Effect of anterior chamber depth on the choice of intraocular lens calculation formula in patients with normal axial length.

PURPOSE: To compare the accuracy of  Sanders-Retzlaff-Kraff II (SRK II) and 3(rd) and 4(th) generation intraocular lens (IOL) formulas and to compare the effect of different anterior chamber depths among the IOL formulas in cataract patients with normal axial length (AL; 22.0-24.5 millimeters, mm).
MATERIALS AND METHODS: A retrospective chart review was performed of patients with normal AL who underwent cataract surgery. The SRK II and 3(rd) generation IOL formulas (Hoffer Q, SRK T, Holladay 1) were compared to the 4(th) generation Haigis formula. For analysis, preoperative anterior chamber depth (ACD) was divided into three subgroups: ≤3, 3-3.5, and ≥ 3.5 mm. The mean error (ME) and mean absolute error (MAE) of each formula was compared for each subgroup against the total. The difference between the ME and MAE of the formulas were compared for each ACD subgroup. P < 0.05 was considered statistically significant.
RESULTS: The study sample was comprised of 309 eyes. The MAE were 0.56 D, 0.52 D, 0.51 D, 0.50 D, and 0.50 D with Haigis, Hoffer Q, SRK T, Holladay 1, and SRK II formulas, respectively. The Haigis formula was a significantly weaker predictor than the SRK T (P < 0.001) and Holladay 1 (P = 0.035) formulas. The ME with Haigis formula was -0.23 D which changed to -0.06 D with ACD ≥ 3.5 mm (P = 0.002). The ME was -0.24 D with SRK II and changed to -0.09D with ACD ≤ 3 mm (P = 0.004). There was no statistically significant difference between Hoffer Q, SRK T, and Holladay formulas 1 in ACD subgroups (P > 0.05, all comparisons).
CONCLUSION: The SRK II formula can predict refraction in patients with normal AL and ACD less than 3 mm with less error and is preferred over other formulas. The Haigis formula is the preferred choice in patients with a normal AL and ACD longer than 3.5 mm. The prediction accuracy of Hoffer Q, SRK T, and Holladay 1 is comparable in normal AL.

INTRODUCTION

Intraocular lens (IOL) power calculation formulas have evolved over the past 30 years to improve the refractive outcome of modern cataract surgery. The efficacy of IOL implantation depends on the accuracy of ocular biometric measurements and the accuracy of IOL power calculation formulas.12 Ocular biometric data used in these formulas include the axial length (AL), corneal power, and anterior chamber depth (ACD).34 Studies have reported that every 1 mm deviation of the corneal diameter, AL, and ACD can result in 5.7 D, 2.7 D, and 1.5 D of refractive error, respectively.5 Today, the accuracy of AL measurements has increased by using the IOL Master;6 thus, the AL contributes to residual refractive error a lot less than ACD. The reported contribution to error from ACD, AL, and corneal power is 42, 36, and 22%, respectively. Therefore, one of the main causes of residual refractive error with IOL formulas is neglecting the role of the ACD.5

Studies have focused on refining IOL calculation formulas from the 1st generation to the 4th generation to improve their accuracy.3789 The SRK II, a 2nd generation formula, is still favored by clinicians and is commonly used in many countries10 including our country, Iran, and Singapore.11 The Hoffer Q, SRK T, and Holladay 1 are 3rd generation formulas which consider AL and corneal height to determine the effective lens position (ELP).5 The Haigis formula is one of the 4th generation formulas which considers the preoperative ACD, in addition to AL, to predict ELP.5

As SRK II is still used in many clinical settings in our country, we designed this study to examine the accuracy and refractive prediction error of SRK II, Hoffer Q, SRK T, Holladay 1, and Haigis in patients who were candidates for cataract surgery and had normal axial length (22.0-24.5 millimeters, mm). We specifically evaluated the role of ACD to determine why the SRK II formula performs well and is preferred by many clinicians.

MATERIALS AND METHODS

A retrospective chart review was performed of patients who underwent cataract surgery with IOL implantation at our center from 2010 to 2011. The surgeries were performed by one surgeon (HH). Inclusion criteria were an AL between 22.0 and 24.5 mm (normal range), availability of preoperative AL, corneal power, preoperative ACD data, 1-month postoperative refraction data, and uncomplicated surgery. This study followed the tenets of the Declaration of Helsinki.

AL, corneal power, and ACD were measured with IOL Master version 5.4 (Zeiss AG, Jena, Germany). All eyes underwent sutureless phacoemulsification under topical anesthesia with a 2.8 mm temporal clear corneal incision. All eyes underwent Acrysof SA60AT lens (Alcon Inc., Fort Worth, TX, USA) implantation and an optimized A-constant of 119. All eyes were targeted for emmetropia.

For analysis, preoperative anterior chamber depth (ACD) was divided into three subgroups: ≤3, 3-3.5, and ≥3.5 mm.

The main outcome measures of this study were the mean error (ME) calculated as the arithmetic mean deviation from the predicted postoperative refractive outcome, and the mean absolute error (MAE) calculated as the absolute mean deviation from the predicted postoperative refractive outcome. To compare errors with these formulas, we used the repeated measure analysis of variance (ANOVA) of the ME and MAE. Post-hoc analysis was used for multiple comparisons among formulas. The Pearson correlation coefficient was used to determine the correlation of ACD with the ME and MAE, one-way ANOVA and post-hoc to compare the ME and MAE of each formula in the different ACD subgroups. The one-sample test was used to compare the errors of each formula in the different ACD subgroups with the total sample. In all analyses, the level of significance was set at 0.05.

RESULTS

Based on the inclusion criteria, data from 309 eyes was extracted for analysis and 50.8% of the samples were right eyes. The mean age of the participants was 67.19 ± 14.81 years. The mean ACD was 3.19 ± 0.41 mm (range, 2.01-4.05 mm), the mean AL was 23.11 ± 0.63 mm (range, 22.02-24.47 mm), and the mean corneal power 44.73 ± 1.62 D (range, 39.57-49.87 D). The mean postoperative manifest refraction spherical equivalent (MRSE) was -0.29 ± 0.58 D.

The ME, MAE, and MAE percentages with each formula are summarized in Table 1. Compared to the Haigis formula (which had the highest MAE), SRK T (P < 0.001) and Holladay 1 (P = 0.035) had the least error. The differences in MAE between other formulas were not statistically significant (P > 0.05, all comparisons). The inter formula differences in the percentage of the MAE of ≤ 0.25D, ≤1.0D, and ≤ 2.0D were not statistically significant (P > 0.05, all comparisons). However, at the ≤ 0.5 D level of error, predictions with Haigis were significantly worse than with the Hoffer Q (P = 0.034), SRK T (P = 0.007), and Holladay 1 (P = 0.032) formulas, whereas the predictions with SRK II were not significantly different from other formulas (P > 0.05, all comparisons).

Table 1

The inter formula mean error and mean absolute error differences in 309 eyes that underwent cataract surgery with intraocular lens implantation

ACD was inversely correlated with the MAE for all formulas. The inter formula differences in the MAE were not statistically significant for all subgroups of ACD [Table 2]. Stated differently, taking ACD into account had no effect on the MAE, and errors in different ACD levels were not significantly different from those in the total sample.

Table 2

The inter formula mean error and mean absolute error differences in the category of anterior chamber depth in 309 eyes that underwent cataract surgery with intraocular lens implantation

As the MAE does not demonstrate the direction of undercorrection or overcorrection, we performed analyses with the ME. The SRK II formula showed similar under- and over- estimation predictions in ACD ≤ 3 mm. Similarly, the Haigis formula had similar under- and overestimation for ACD ≥ 3.5 mm [Table 3]. In other words, the prediction with SRK II was closer to emmetropia at lower ACD readings and prediction with Haigis was closer to emmetropia at higher ACD readings. The inter formula ME differences were statistically significant at ACD levels ≤3.0 mm [Table 2]. Post-hoc analysis showed that the ME with SRK II differed significantly from Hoffer Q (P < 0.001), SRK T (P < 0.001), and Holladay 1 (P < 0.001) formulas. In this ACD category, Haigis and SRK II did not differ significantly in terms of ME, although SRK II predictions were closer to emmetropia when compared to Haigis (P > 0.05). Inter formula differences in the ME were not statistically significant in the ACD 3.0-3.5 mm subgroup (P > 0.05). At ACD ≥ 3.5 mm, the Haigis ME differed significantly from SRK II (P = 0.025), Hoffer Q (P = 0.049), SRK T (P = 0.021), and Holladay 1 (P = 0.035).

Table 3

The percentage of the mean error of the formulae in the category of anterior chamber depth in 309 eyes that underwent cataract surgery with intraocular lens implantation

To examine the effect of ACD, the ME of each formula was compared with the ME for each subgroup of ACD. We found a statistically significant difference between SRK II ME in the total sample and in the ACD ≤ 3.0 mm subgroup (-0.24 and -0.09D, respectively; P < 0.001). There was also a statistically significant difference between the Haigis ME in the total sample and in the ACD > 3.5mm category (-0.23 D and -0.06 D, respectively; P < 0.001). Other comparisons within these two formulas or the other 3 formulas showed no statistically significant differences.

DISCUSSION

The outcomes of the present study indicate that for normal AL (22.0-24.5 mm), Haigis is less accurate than SRK T and Holladay 1 formulas. There were no statistically significant differences between errors with SRK T, Holladay 1, Hoffer Q, and SRK II formulas and their accuracies were comparable. We considered preoperative ACD, and the predictions changed for these formulas. At ACDs higher than 3.5 mm, predictions with Haigis were closest to emmetropia, and at ACDs less than 3 mm, SRK II was closest to emmetropia. We know that there is a reduction in ELP as preoperative ACD decreases, and lower IOL power is needed. Hence, it can be concluded that the tendency for underestimation observed with SRK II, which usually suggests lower IOL powers compared to other formulas, compensates for this error and results in predicted refraction that is closer to emmetropia.

According to two major population-based studies in Iran, the mean ACD in the population over 40 years old is less than 3.0 mm [between 2.50 mm and 2.69 mm with the Orbscan (Bausch and Lomb Inc., Rochester, NY, USA)11 and 2.69 mm with the Pentacam (OCULUS Optikgerδte GmbH, Wetzlar, Germany)]12, and it appears that in Iranian cataract surgery candidates with normal AL, predictions with SRK II are accurate as well as formulas. Notably, SRK II is a 2nd generation formula and its predictions have been reported to be less reliable than other formulas.13 Hoffer Q, SRK T and Holladay 1st are 3rd generation formulas and Haigis is a 4th generation formula that takes ACD into account. However, SRK II is the best choice in certain patients (AL 22.0-24.5, and ACD ≤ 3.0).

Comparing Holladay 1, Hoffer Q, and SRK T formulas in cases with AL between 22.0 mm and 24.5 mm, Hoffer14 suggested Holladay 1 and Hoffer Q are the optimal formulas. In a study by Narvaez et al.,15 the accuracy of these three formulas were comparable. In a study by Aristodemou et al.,3 the MAE with different formulas was similar for AL of 22.0-23.5 mm while Holladay 1 had slightly better predictions than other formulas for AL 23.5-24.5 mm. A study by Elder16 reported that that SRK T formula had less than 0.5 D error for the MZ30BD IOL (Alcon Inc., Fort Worth, TX, USA) in 56% of the cases. In our study, this value was 62.4% for the SA60AT IOL. Based on these outcomes we believe that predictions with Holladay 1, Hoffer Q, and SRK T formulas are quite similar in eyes with normal AL, and there are no clinically significant differences among the formulas. The minor differences between studies can be due to differences in the type of IOL, differences in devices measuring biometric components including AL, differences in the number and expertise of surgeons, as well as differences in the sample size and the power of the study. However, a very important cause of differences among study results mentioned here is the variation in anatomic characteristics and ocular parameters in different populations. The normal range of ACD in a given population can help with the choice of IOL calculation formula, and could differ among different populations. As demonstrated in this study and based on our clinical experience, SRK II is the suggested formula for patients with normal AL and ACD ≤ 3.0 mm because its tendency for underestimation of the IOL power leads to the least deviation from emmetropia.

In our study, the Haigis formula had 0.56 D ME in the total sample and its predictions were significantly less accurate than SRK T and Holladay 1 formulas. This formula showed no difference with 3rd generation formulas such as SRK T, Holladay 1, and Hoffer Q at ACD 3.0-3.5 mm. However, in patients whose ACD is at least 3.5 mm, the calculation IOL power from the Haigis formula was close to emmetropia. In European populations, the mean ACD has been reported at least 1.0 mm higher than our study sample.17181920 In contrast, in a Singaporean population, where the mean ACD (3.08 mm)21 is lower than the European and American populations, the accuracy of SRK II in the prediction of refractive results is good and it is routinely used in practice.11 This observation could explain better outcomes from 4th generation formulas such as the Haigis formula in European patients. Our results and the findings from other studies are presented in Table 4.

Table 4

The accuracy of intra-ocular lens power calculation formula reported in different studies

In conclusion, differences in ocular biometric characteristics cause differences in accuracy of IOL formulas. These differences allow different IOL formulas to be used for different biometric characteristics. In populations which have higher mean ACDs, 4th generation formulas may provide better and more accurate predictions. However, in populations such as the Iranian population where the mean ACD is lower, SRK II can still be used as a low error formula.