Accuracy of six contemporary toric formulas: Analysis of 6,753 Eyes

To compare the predictive accuracy of four toric IOL formulas (Barrett, EVO, Hoffer ® QST, Kane) integrated in the ESCRS calculator. Barrett and EVO were tested with measured (Barrett M and EVO M) and predicted (Barrett P and EVO P) posterior corneal curvature.Design: retrospective database studySetting: multiple surgical centers in the USA.

We analyzed 6753 eyes implanted with non-toric IOLs. Preoperative biometry, including posterior corneal curvature, was obtained using the IOLMaster 700. Predicted residual astigmatism was calculated via the proprietary websites of the four toric formulas integrated into the ESCRS IOL Power Calculator. Primary endpoints were the mean vector magnitude error and proportions of eyes with vector magnitude prediction errors ≤0.50 D.

Kane demonstrated the lowest mean vector magnitude prediction error (0.537 D), significantly lower than EVO P (0.561 D), EVO M (0.57 D), Barrett M (0.579 D), Barrett P (0.583 D), and Hoffer ® QST (0.622 D) (all p<0.001). This result was consistent in with-the-rule, against-the-rule, and oblique subgroups. Kane also achieved the highest proportion of eyes with prediction error ≤0.50 D (p<0.001). For all formulas, centroid values were significantly different from zero. The Kane formula showed the lowest total variances, significantly lower than all other formulas (p<0.001).

The Kane formula showed statistically superior predictive accuracy with the lowest errors and variances and the highest proportion of eyes within 0.50 D. However, differences among formulas were small, and their clinical relevance for toric IOL outcomes remains uncertain.