaster, Pentacam, OPD scan and Medmont may be considered interchangeable; however, assessment of astigmatism shows greater variability between instruments, sessions and observers. [ABSTRACT FROM AUTHOR]