[Comparison between ray-tracing and IOL calculation formulae of the 3rd generation]

The aim of this study was to compare the different calculation methods in large patient collectives, including eyes with extreme axial lengths.

The prediction errors of the Haigis, SRK/T, Hoffer-Q and Holladay formulae and of the OKULIX ray-tracing are compared in 2888 normal eyes implanted with 8 IOL models. The 5 methods are adjusted to zero mean prediction error for each subcolletive implanted with a particular IOL model, in the formulae by variation of the "formula constants" and in the ray-tracing by adjusting the mean anterior chamber depth. 249 short eyes (mean axial length 21.3 mm) are than compared with the same adjusting parameters. Two collectives from two hospitals with very long eyes (59 eyes with mean axial length of 30.4 mm and 50 eyes with mean axial length of 31.4 mm) and two extremely short eyes (16.7 mm and 16.72 mm) of the same patient are additionally included into the investigation.

In normal eyes, standard deviations of the mean prediction errors ( approximately 0.59 D), mean absolute errors approximately 0.43 D) and median of the absolute error approximately 0.33 D) do not differ between the five methods. The differences increase with the distance from "normal" eyes and are up to 6 D in the extremely short ones.

As long as only axial lengths and corneal radii are used as input parameters, the choice of the calculation method appears not to be relevant in the case of normal eyes, because other errors are dominant. Other than the formulae, the ray-tracing method can be applied to non-normal eyes (extremely short or long ones) without bias induced by the calculation method. In particular, additionally measured data such as topography or spatially resolved corneal thickness can be used, e. g., in eyes after refractive surgery.