Agreement of keratometry measurements for toric IOL planning in routine clinical practice: KR-800, IOLMaster 500, and Pentacam HR.

Accurate keratometry is essential for toric intraocular lens (IOL) planning, yet most prior studies evaluated inter-device agreement under standardized research conditions that may not reflect everyday practice. This retrospective study evaluated keratometry agreement among the KR-800, IOLMaster 500, and Pentacam HR under real-world clinical conditions in 227 eyes of cataract surgery candidates. Corneal astigmatism was converted to power vector components: J0 (with/against-the-rule astigmatism) and J45 (oblique astigmatism). Population-level agreement was assessed using linear mixed-effects models and intraclass correlation coefficients (ICC), while individual-level agreement was evaluated using Bland-Altman analysis. Agreement in toric IOL candidacy and the predicted residual astigmatism arising from inter-device differences were also examined. No significant population-level differences were observed among devices (J0: p = 0.138; J45: p = 0.233), with ICC values of 0.894 for J0 and 0.748 for J45, respectively. However, the 95% limits of agreement were wide (0.80-1.21 D). When any device pair was considered, the predicted residual astigmatism exceeded 0.50 D in 64.8% of patients, and the Pentacam HR classified fewer eyes as toric IOL candidates than the reflection-based devices. Among toric IOL candidates (≥1.0 D astigmatism), 21-33% demonstrated axis differences exceeding 10°, a threshold associated with clinically relevant reduction in astigmatic correction. Axis differences were inversely correlated with astigmatism magnitude (Spearman's rho = -0.220 to -0.272, all p-values < .001). The three devices showed acceptable agreement at the population level, and notably, the average magnitude of disagreement was limited. At the individual level, however, differences exceeded clinically relevant thresholds in a meaningful proportion of patients, including some toric IOL candidates. Using a single device consistently throughout surgical planning may be prudent, although the clinical impact of these differences was not directly assessed and warrants prospective validation.