Improving accuracy in predicting IOL power in cataract surgery patients
New biometric devices and data analysis techniques combine in the quest for optimum outcomes in cataract surgery
New biometry instruments that measure ocular dimensions with micron precision and intraocular lens (IOL) calculation formulas using increasingly sophisticated methodology are showing promise in providing much-needed improvements in the predictability of the IOL power in the eyes of cataract surgery patients.
The sophistication of IOL designs has also increased with the potential for precise correction of astigmatism and presbyopia, as well as an enhanced quality of vision. However, much of the benefit of the newer lens designs is lost where there is only slight residual refractive error.
Currently, on average around three-quarters of eyes undergoing cataract surgery achieve the requisite accuracy of being within plus or minus half of one dioptre of emmetropia. But there remain around 25% of patients who are not within that range and around 10% of patients who are more than 1.0D outside of predicted refraction, said David J Spalton FRCS, FRCP, FRCOphth, President of the ESCRS.
“We now realise that with our presbyopia-correcting lenses or toric lenses, the biometry has to be a lot better than it has been up until now if these lenses are going to work properly. And since 95% of refractive error is sphere and cylinder, the benefits of asphericity get lost in the noise unless your biometry and IOL calculation is really spot on,” he told EuroTimes in an interview.
“I also think we are entering a new era with regard to IOL calculation because the newer formulas, such as the Olsen, Barrett Universal II and Warren Hill’s Radial Basis Function formula, are all becoming much more accurate in offering a greater chance of achieving emmetropia,” he added.
IOL calculation formulas include a range of measurements, keratometry and axial length (AL) being essential to all, and anterior chamber depth (ACD) included in most, with lens thickness and corneal diameter measurements included in some. Originally these types of measurements required several tests with several machines. In more recent years there has been a trend towards all-in-one biometers that can provide all of those measurements and more in a single short test.
One example is optical biometry, which, almost from its launch in1999 in the form of the IOLMaster® (Zeiss), emerged as the gold standard for obtaining AL measurements. However, its precision based on partial coherence interferometry (PCI), was more limited when measuring ACD, and lens thickness.
A drawback of the multimode laser used with PCI is its spectral side maxima, which might lead to secondary interference maxima of the retinal peak, resulting in misinterpretation of the retinal signal in some cases. This, however, was addressed by a software upgrade (IOLMaster Version 5). This software allows averaging of consecutive optical scans, resulting in a composite scan. The new algorithm has been useful in 30% of cases that could not be measured with the previous software version.
A new non-contact optical biometer (Lenstar, Haag-Streit AG) using optical low-coherence reflectometry (OLCR) was commercially introduced in 2008. The technique was developed in the 1980s in the telecommunication industry for reflection measurements.
“With this new piece of technology, for the first time we were truly able to measure the ACD and for the first time were able to truly measure the lens thickness, and also get better measurements in the eyes with very dense cataracts and posterior subcapsular cataracts,” said Sathish Srinivasan FRCSEd, FRCOphth, FACS, University Hospital Ayr, Scotland, UK.
The new IOLMaster 700 (Zeiss) uses swept-source optical coherence tomography (SS-OCT) which enables more accurate ACD, measuring lens thickness and still better measurements of the cornea. It also provides more accurate keratometry and may eliminate the need for corneal topography in eyes with irregular corneas.
“The IOLMaster 700 is quite different and it is a game changer because it uses SS-OCT. That means we get a proper two-dimensional cross-sectional image of the entire eye. You also have a pachymetry map of the cornea. Software updates to actually give the posterior corneal curvature data will be coming out soon,” said Oliver Findl MD, Hanusch Hospital, Vienna, Austria.
The posterior corneal curvature is generally not an important factor in most cataract patients, but can be crucial in toric IOL calculations, and for eyes that have undergone previous corneal refractive surgery or keratoplasty procedures. Corneal topographers have been the device of choice for anterior and posterior curvature mapping, although the Barrett Toric IOL Calculator uses a fudge factor.
The values measured by ocular biometry in cataract patients make up the optical elements of the eye. Keratometry provides the refractive power, the AL provides the eye’s overall focal length, and the ACD provides an idea of where the lens is and therefore the power of the IOL needed to provide the eye with desired refraction.
However, the different formulas vary in terms of what parameters they include. All require AL and K, including several early formulas that required only those two parameters, such as the Hoffer Q, Holladay 1, SRK/T, and T2. The Haigis formula adds ACD, but does not use the K. Further parameters in other formulas include lens thickness, corneal diameter, age and preoperative refraction.
Since not all biometers can provide all of the measurements necessary for some third- and fourth-generation formulas, results with the different formulas will differ with different devices. At the same time, some formulas with fewer inputs perform as well or better than those with more inputs.
A recently published study (Cooke et al, Cataract Refract Surg, 2016; 42:1157–1164) showed that, in a retrospective series of 1,450 eyes implanted with the Acrysof SN60WF IOL (Alcon), the Olsen formula performed the best out of nine formulas at all ALs when Lenstar values were used, but performed the worst of all nine formulas at all ALs when PCI measurements were used and the lens thickness values were not included.
That may indicate the importance of lens thickness in the formula. However, when PCI measurements were used, the Barrett Universal II formula achieved the best results at all ALs, despite the formula’s requirement for unavailable lens thickness values.
Moreover, the Barrett formula performed almost as well as the Olsen formula when using Lenstar measurements, even though, unlike the latter formula, its inputs do not include patients’ age or central corneal thickness. Another finding of the study was that the Holladay 2 formula performed better when the preoperative refraction data was omitted.
FORMULA EVALUATION PROTOCOLS
Among the difficulties encountered when collating data for use in evaluating and comparing IOL formulas are the discrepancies in data collection and errors in methodology, according to an editorial by the IOL Power Club, a group of well-known creators of IOL calculation formulas and a respected statistician, in the March 2015 issue of the American Journal of Ophthalmology.
They point out, for example, that ACD is variously defined as the distance between the corneal epithelium and the anterior surface of the lens, and as the distance between the corneal endothelium and the lens. While the latter may be the original definition of ACD, they suggest that the endothelium-to-lens distance should instead be termed the ‘aqueous depth’ and ACD be used for the epithelium-to-lens distance.
Even the term ‘target refraction’ can be confusing and contradictory, since some authors define it as the desired refraction, whereas it is more appropriately defined as the refraction achievable with a given lens in a given eye according to the formula used.
Speaking with EuroTimes in an interview, Kenneth J Hoffer MD, one of the co-authors of the editorial, said that the use of the term ‘average’ for biometric values also needs to be better qualified. He pointed out that his own research indicates that biometric averages vary by gender and race.
“We did a study collecting data in 250,000 eyes published over the past 20 years and discovered that men have an AL a half millimetre longer than women when you adjust for height and weight, and a half a dioptre flatter cornea,” said Dr Hoffer, Stein Eye Institute, University of California, Los Angeles, USA.
He noted that average values are an important component of the Holladay 2 formula. In a retrospective, study involving 2,707 eyes, he was able to show that adjusting the average values in the Holladay 2 to the average values for the patient’s gender and race resulted in a 30-40 % better median absolute error prediction accuracy compared to the Hoffer Q, the Holladay 1 and the SRK/T formulas.
The new formula, called the Hoffer H-5 (H for Holladay; 5 for 5th generation), would have resulted in 84% of eyes being within ±0.20D of predicted refraction, he said. The H-5 formula will soon be available on the IOLMaster 700, the Lenstar and the Aladdin.
THE END OF FORMULAS?
Meanwhile, new research tools such as artificial intelligence are making inroads into IOL calculation with the introduction of Warren Hill’s Radial Basis Function software. Using pattern recognition software and an ever-growing database, the software detects correlations between postoperative outcomes and three IOL calculation inputs, such as AL, ACD, and steep and flat keratotomy values, and on that basis determines the best IOL for a particular patient.
“They’ve done both a prospective and retrospective analysis to show that it works not only in the retrospective database but also when you enter prospective data. So they have been incredibly thorough about it,” Prof Spalton told EuroTimes.
The Hill-RBF Calculator has been launched in 2016 and it is available for free online. Surgeons using the calculator are encouraged to enter optional values such as lens thickness and corneal diameter measurements, to provide data for future versions of the calculator.
One limitation of the radial basis function IOL power calculator in its present form online is that it is based on data obtained with only one optical biometer, the Lenstar LS 900, and only one type of IOL. And, though the user enters the type of biometer used and the A constant of the lens to be implanted, the website cautions users of the calculator that for optimal results the same biometer/lens as that of the original dataset should be used.
Dr Findl told EuroTimes that he also questions whether the traditional formulas of the past are the way forward. He noted that the accuracy of modern biometry instrumentation in measuring the eye’s dimensions and the IOL’s postoperative position in relation to the intraocular anatomy, mean that the term ‘effective lens position’ is out-of-date.
“I think we should replace the term ‘effective lens position’ with ‘anatomic lens position’. And I personally believe we should be moving away from only formulas and moving more towards ray tracing. We had formulas because, 15 or 20 years ago, our computers lacked the necessary speed. Today, with any smartphone we can do the calculation in a very short time,” Dr Findl added.
David J Spalton: firstname.lastname@example.org
Sathish Srinivasan: email@example.com
Oliver Findl: firstname.lastname@example.org
Kenneth J Hoffer: email@example.com