Measured versus predicted posterior corneal astigmatism in toric IOL planning: a retrospective paired calculator comparison using Pentacam and CASIA2 tomography inputs

To determine how frequently toric intraocular lens (IOL) planning recommendations change when measured posterior corneal astigmatism (PCA) is entered into the Barrett toric calculator instead of its internal prediction, whether this effect differs between Pentacam and CASIA2 cohorts, and how recommendations compare with the Johnson & Johnson (J&J) toric calculator. This retrospective, single-center study included 91 eyes of 91 patients (one eye per patient) undergoing toric IOL implantation (TECNIS DIU platform), assigned to two non-overlapping tomography cohorts (59 CASIA2, 32 Pentacam). For each eye, planning outputs were reconstructed uniformly in March 2026 using identical biometry (IOLMaster 700) and anterior keratometry across three scenarios: Barrett with predicted PCA (B-PRED), Barrett with measured PCA (B-MEAS), and J&J. A prespecified planning discrepancy was defined a priori as a toric step difference ≥ 1 and/or an axis difference ≥ 10°. An all-eyes sensitivity analysis (145 eyes) used generalized estimating equations. A planning discrepancy between B-MEAS and B-PRED occurred in 29.7% (27/91; 95% CI: 21.3–39.7%), driven by step differences (23.1%) and axis disagreement (12.1%). B-MEAS predominantly recommended a lower toric step (20.9% versus 2.2% higher). The discrepancy rate between J&J and B-MEAS – a combined calculator/PCA-workflow contrast – was 59.3% (54/91), with J&J systematically recommending a higher step (52.7%). The pure inter-calculator contrast (J&J versus B-PRED) yielded 45.1%, significantly exceeding the PCA-handling contrast (McNemar p = 0.049); switching the Barrett reference to measured PCA added a further increment (to 59.3%; p = 0.001). No statistically significant between-cohort differences were detected ( p = 0.481 and p = 1.000), although only differences of approximately 27–30% points were detectable. In exploratory analysis, anterior astigmatism magnitude, PCA magnitude (OR = 1.73 per 0.1 D), and against-the-rule class were associated with the J&J versus B-MEAS discrepancy (all p ≤ 0.002). Entering measured PCA into Barrett Toric changes the recommendation in approximately 30% of eyes, predominantly reducing the cylinder step; calculator choice contributes at least as much planning variability. No significant between-cohort differences were detected, but the design permits no equivalence claim. As planning-level findings without outcome validation, these results warrant standardized documentation of calculator settings rather than a preference for any pathway.