I read with interest the recently published paper by Carifi and associates regarding formula accuracy in very small eyes (<21 mm) and some questions arose. Why is the “target refraction” mentioned at all? It is of absolutely no consequence when evaluating the prediction error of intraocular lens (IOL) power calculation. If you target for −10.00 diopter (D) and the formula recommends a 35.0 D IOL predicting a postoperative refraction of -10.25 and the postoperative refraction is −9.88 D, then the prediction error is 0.37 D [(−10.25)−(−9.88)], not 0.12 D (from target). The only comparison should be the difference between the refractive error predicted by the formula (not the surgeon’s target) and the actual postoperative refraction of the patient.
The authors report that the prediction error leads to patients being slightly myopic with the Hoffer Q formula without mentioning that state’s being far superior to their being hyperopic. The only formula referenced is the Hoffer Q (reference 17), but it is missing its 2 crucial errata noted below. The other formulas studied should have had references.
The authors state that “…a trend toward less accuracy of the prediction was observed with the SRK/T and the Holladay 1…” but do not reference that this was first published in their reference 17. The authors state that the obsolete, error-prone SRK II regression formula was included as a control. This formula should not have even been included, since it is inaccurate, is obsolete, and is certainly not a control.
The authors do point out that the Food and Drug Administration (FDA)-approved labeled diopter tolerance is a very high ±1.00 D in IOLs >30 D, as were used in their study. They suggest exact diopter labeling of IOLs but do not reference the first such FDA study on such lenses. The authors fail to mention that their reference 18 already showed a tendency that the Holladay 2 formula equals the Hoffer Q in <22 mm short eyes and it also noted that the Holladay 2 requires the collection of much more data to achieve that accuracy, whereas the Hoffer Q only depends on axial length and corneal power.
Finally, what was the need for a study of only 28 short eyes after the 2011 publication of the very large study using optical biometry by Aristodemou and associates (reference 19) proved statistically that the Hoffer Q was superior in short eyes? There are not enough eyes in this study to prove any statistical significance in the difference of prediction error. Adding the Haigis and Holladay 2 formulas is an addition to that study, but the Haigis formula must be carefully triple optimized to obtain a mean error of zero and this was not done in this study, which is unfair to the Haigis formula.