PURPOSE: To compare the predictive accuracy of four optical modelling scenarios for postoperative refraction after myopic Implantable Collamer Lens (ICL) implantation using patient-specific clinical biometry and to validate the results against clinical postoperative refraction and vault data. METHODS: This retrospective study analysed 215 eyes following myopic ICL implantation. Four optical vergence models were implemented using patient-specific preoperative biometry and measured postoperative vault. The scenarios differed in their treatment of corneal and phakic intraocular lens (pIOL) thickness: (A) thick cornea and thick pIOL; (B) thin cornea and thick pIOL; (C) thick cornea and thin pIOL; (D) thin cornea and thin pIOL. Postoperative spherical equivalent refraction was defined by objective autorefraction, with a sensitivity analysis for the subjective refraction in a subset of eyes. Prediction accuracy was assessed using mean error, mean absolute error, refractive accuracy thresholds, and Bland-Altman analysis. Two-dimensional dispersion was evaluated using standard distance deviation and standard deviational ellipses (SDE). Ellipse-aligned variance was compared using Brown-Forsythe tests with Bonferroni correction. The prediction accuracy of all scenarios were contrasted against the standard. RESULTS: Scenarios assuming a centrally thick pIOL (A and B) systematically overestimated postoperative refraction, with biases exceeding + 1.0 D and wide limits of agreement (> ± 3.0 D). Thin-pIOL models demonstrated substantially improved agreement. Scenario C showed the lowest bias (0.30 ± 0.48 D), narrowest limits of agreement (- 0.62-1.21 D), smallest SDE area, and significantly reduced variance along both ellipse-aligned axes compared with Scenarios A and B (p < 0.001). No significant dispersion difference was observed between Scenarios C and D. Scenario C achieved prediction accuracy that was very close to that of the manufacturer's calculation algorithm. CONCLUSION: For myopic ICLs with central thinning and a central aperture, optical models assuming a thin pIOL provide postoperative refractive prediction. Model selection should prioritize optical relevance along the visual axis rather than anatomical complexity.
Oleszko et al. (Wed,) studied this question.