We thank Savini and associates for their attention and suggestions for our study.
Regarding the first issue, what we did was exactly what they suggested. As stated in the methods section of our article, the lens power that could yield a target refraction closest to −3.50 diopters (D) postoperatively was selected in cases of high myopia and the lowest possible myopic value chosen as target refraction for controls. The postoperative refractive error was calculated as the difference between the measured postoperative refraction and the predicted postoperative refraction corresponding to the selected IOL power. According to their expression in response, perhaps it was the inconsistent understanding of “targeted refraction” that confused them. Here, the targeted postoperative refraction referred to the refraction predicted by the formula rather than the surgeon’s target of −3.50D or the “target refraction” displayed on the top right corner of the IOLMaster report. Thus technically, the postoperative refractive error in our study was exactly the prediction error they want to know, though in different words. Similar usage of “target (ed) refraction” could be seen in various studies.
With regard to the second issue, as suggested, we conducted the constant optimization using the program included in the Zeiss IOLMaster 500 this time. Because only 1 patient in this study was implanted with a negative-diopter IOL, we took it out to avoid the possible confounding factors in further analysis.
The new constants acquired were 120.2 for SRK/T formula and 3.48 for Holladay 1 formula (SF). The mean arithmetic prediction error reduced from 0.57 ± 0.73 to 0.39 ± 0.75 for SRK/T formula and from 1.13 ± 0.76 to 0.44 ± 1.00 for Holladay 1 formula (both P < .001). Of note, with new constants, the prediction errors of those with previously severe hyperopic shift decreased while those with relatively accurate IOL prediction tended to have greater myopic shift. Thus, the significant reduction toward zero was perhaps the compromise between decreased hyperopic shift and increased myopic shift after using new constants. For individuals, the prediction errors still existed.
In addition, with the optimized constants, our data reanalysis of patients with positive IOL still revealed a statistically significant correlation between fixation stability and prediction error (SRK/T formula: 63% BCEA r = 0.381, P < .001, 95% BCEA r = 0.377, P < .001; Holladay 1 formula: 63% BCEA r = 0.393, P < .001, 95% BCEA r = 0.389, P < .001; Pearson correlation analysis) and that 63% BCEA was still an independent influencing factor of prediction error (SRK/T formula: β = 0.173, P = .041; Holladay 1 formula: β = 0.146, P = .028; backward stepwise multiple linear regression analysis).
Hence, the reliability of our conclusion did not change after constant optimization. And fixation stability should be taken into consideration to minimize the deviation from target refraction postoperatively in the future.
Currently, we are in the process of conducting a prospective study to corroborate this hypothesis and we will try to optimize the constants for negative-diopter IOLs for high myopic eyes in the future.
Finally, we appreciate and respect the suggestions from Savini and associates, which help make our study more rigorous.