Eyeworld

EW CATARACT 42 December 2017 Presentation spotlight by Stefanie Petrou Binder, MD, EyeWorld Contributing Writer length differences of more than 0.3 mm, K differences of more than 1.0 D, and IOL power differences greater than 1.0 D. These outcomes should go up as red flags, and the measure- ments should be repeated. Mon- ocular screening can identify axial length signal/noise ratios (S/N ratio) of less than 2.0 or if the K reading has a standard deviation in excess of 0.20 D. When these numbers pop up, those measurements should be repeated in patients because those are red flags," he explained. "If we knew everything about the optical system exactly, then Snell's law of refraction and ray tracing would be perfect in their predictions," Dr. Holladay said. "But current accuracy limitations includ- ing axial length, K, and refraction measurement errors can only be improved by improving the tech- nology. Corneal posterior surface measurements are evolving. IOL radii, asphericity, and thicknesses are proprietary. You can't get that from manufacturers for use in ray tracing, and that is something that we should change. Prediction of ELP is not exact, and it is continually evolving. It will never be exact, how- ever, because it is a prediction." According to Dr. Holladay, ray tracing will ultimately provide eye surgeons with the greatest accuracy, once IOL properties are made avail- able and the posterior cornea can be measured reliably. He highlighted screening techniques to identify po- tential errors preoperatively to help direct patient flow. He also thinks it is important that all surgeons personalize their lens constant to get the best results in their patients. EW References 1. Melles RB, et al. Accuracy of intraocular lens calculation formulas. Ophthalmology. 2017 Sep 23. Epub ahead of print. 2. Norrby S. Sources of error in intraocular lens power calculation. J Cataract Refract Surg. 2008;34:368–76. Editors' note: Dr. Holladay has no financial interests related to his comments. Contact information Holladay: holladay@docholladay.com depth, lens thickness, horizontal white-to-white, age, and pre-cataract refraction. The second most relevant source of error according to the study's outcomes is refraction, accounting for 27% of the prediction error. Axial length measurement accounts for 17%, the corneal power accounts for 12%, and the pupil for 8% for a standard deviation of 0.391 D and mean absolute error of 0.31 D of IOL prediction error, the last of which is not usually measured or used in power calculations, according to Dr. Holladay. Using variables to predict IOL power Surgeons have four IOL power selection methods at their disposal, which include different IOL ver- gence formulas for thin and thick lenses, ray tracing, and the neural network. Each method uses a com- bination of variables to predict IOL power as accurately as possible. Thin IOL vergence formulas include the Binkhorst formula, Holladay, and SRK. They require five measurements, including anterior K, refraction, IOL power, ELP, and axial length. Thick IOL vergence formulas include the Holladay 2, Olsen 2, and Barrett 2. Nine variables are required here, which include those used for IOL power predictions in the thin vergence formulas plus information on posterior K and IOL radii and thickness. Ray tracing has huge potential, using the above variables with the addition of topographic information from the front and the back of the cornea, as well as the radii and asphericities of the front and back of the IOL (10+ variables). Neural network works with the five variables (axial length, K, refraction, anterior chamber depth, and IOL power), computing 120 possible combinations between the variables to internally predict ELP precisely. According to Dr. Holladay, screening is a step the surgeon needs to take to improve measurement error. "Data screening identifies measurement error and allows the surgeon to repeat measurements, when necessary. Binocular screening is important especially with axial standard deviation of 0.50 D, 67% of cases are within that (± 0.50 D). The mean absolute error is always 80% of the standard deviation, so if the average standard deviation is 0.50 D, the mean absolute error, on average, is 0.40 D. Good surgical outcomes today show a standard deviation of 0.40 D and a mean absolute error of 0.32 D, which corresponds to 78% of patients achieving within 0.50 D of the predicted refraction. The best I have ever seen is a standard devia- tion of 0.31 D and a mean absolute error of 0.24 D, which is equiva- lent to 90% of 'all comer' patients achieving within 0.50 D." In a recent study, 1 Dr. Holladay and co-researchers investigated the standard deviation associated with IOL power prediction in current surgical practice, incorporating data from 13,301 patients and more than 140 surgeons. They found that the standard deviation in this large col- lective was 0.4439 D and mean abso- lute error was 0.34 D, corresponding to 77% of cases that achieved within 0.50 D predicted refraction. Sources of error Although prediction outcomes nearing 80% on average are very satisfactory for most surgeons, there is some room for improvement to reach 90%, which might be achieved through a better understanding of potential sources of error. A study that investigated the sources of error in IOL power calculations by Sverker Norrby identified 16 variables that contributed to prediction error, the first five of which explain almost 99% of error. 2 The highest source of error is from the prediction of the effective lens position (ELP)— the prediction of where the IOL is ultimately placed, for which precise anatomic eye measurements can greatly reduce error. It accounts for 35% of the prediction error, with a standard deviation of 0.31 D and a mean absolute error of 0.24 D. ELP was first predicted by Binkhorst in the early 1980s using only the measurement of the axial length. Today, seven variables have been identified as influencing the prediction of ELP, among them: axial length, K, anterior chamber IOL power calculations using state-of-the-art methods for normal and complex eyes have reached high levels of reliability, however, certain limitations to accuracy are inherent O ne of the challenges eye surgeons encounter when calculating IOL power is that no single diagnostic modality or formula can be applied to all eyes. Ophthalmic surgeons need to take different approaches to determine IOL strengths for phakic and pseu- dophakic eyes, eyes with astigma- tism, long and short eyes, as well as eyes that have undergone previous refractive surgery. The success and safety of IOL implantation reflects decades of evolution in surgical technique and measurement methods, as well as in understanding of the sources of potential error. Reaching prediction limits "Precise measurements and IOL calculation formulas are essential to achieve the high accuracy refractive outcomes demanded of premium IOLs," said Jack Holladay, MD, MSEE, clinical professor of ophthal- mology, Baylor College of Medicine, Houston, who spoke on the topic at the 2017 ASCRS•ASOA Symposium & Congress. "As we stand today, the prediction error 'floor' is associated with a standard deviation of 0.31 D and a mean absolute error of 0.24 D. These values are close to the best that we can achieve. With these values, 90% of our patients would achieve within a half diopter of their predicted refraction for 'all comers,' meaning all cases coming to our practice." Dr. Holladay explained that prediction error, the difference between the actual and predicted refraction, was always Gaussian (normal distribution). "The average outcomes of standard deviation for doctors today is 0.50 D (not 0.31 D above)," he said. "Gaussian statis- tics tell us that if something has a Limiting prediction error to improve IOL power accuracy

• Blank Page
• Blank Page (1)