Prediction accuracy of preinstalled formulas on 2 optical biometers

To evaluate how well partial coherence interferometry (PCI) (IOLMaster) and optical low-coherence reflectometry (OLCR) (Lenstar LS 900) predict postoperative refractions using only the formulas that come preinstalled on the machines.

Private practice, Saint Joseph, Michigan, USA.

Retrospective consecutive case series.

Eyes were measured with 2 biometers before cataract surgery. Six formulas were ranked by machine. Formulas were also ranked for extremely long and short eyes by averaging the ranks of 6 statistics (mean error, mean absolute error, standard deviation [SD], maximum error, and percentage of eyes within ±0.5 diopter [D] and ±1.0 D of prediction).

Formulas were ranked by the SD of the prediction errors. The OLCR device outperformed the PCI device using the preinstalled formulas. The Olsen formula performed the best (0.378) but was preinstalled on the OLCR device only. Other formulas had the following SDs (OLCR device first, PCI device second): Haigis (0.393, 0.401), Holladay 1 (0.408, 0.414), Hoffer Q (0.428, 0.432), SRK/T (0.433, 0.440), and SRK II (0.623, 0.633). Rankings for long eye were (first to last) were Olsen, Haigis, SRK/T, Hoffer Q, and Holladay 1. Rankings for short eyes were Olsen, Haigis, Holladay 1, SRK/T, and Hoffer Q.

The OLCR device outperformed the PCI device using the Olsen formula. The Olsen formula also ranked first for short eyes and long eyes. Other formulas performed about the same on both machines. The SRK II formula should be avoided.