We read with interest the article “Accuracy of intraocular lens power calculation formulae in children less than two years.” We thank the authors for referring to our paper. The authors retrospectively analyzed 128 eyes of 84 children less than 2 years of age who underwent primary intraocular lens (IOL) implantation for congenital cataract. Surprisingly, they found that SRKII was the most predictable formula in their series.

The most striking observation about this study is the magnitude of the calculated prediction errors, up to 14.3 diopters (D), with large absolute prediction error for all the formulae: SRKII, 2.27 ± 1.69 D; SRK/T, 3.23 ± 2.24 D; Holladay 1, 3.62 ± 2.42 D; HofferQ, 4.61 ± 4.12 D. In contrast, we observed that SRKII was the least predictable formula (mean absolute prediction error, n = 135: SRKII, 1.11 ± 0.94 D; SRK/T, 0.84 ± 0.71 D; Holladay 1, 0.76 ± 0.64 D; HofferQ, 0.76 ± 0.61 D). While our study was not powered to compare eyes with axial length (AL) <20 mm, the mean absolute prediction errors observed for this group (n = 12) were: SRKII, 1.82 ± 1.00 D; SRK/T, 0.65 ± 0.53 D; Holladay 1, 0.42 ± 0.39 D; HofferQ, 0.73 ± 0.51 D. Absolute prediction error >2 D was found in only 2 eyes using all formulae, suggesting other sources of measurement error in these cases.

While our study had few eyes with AL <20 mm, the Infant Aphakia Treatment Study (IATS) reported prediction errors in a group of 49 infant eyes in which the mean AL was 18.1 ± 1.4 mm. While results comparing formulae have not been reported, the absolute prediction error using Holladay 1 was 1.8 ± 1.3 D, which is half that reported in the present study.

Additionally surprising is the finding of negative mean prediction error, suggesting an undercorrection, or more residual hyperopia than expected, for all formulae. This was particularly noted when using formulae designed for shorter eyes (Holladay 1, −3.39 ± 2.74 D and HofferQ, −4.39 ± 3.42 D). For eyes <22 mm we observed less, not more, residual hyperopia by using Holladay 1 or HofferQ. For IATS pseudophakic eyes with AL ≥18 mm the mean prediction error was minimal (−0.1 ± 1.6 D), and for AL <18 mm (1.8 ± 2.0 D) using Holladay 1 there was a trend for overcorrection, or less residual hyperopia than expected.

Prediction error is expected to be higher in young children and shorter eyes; however, it seems too large in this series. The large prediction error questions the reliability of various measurements, and it seems that other error sources outweigh any error introduced by formula error. The authors note that multiple surgeons performed the surgeries and IOL power calculations and used different IOL types. Contact ultrasound was performed by various individuals, rather than immersion ultrasound, which may introduce more error. Multiple optometrists performed the refractions, and retinoscopy was performed after sutures were removed at 4 weeks, but the range of time during which refraction occurred is unclear. Since eye growth occurs rapidly in the first year of life, delays in postoperative refractions may less accurately reflect prediction error because of eye growth and myopic shift. Further, the authors do not comment on IOL power implanted, and high IOL power can magnify prediction error.

While IOL power calculation remains challenging in pediatric eyes and many factors contribute to error, we would not advise changing the formula selection to favor SRKII based on this report.