Axial lengths as measured by the Intraocular Lens Master (IOLm) and ultrasound biometry.
Bland-Altman analysis demonstrating that the difference between the predictions in the 2 modalities is not associated with the magnitude of the predictions of the 2 modalities. D indicates diopters; IOLm, Intraocular Lens Master.
Comparison of the prediction accuracies of final refraction made by the Intraocular Lens Master (IOLm) and ultrasound biometry shows that the 2 modalities were highly correlated. The 4 eyes that had a greater than 1-diopter (D) difference in prediction are represented by the data points that fall outside the dotted lines.
Overlap between the patients (129 eyes total) in whom both technologies were unable to predict the final refraction within 1 diopter (D). IOLm indicates Intraocular Lens Master.
Bhatt AB, Schefler AC, Feuer WJ, Yoo SH, Murray TG. Comparison of Predictions Made by the Intraocular Lens Master and Ultrasound Biometry. Arch Ophthalmol. 2008;126(7):929-933. doi:10.1001/archopht.126.7.929
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To ascertain whether a new instrument that uses partial coherence interferometry technology (Intraocular Lens Master [IOLm]) or ultrasound biometry provides a more accurate prediction of refractive outcomes in cataract surgery.
This was a retrospective medical record review of 421 eyes of 304 patients who underwent cataract surgery with the IOLm and ultrasound biometry from January 3, 2002, to December 2, 2005. The mean difference between the prediction with each technology and the final spherical equivalent was compared. Circumstances in which both technologies failed to make accurate predictions were investigated.
The mean (SD) of the difference between predicted refraction and final spherical equivalent was −0.43 (0.84) diopters (D) for the IOLm and −0.60 (0.87) D for ultrasound biometry, indicating that on average the IOLm was a closer predictor than ultrasound biometry of the final spherical equivalent (P < .001). The IOLm had a 5% higher likelihood of predicting a spherical equivalent within 0.25 D than ultrasound biometry (P = .06), an 8% higher likelihood of predicting a spherical equivalent within 0.50 D (P < .001), and an 8% higher likelihood of predicting a spherical equivalent within 1.00 D (P < .001).
To our knowledge, this is the most eyes examined for cataract surgery in a prospective fashion with both the IOLm and ultrasound biometry. The IOLm is a better predictor of postoperative refraction than ultrasound biometry, particularly within close ranges.
The use of traditional ultrasound biometry for intraocular lens (IOL) selection before cataract surgery requires a skilled operator, good corneal surface contact, and significant time expenditure. The introduction of noncontact optical biometry has revolutionized preoperative IOL selection by eliminating these obstacles. One instrument (Intraocular Lens Master [IOLm]; Carl Zeiss Meditec, Jena, Germany), which uses partial coherence interferometry technology, was introduced in 2000. Since then, it has been touted for its fast operation without requiring corneal contact. Several studies1- 8 have been conducted to compare the accuracy of the predictive value of the IOLm with that of traditional axial biometry. Recent publications have reported mixed conclusions about which technology has a better predictive value and in which clinical cases 1 or both technologies will fail to predict postoperative refraction. In a prospective study of 162 consecutive eyes undergoing cataract surgery, Gantenbein and Ruprecht1 concluded that contact axial biometry offered a better prediction of final refraction than the IOLm but that the IOLm was an easier and faster tool to use. In a prospective study of 140 consecutive eyes undergoing cataract surgery, Kutschan and Wiegand2 found that both contact ultrasound biometry and the IOLm were similar in their predictive capabilities and concluded that the IOLm was easier to use. Rajan et al3 found that the use of optical biometry offered a better predictive value than the use of applanation axial biometry measurement. Verhulst and Vrijghem4 and Skorkovska et al5 found that, in eyes with significant nuclear sclerotic cataract, axial biometry was still needed for accurate axial length measurement. More recently, Ueda et al7 reported that axial length measurements taken with the IOLm were slightly affected by the cataract density but to a lesser extent than ultrasound biometry. Two small recent case series6,8 have examined the effect of macular disease on the 2 techniques with a qualified suggestion that the IOLm may be more accurate in these cases, but the conclusions have limited power because of the few patients. In this study, to our knowledge, we evaluated the most eyes ever examined in a prospective fashion with both traditional axial biometry and the IOLm. We sought to compare the accuracy of these methods and to determine in which clinical scenarios patients' final refractions were not predicted accurately by either method.
This was a retrospective medical record review of 421 eyes of 304 patients who underwent cataract surgery at the Bascom Palmer Eye Institute from January 3, 2002, to December 2, 2005. The study was approved by the institutional review board of the University of Miami. All eyes included in this study underwent preoperative measurements of axial length with the IOLm and ultrasound biometry and underwent refraction within 1 to 6 months postoperatively. Eyes that did not fulfill the previously mentioned criteria were not included in this study. All preexisting ophthalmic diseases and postoperative complications were noted. Preoperative axial length measurements were taken on each eye with the IOLm. Ultrasound axial biometry was performed by members of the experienced echography department at the Bascom Palmer Eye Institute. Because of the department's experience with many eyes and the expertise possessed by the echographers in this study, the ultrasound axial length measurement for each eye was primarily obtained using the contact method instead of an immersion method. Immersion ultrasound biometry was used in eyes in which contact biometry axial length measurements were difficult to obtain. After axial length measurement with each method, the SRK-T formula customized for each modality was used to arrive at a prediction for postoperative refraction. The A constants in each SRK-T formula were individualized for the IOL that was chosen by the surgeon. Every available refraction within a period of 1 to 6 months postoperatively was recorded, and the latest available refraction was used as the final refraction in the statistical analysis. The final refractions were converted to spherical equivalents and compared with the preoperative predictions made by the IOLm and ultrasound biometry.
The mean axial lengths (Figure 1) measured by the IOLm and ultrasound biometry were 23.97 and 23.92 mm, respectively. The mean difference between the axial length measured by the 2 modalities was 0.042 diopter (D) (95% confidence interval, 0.026-0.058 D). The mean (SD) of the difference between predicted refraction and final spherical equivalent was −0.43 (0.84) D for the IOLm and −0.60 (0.87) D for ultrasound biometry. This finding demonstrates that, on average, the IOLm was a closer predictor than ultrasound biometry of the final spherical equivalent (P < .001). A Bland-Altman analysis was performed to confirm that the difference between the predictions in the 2 modalities was not associated with the magnitude of the predictions (Figure 2). A comparison of the predictions of final refraction made by the IOLm and ultrasound biometry (Figure 3) shows that the 2 modalities were highly correlated. In fact, the differences between the predictions with both modalities differed by more than 1 D in only 4 eyes. The 4 eyes that had a greater than 1-D difference in prediction are represented by the data points in Figure 3 which fall outside the dotted lines. The ability of each modality to predict the final refraction within a certain range of the final spherical equivalent was also analyzed. The prediction of postoperative refraction with the IOLm was within a difference of 1 D from the final spherical equivalent for 333 eyes (79.1%). Three hundred eyes (71.3%) were within 1 D with predictions made by ultrasound biometry. A total of 191 eyes (45.3%) were within 0.5 D of the final spherical equivalent using the IOLm, and 158 eyes (37.5%) were within this range with ultrasound biometry predictions. Ninety-nine eyes (23.5%) were within a range of 0.25 D of the final spherical equivalent using IOLm predictions, and 79 eyes (18.8%) were within the same range with the ultrasound biometry predictions.
Table 1 and Table 2 demonstrate the accuracy of the IOLm and ultrasound biometry predictions. For the IOLm, in 88 eyes (20.9%), predictions for the final refraction differed from the final spherical equivalent by 1 D or more. Further investigation demonstrated that 24 of the previously mentioned eyes (27.3%) had preexisting macular disease. With ultrasound biometry, 121 eyes (28.7%) had predictions of the final refraction that differed from the final spherical equivalent by 1 D or more. Within the group of 121 eyes that were outside 1 D with ultrasound biometry predictions, 31 (25.6%) had previous macular disease. Seventy-nine eyes differed from the predicted spherical equivalent by more than 1 D using both technologies. Figure 4 demonstrates the correlation between the 2 technologies in eyes in which the predicted refractions differed by 1 D or more from the final spherical equivalent.
Table 1 indicates that with the use of the IOLm 73.9% of patients required more myopic correction after surgery than predicted, whereas only 25.7% of patients required more hyperopic correction than predicted, with the remaining patients exactly equal to the prediction. Table 2 indicates that, with the use of ultrasound biometry, 75.3% of patients required more myopic correction after surgery than predicted, whereas only 23.5% of patients required more hyperopic correction than predicted, with the remaining patients exactly equal to the prediction.
Since its introduction in 2000, the IOLm has created much debate regarding its accuracy and applicability compared with traditional ultrasound biometry. One previously reported limitation of the IOLm has been the presence of macular thickening or irregularity given that the IOLm was calibrated using healthy eyes with normal retinal thickness to match the measurements of traditional A-scan ultrasonography.6,9 Two small recent case series6,8 have examined the effect of macular disease on the 2 techniques with a qualified suggestion that the IOLm may be more accurate in these cases. However, these studies had limited power based on few patients. Our study sought to answer these unresolved questions by evaluating many eyes with both technologies.
In this study, we assessed the mean difference between the preoperative predictions and the final spherical equivalent with each technology, the correlation of the 2 technologies in the preoperative prediction for the same eye, and the ability of each technology to predict the final spherical equivalent within specified ranges. The statistical analysis demonstrated that little difference existed in the average axial length measurements as measured by the 2 technologies (Figure 1). With the use of the SRK-T formula to obtain predictions for postoperative refraction, the IOLm was 0.17 D more accurate than ultrasound biometry in predicting the final spherical equivalent. This difference was statistically significant (P < .001). The IOLm also offers a slightly better prediction of the postoperative refraction than ultrasound biometry within 0.25- , 0.50- , and 1.00-D ranges. In fact, the IOLm had a 5% higher likelihood of predicting a final spherical equivalent within 0.25 D than ultrasound biometry (P = .06). The IOLm had an 8% higher likelihood of predicting a final spherical equivalent within 0.50 D than ultrasound biometry (P < .001) and an 8% higher likelihood of predicting a final spherical equivalent within 1 D (P < .001). Clinically, the difference in prediction ability between the 2 modalities may not be very important.
The groups of eyes that had differences between the prediction and final spherical equivalent of 1 D or greater were assessed for each technology. These groups were correlated between the 2 technologies, as demonstrated by the overlap in Figure 4. This finding suggested that neither modality did a significantly better job at predicting the final spherical equivalent in this group of eyes.
The predictions made by the IOLm and A-scan ultrasonography were highly correlated (statistically associated) as demonstrated in Figure 3. In fact, only 4 eyes had predictions made by the IOLm and ultrasound biometry that differed by more than 1 D from each other. It is theorized that 2 of these eyes differed by this degree because of previous retinal detachments that may have altered the accuracy of the IOLm technology given its measurement of the axial length from the corneal vertex to the retinal pigment epithelium rather than the vitreoretinal interface. A third eye exhibited symptoms of severe dry eye, and a fourth eye had silicone oil. The Bland-Altman analysis of this sample demonstrated that there was no trend toward greater differences between the predictions of the 2 modalities as the magnitude of the predictions increased with each modality.
We closely examined 7 eyes that had differences of more than 2.00 D between the prediction made by both technologies and the final spherical equivalent. These patients are described in Table 3 along with possible explanations of why the predictions were inaccurate in these patients. This subset of eyes includes 2 patients with postoperative cystoid macular edema, 1 patient with immediate corneal decompensation, 1 patient with a history of pars plana vitrectomy for a macula-off retinal detachment, 1 patient with a history of preexisting epiretinal membrane, 1 patient with a history of previous vitrectomy and membrane peel for an epiretinal membrane, and 1 patient with preexisting severe anterior basement membrane epitheliopathy. Patients 4 through 7 in Table 3 had preoperative disease, which interferes with the ability of the IOLm and ultrasound to accurately predict the final spherical equivalent. Patients such as these may require preoperative counseling regarding the uncertainty of accurate IOL selection with either technology.
In conclusion, the IOLm is a slightly better predictor of postoperative refraction than ultrasound biometry. In particular, it is slightly better at predicting postoperative refraction within close ranges (within 1 D) than ultrasound biometry. Using the predictions of the IOLm tends to result in slightly more postoperative hyperopic correction than expected compared with the use of ultrasound biometry predictions. In the patients in whom the final refraction differed by more than 1 D from the prediction with either modality, a great overlap was seen, suggesting that neither technology was able to provide a better prediction than the other. The IOLm may offer the practicing physician a slight advantage over ultrasound biometry because of increasing patient expectations for precise postoperative refraction and the easier and quicker operation compared with ultrasound. However, for the physician who wants to continue using ultrasound biometry, there is no significant disadvantage.
Correspondence: Timothy G. Murray, MD, MBA, Bascom Palmer Eye Institute, Department of Ophthalmology, University of Miami School of Medicine, 900 NW 17th St, Miami, FL 33136 (email@example.com).
Submitted for Publication: November 13, 2007; final revision received December 20, 2007; accepted January 6, 2008.
Financial Disclosure: None reported.
Additional Contributions: We thank the Bascom Palmer Echography Study Group, which consists of Ernesto Bermudez, OD, Maria Bretana, MD, Fiona Ehlies, BSc (Hons), RDMS, Brandy Hayden, BS, Linda Kelley, CRA, COA, RDMS, and Pat Superfine-Rivera, BGS.