## An Advanced Method for IOL Power Calculation

Gary Wörtz, MD: Open, outspoken. It's Ophthalmology off the Grid—an honest look at controversial topics in the field. I'm Gary Wörtz.

For ophthalmologists, numbers hold a lot of weight. We rely heavily on measurements to understand our patient's conditions and inform our surgical decision-making pre-, intra-, and postoperatively. However, as we all know, there's always room for error. Even the most sophisticated calculation methods have their limitations. Hopefully, things are about to change.

One individual who has made tremendous strides in eliminating the guesswork and overcoming these limitations is Dr. Warren Hill. I touched base with Warren to learn more about his efforts to develop a new self-validating method for IOL power selection. Listen in as he takes us through this project and shares how 6 years of research and a strong collaboration brought this exciting tool to ophthalmologists.

Speaker 1: Ophthalmology off the Grid is an independent podcast supported with advertising by Alcon.

Gary: Today, I have Dr. Warren Hill with me, and I'm just so honored and thrilled that Warren could spend a little bit of time talking about not only his current project, the Hill RBF, which is something we're all very, very excited about, but also just a little bit of background on his experiences with biometry. Maybe start from the beginning, all the way through developing the website and then walking us through the project he's been on most recently with the Hill-RBF. Before we even get started, I just want to say thank you for all you've done for ophthalmology, all you've done for me over the years. We've had a few little cases that you've helped me out on, and I've appreciated that.

Warren, I just want to say thank you. Maybe as we get started here, you can just give us a little bit of a historical perspective on biometry and what piqued your interest in biometry when you started your practice, or maybe if it wasn't even at the beginning, when did you start thinking that biometry was fun, and when did you find out that this was your passion?

Warren Hill, MD: Thanks, Gary, for the kind words. It's a delight to be here with you and share some of the thoughts that we're going to go through in the next few minutes. This is coming up on my 32nd year in practice. When I first began, what I did was see patients for other physicians. Then, in my area in the mid-1980s, there were a lot of very large clinics. When physicians would have a problem, sometimes patients would drift over to my office, and I would fix the problem and then send the patient back. Often that involved a lens exchange, a corneal transplant, perhaps vitrectomy. Back then, general ophthalmologists did a lot more of different types of procedures than what we do right now.

What I found was, probably 8 or 10:1, the most common problem I saw was the wrong power IOL. This became the mission for me, because more and more patients were being sent to me to help figure this out. In the late 80s, I kind of made friends with Jack Holladay, who was very gracious and very kind and helped me understand things. I grew up around the whole field of IOL calculations, mostly out of necessity. Eventually, physicians would send me patients before the refractive surprise came, and I guess you get good at what you do a lot of, and, over time, that sort of became one of the centers of my practice. To this day, I'm still mostly doing referral anterior segment surgery with an emphasis on IOL power calculations for other physicians.

The website that I have that you referred to was also born out of necessity. We used to get a lot of questions about, how do I handle staphyloma or silicone oil, or what do I do in the setting of extreme axial myopia. The right-hand column of that website became a depository of answers to questions I used to get very frequently, and it was more out of survival than anything else just to refer people to that so we didn't have to go though the same answer time and time again. Over the years, that website just kind of evolved into its present form.

Gary: It's interesting, it sounds like you basically discovered an unmet need in the marketplace, and in your area, you decided to be willing to fall on the sword, so to speak, to do the really tough things, and because of your willingness to do that and your willingness to be an educator and mentor, it has really opened up doors probably, that you maybe never dreamed of when you were starting your practice and taking care of all these train wrecks or patients who maybe that no one else wanted to take care of. Is that a fair assessment?

Warren: Actually, yeah. There's a little more to it than that. My father-in-law, who was a physician, always used to say "The best way to learn something is to try to teach it to someone else." So, in the 90s, I used to go to meetings, and I still do, and I would give lectures on IOL calculations. That's really when the light came on for me about how broad the problem was, how little we had to work with at the time, and how much this was needed. Again, my knowledge evolved really through the questions of those people I was looking to help and through courses I was teaching, and with more and more exposure and more and more interest came more and more knowledge.

Part of the interesting thing about the website is that people will send me cases and say "Gee, what do I do with this" or "Why did this happen?" and I would work through the problem with them and come up with the right answer; and what they would get is the right answer, of course, and what I would get is knowledge. And with every single one of these that came down the pike, I gained more and more experience, and through that experience that came from other people, evolved what I know today.

Gary: Well, one of the things that, you know this is an area that I'm very interested in, as I have my own ideas about effective lens position, and maybe an IOL that someday will help solve that. I kind of had an epiphany moment after we spoke one time and attending some lectures, doing some reading, and really just thinking about biometry, I thought, "You know, it all comes down to effective lens position, and all these formulas essentially are just taking various bits of data, whether that's just the axial length; or the axial length and the Ks; or axial length, Ks, anterior chamber depth, and the list goes on and on. But, essentially, I kind of boiled it down in my mind that all of these formulas are doing is taking different bits of data, weighting those differently based on the population they used to derive their formula, and coming up with where the best guess for the effective lens position will be.

When I had that epiphany, I just thought, “Is that really true, is that what we're really doing?” What are your thoughts on that? Is that a fair assessment at a 50,000-foot level of what the formulas have done for us in the past?

Warren: Well, actually let's go up to 100,000 feet, and this is going to be a surprise to most people, but the formulas we're using right now really came from Dr. Gauss in 1840, and keep in mind that's the first half of the 19th century. What Dr. Gauss published was a basic formula: object vergence, plus lens vergence, equals image vergence. If we fast-forward to current times, if you look at any modern formula, if we have a formula that solves for the IOL power, that's lens vergence, equals image vergence, minus object vergence; it's still Dr. Gauss’ formula.

Probably about 30 or maybe more percent of the accuracy of that formula is the effective lens position, and, as you know, the effective lens position can only be estimated; it can't be calculated. So, a large part of all theoretical formulas is based on something that can only be estimated, and that's kind of the blind spot. If the eye didn't read the textbook, if the eye didn't line up with the assumptions of the formula, that is the refractive surprise you get.

It might be a surprise to people that if you look at a series of patients, and I've looked at about 260,000 implantations as part of the formula optimization database, the vast majority, greater than 90% of surgeons are at about 78% within 0.50 D. Said differently, that's 22% who are outside the 0.50 D accuracy. Only 6% of surgeons are at 84% or better, and less than 1% of surgeons are at 92%. So, using the older formulas, like third-generation two-variable formulas, these are the kind of accuracy results you're going to get. We really need to move up to theoretical formulas from this century, is kind of the joke we like to make, that do a much better job with the effective lens position. Those formulas would be like Olsen's formula, Thomas Olsen's C formula, and then Graham Barrett’s Universal II formula. These are really two gifts to ophthalmology, and they've taken theoretical formulas way, way passed what we thought in the past. As long as you have a theoretical formula, you're still tied to the effect of the lens position.

Gary: Right and it sounds like you've found another unmet need in terms of figuring out how do we move the ball beyond the 78% plus or minus 0.50 D for the vast majority of physicians, and how do we get a formula that is going to perform well, not just for the eyes that read the textbook, and the eyes that are the "normal shape and size eyes." It sounds like the Hill-RBF is maybe our next best advancement in ophthalmology to get it right for eyes across the spectrum. Walk me through a little bit about the genesis of this idea, what made you think of it, and how you decided to tackle such a huge project.

Warren: First, let's start at the beginning; this all began about 7 years ago, and it began with working with MathWorks, with Pete Maloney, who is an engineer at MathWorks. It has evolved into a project with 25 surgeon beta-testers in 14 countries, so this isn't just my work, of course, this is very much a team effort. Right now, our core investigators are Doug Koch and Li Wang at Baylor University (these two need no introduction to anybody in ophthalmology); a young ophthalmologist in Tel Aviv, Israel, his name is Adi Abulafia, someone you're going to be hearing a lot about in the future, this young man has a gift for analysis; and then David Goldblum, who is a professor of ophthalmology at the University of Basel in Switzerland. Our beta-testers are in Europe, the Middle East, Africa, North and South America, Asia, India, and Australia, so we have a really wide range of people involved in this project.

So, the idea was to get rid of the effective lens position using a completely different mathematical model. That mathematical model is an engineering-based statistical model of which there are many different forms. There are things like Gaussian process methods, stochastic process methods, polynomial models, and what we settled on was some thing called a radial basis function. A radial basis function is like a form of artificial neural network, except that it handles the data a little bit differently, and we use it for pattern recognition.

The way our theoretical formula works is you have the measurement points that kind of go into a black box, which is usually a regression-derived algorithm, and then out the other end comes an IOL power for a given spherical equivalent. Well, what this does is we use what are called boundary model pairs, like axial length versus anterior chamber depth, anterior chamber depth versus central corneal power, central corneal power versus spherical equivalent. We create what are called a boundary model around each one of these; in other words, data points within these theoretical boundary models are considered valid, and, if a data point falls outside of it, it's invalid. You can actually flag this for the user to tell them if the lens power that we are going to estimate is at a certain accuracy level.

So, instead of applying preoperative measurement data points to a regression-derived algorithm, instead, we look at it as a pattern and apply a boundary model for internal validation. Right now our boundary model is at about 90% accuracy within 0.50 D. So rather than just run regression line across axial length versus anterior chamber depth, we actually were able to keep track of all the data points simultaneously, so it's a completely different mathematical model. In engineering circles, this is completely normal, this is just a standard tool, but this is something new to ophthalmology.

Gary: It sounds like this is something that is already well-established but also sounds like something that may learn and advance and get better as time goes on with more and more data points, is that correct?

Warren: You bring up really an excellent point: These engineering based statistical models are big data exercises. Right now, we're working with about 3,445 points to create our pairwise boundary models and do our panel recognition. This last week we just finished adding about 10,000 cases to the database for this, including IOLs down as far as -5.00 D, and we specifically concentrated on implantations of IOLs for very, very high hyperopes, which is always one of the more problematic parts of cases that we do. As the size of the database increases, the depth and breadth of the boundary model increases, and the overall accuracy gets better and better and better.

So we are going to be at about 10,000 or 12,000 cases in the next version, which I hope to have finished by the end of the year. Then, at the end of next year, we may be up to 30,000 cases, and eventually our hope is that with this model and the input from physicians around the world who have volunteered to help us, we may not ever have to worry about the spherical equivalent again. Just like when we went from ultrasound to optical biometry, you remember what an epiphany that was for all of us. Going from an applanation A-scan to the IOLMaster basically took the axial length completely out of the picture; we just didn't have to worry about it. Our goal with this project is to make the spherical equivalent mostly a nonissue and then move on to other things.

Gary: I guess one question I would have is, and this requires you to take off your developer and scientist hat and put on the hat of a physician who's busy in practice, we all are creatures of habit, and that can be protective, that in some ways prevents us from doing things that may not be well-established, but it can also be detrimental in some ways, where, as surgeons, were maybe reticent to change from that good old Holladay I formula we've been using for 10 years or longer. In a busy practice, you get this data that says, "All right, we've got a new formula everyone, stop the presses. Change from what you've been doing and move over to this." What do you advise fellow surgeons who this whole concept may be a little bit new to, and maybe the security blanket of that formula is maybe not perfect but they're at least used to the variability, how do you counsel them to take the plunge and feel safe to move over to something that is a new model?

Warren: Gary, this is really an excellent observation. Lets backtrack a little bit to the 1960s. There was a book I read back then by a man named Everett Rogers called The Diffusion of Innovations, and if you don't know the book, if you haven't read or are not familiar with his work, you're certainly familiar with the terminology because we use it everyday.

Gary: Yeah, I'm very familiar; love it.

Warren: What he said was that there are basically five groups. It is distributed like a bell curve. About 34% of any group that's exposed to a technology is termed the early majority, and these are people who try something only after somebody tries it first, and I think that's probably what you're referring to. Another 34% are the late majority, and these are people that kind of border on cynical or are looking for a lot of control, perhaps administrative types that oversee health plans. There is another 16% that are kind of like the laggards, or people who will only change technologies if it is absolutely forced on us. Then there is the 13.5% of people who are the early adapters, and these are people who believe in it and do it for themselves, regardless of what anybody tells us. There is a very, very small group of the innovators—the Graham Barrett’s, the Doug Koch’s, the Thomas Olsen’s—who give us new technology.

Now, to the left of this curve, all of these people are interested in enhancing outcomes; to the right of the curve, they're interested in maintaining familiarity, and a lot of ophthalmologists are in that last group of maintaining familiarity. Where I'm going with this is when we first came out with this and I put up the online calculator, I wasn't sure ophthalmologists would warm up to this until a couple of years. We had 9,000 calculations on the calculator in the first 5 days, and, in a 17-week period, we had about 38,000 calculations.

So what's happened is not only is this an unmet need, but I think ophthalmologists as they become more and more interested and adapt to new technology are more willing to try new things than they were perhaps in the years in my generations years ago. So the acceptance of this is beyond anything I could've imagined.

Gary: So maybe your observation would be that the curve is actually shifted to the left, where there are maybe a few more early adapters than you had anticipated.

Warren: Yeah, judging by the traffic that we're now getting on the website and the sales of the Lenstar, which is where this formula now resides. The sales of the Lenstar have now gone through the ceiling since we put this on it, and this is really what's driving that in some part.

So I think it's two things; one is that the newer generation of ophthalmologists is not quite as concerned about technology as you and I may have been early in our careers, and they're accepting. They're accepting of trying something new, and the use has just been amazing, which is great because the website is going to be used for what's known in the business as crowd-sourcing, where as each calculation is put on that it's captured and then the user gets an email t3 months later asking for results. We're able to build up the database from that, so this is becoming actually a worldwide collaborative effort a lot quicker than anybody could have ever imagined.

Gary: Well, and it starts with leadership, and it starts with someone who has a vision and sees an unmet need and is willing to put in the blood, sweat, and tears and put their name behind something that, at the beginning, you don't know if it's going to work, or if it's going to be accepted, or frankly if you're going to be laughed at if this fails—not that we would ever laugh at you, but I'm saying in general when you innovate it's an exercise that has to be fueled by passion, but it benefits the field, it benefits patients. This is just such an area that we all recognize we could do better, and to be honest that's why I'm doing what I'm doing with trying to develop a lens that could help us predict even better based on some different structural elements within my lens, so I'm kind of on the same side of this situation trying to innovate.

When I go to conferences and I go to trade shows and we walk around and listen to lectures, and I'm trying to think do I need another box? Do I need a better topographer? Do I need dual Scheimpflug? Do I need intraoperative aberrometry? What is the magic box that I need? And what's kind of interesting is it maybe isn't the data that we have, maybe it's the way that we're using the data, and it seems you have approached this in a lateral manner and have said it's not necessarily the data, we have ways to measure the eye; it's how are we applying it to the eye that's going to make the biggest difference. Is that a fair assessment?

Warren: Yeah, actually to take this one step further, I think what we're seeing is really converging technologies. So, the next step for ophthalmology that I think will be as big an epiphany as when we went from applanation ultrasound to optical biometry, will be being able to reach a 90% level for 0.50 D accuracy for everyone. We're seeing three technologies that are actually approaching that all at the same time. One is theoretical formulas and Graham Barrett’s Universal II formula with really good measurements can actually do this, and his formula is just an amazing theoretical formula. The other is intraoperative aberrometry, and one of the newer forms of intraoperative aberrometry, the Holos device, has an algorithm that actually Graham Barrett wrote; instead of doing the preoperative measurements to come up with IOL power, it looks at an aphakic measurement and then adds power to an existing optical system using something called the RX formula that Graham developed, and that can be at 90%. And then now the RBF method is moving us in that direction as well.

Mike Snyder, Steve Scoper, and I did a prospective study at the earlier part of this year, and that's pretty much where we ended up. For all cases, consecutive cases by three surgeons, we ended up at 91%. We're getting to that place where we all want to be, at least at this stage in ophthalmology. It's possible to get there by three different methods, and I see that as wonderful; now it becomes a horse race, and everybody's trying to get a little bit better and a little bit better, and this is a very exciting time. With your technology as well, you have the option of not only changing the IOL if the spherical equivalent is not correct, but changing the type of IOL if the patient wants a multifocal, or a toric, or a drug delivery system. The technologies are just blossoming in all kinds of different directions. This is a really fun time to be an ophthalmologist

Gary: So, as we get sort of toward the end of this conversation, one question I would have for you is: Are we where you thought we would be when you started your career? And maybe there's two parts to that question: What has surprised you, and is there anything that you thought we'd already be at and we're not there yet? So maybe surprised in a positive direction and surprised in a not so positive or in an unmet need direction.

Warren: We'll start with the latter. What still surprises me is that we, as surgeons, are judged by our patients and our peers by our refractive outcomes—no question, that's a fact, undisputed—but yet, so many surgeons hand over the IOL power calculation part of their practice to technicians in a completely unsupervised environment. The old joke is “automate and delegate.” It seems a little bit of a paradox that we all do wonderful surgery and our patients look fantastic the next day, but yet we're using technology from the 1980s or 1990s as far as selecting IOL power for the outcome. None of us would drive a car from 1998, use a cell phone from 1991, or use a telephone from 1993, and yet that's the technology that so many people are using, so that's a little bit of a surprise to me.

Now, on the good part, the good side, I'm encouraged to see people making the move to this artificial intelligence RBF calculator and also using Graham Barrett’s formula. Both of these are available online at no charge to anybody, so there's no cost to use something like this. I am excited about that. Another thing that excites me is the explosion of technology that we're going to be seeing over the next 2 years regarding preoperative measurements. A lot of companies are moving to swept-source OCT. There's a couple of wonderful devices on the market right now, and I think everybody's going to move to that and when we start measuring the cornea using swept-source OCT, that second part of the equation with a lot of variability, which is the anterior cornea, that may go away as well.

Right now, the greatest variability in selecting IOL power is the calculation method and the measuring of the cornea. If we can take the calculation method to very high levels and get rid of the noise from a reflective keratometry to swept-source OTC, then I think it's a whole new ball game and I can say we may not have to worry about the spherical equivalent anymore, now we can worry about the other things that we haven't been concentrating on so much because we've been so worried about getting the power correct.

Gary: Well, Warren, I agree with you. I think those are both areas that, on one hand, we need to strive to improve, and, in other ways, we're just so fortunate and lucky that we have great physicians and also partners in industry that are coming up with technologies that are going to continue to make us better and make us heroes in the eyes of our patients.

So, beyond that, I just want to say a hearty thank you. Personally, there have been a number of times that I've had a significant questions about patients that either required an exchange or other head-scratchers, and as busy as you are taking care of the world's IOL conundrums, you've always made time for me personally, and that goes back to when I was first out of residency and continues to this day. So, I just want to say a hearty thank you from all of the ophthalmologists worldwide for what you've done and also a very personal thank you for being a resource that I know I can count on and trust. So, thank you so much Warren for everything you've done for us.

Warren: Well, I appreciate the kind words, and it was wonderful to spend some time with you this evening.

Gary: Absolutely. Well, thanks again, Warren, and you're always welcome to come back if you ever have anything you'd like to share, so thank you very much.

For more on the Hill RBF calculator visit RBFCalculator.com

This has been Ophthalmology Off the Grid, stay up to date on all of our episodes by subscribing, and if you've got an idea, comment, or opinion you'd like to share, please leave us a review. This is Dr. Gary Wörtz. Thanks for tuning in.

Speaker 1: Ophthalmology off the Grid is an independent podcast supported with advertising by Alcon.

Gary Wörtz, MD: Open, outspoken. It's Ophthalmology off the Grid—an honest look at controversial topics in the field. I'm Gary Wörtz.

For ophthalmologists, numbers hold a lot of weight. We rely heavily on measurements to understand our patient's conditions and inform our surgical decision-making pre-, intra-, and postoperatively. However, as we all know, there's always room for error. Even the most sophisticated calculation methods have their limitations. Hopefully, things are about to change.

One individual who has made tremendous strides in eliminating the guesswork and overcoming these limitations is Dr. Warren Hill. I touched base with Warren to learn more about his efforts to develop a new self-validating method for IOL power selection. Listen in as he takes us through this project and shares how 6 years of research and a strong collaboration brought this exciting tool to ophthalmologists.

Speaker 1: Ophthalmology off the Grid is an independent podcast supported with advertising by Alcon.

Gary: Today, I have Dr. Warren Hill with me, and I'm just so honored and thrilled that Warren could spend a little bit of time talking about not only his current project, the Hill RBF, which is something we're all very, very excited about, but also just a little bit of background on his experiences with biometry. Maybe start from the beginning, all the way through developing the website and then walking us through the project he's been on most recently with the Hill-RBF. Before we even get started, I just want to say thank you for all you've done for ophthalmology, all you've done for me over the years. We've had a few little cases that you've helped me out on, and I've appreciated that.

Warren, I just want to say thank you. Maybe as we get started here, you can just give us a little bit of a historical perspective on biometry and what piqued your interest in biometry when you started your practice, or maybe if it wasn't even at the beginning, when did you start thinking that biometry was fun, and when did you find out that this was your passion?

Warren Hill, MD: Thanks, Gary, for the kind words. It's a delight to be here with you and share some of the thoughts that we're going to go through in the next few minutes. This is coming up on my 32nd year in practice. When I first began, what I did was see patients for other physicians. Then, in my area in the mid-1980s, there were a lot of very large clinics. When physicians would have a problem, sometimes patients would drift over to my office, and I would fix the problem and then send the patient back. Often that involved a lens exchange, a corneal transplant, perhaps vitrectomy. Back then, general ophthalmologists did a lot more of different types of procedures than what we do right now.

What I found was, probably 8 or 10:1, the most common problem I saw was the wrong power IOL. This became the mission for me, because more and more patients were being sent to me to help figure this out. In the late 80s, I kind of made friends with Jack Holladay, who was very gracious and very kind and helped me understand things. I grew up around the whole field of IOL calculations, mostly out of necessity. Eventually, physicians would send me patients before the refractive surprise came, and I guess you get good at what you do a lot of, and, over time, that sort of became one of the centers of my practice. To this day, I'm still mostly doing referral anterior segment surgery with an emphasis on IOL power calculations for other physicians.

The website that I have that you referred to was also born out of necessity. We used to get a lot of questions about, how do I handle staphyloma or silicone oil, or what do I do in the setting of extreme axial myopia. The right-hand column of that website became a depository of answers to questions I used to get very frequently, and it was more out of survival than anything else just to refer people to that so we didn't have to go though the same answer time and time again. Over the years, that website just kind of evolved into its present form.

Gary: It's interesting, it sounds like you basically discovered an unmet need in the marketplace, and in your area, you decided to be willing to fall on the sword, so to speak, to do the really tough things, and because of your willingness to do that and your willingness to be an educator and mentor, it has really opened up doors probably, that you maybe never dreamed of when you were starting your practice and taking care of all these train wrecks or patients who maybe that no one else wanted to take care of. Is that a fair assessment?

Warren: Actually, yeah. There's a little more to it than that. My father-in-law, who was a physician, always used to say "The best way to learn something is to try to teach it to someone else." So, in the 90s, I used to go to meetings, and I still do, and I would give lectures on IOL calculations. That's really when the light came on for me about how broad the problem was, how little we had to work with at the time, and how much this was needed. Again, my knowledge evolved really through the questions of those people I was looking to help and through courses I was teaching, and with more and more exposure and more and more interest came more and more knowledge.

Part of the interesting thing about the website is that people will send me cases and say "Gee, what do I do with this" or "Why did this happen?" and I would work through the problem with them and come up with the right answer; and what they would get is the right answer, of course, and what I would get is knowledge. And with every single one of these that came down the pike, I gained more and more experience, and through that experience that came from other people, evolved what I know today.

Gary: Well, one of the things that, you know this is an area that I'm very interested in, as I have my own ideas about effective lens position, and maybe an IOL that someday will help solve that. I kind of had an epiphany moment after we spoke one time and attending some lectures, doing some reading, and really just thinking about biometry, I thought, "You know, it all comes down to effective lens position, and all these formulas essentially are just taking various bits of data, whether that's just the axial length; or the axial length and the Ks; or axial length, Ks, anterior chamber depth, and the list goes on and on. But, essentially, I kind of boiled it down in my mind that all of these formulas are doing is taking different bits of data, weighting those differently based on the population they used to derive their formula, and coming up with where the best guess for the effective lens position will be.

When I had that epiphany, I just thought, “Is that really true, is that what we're really doing?” What are your thoughts on that? Is that a fair assessment at a 50,000-foot level of what the formulas have done for us in the past?

Warren: Well, actually let's go up to 100,000 feet, and this is going to be a surprise to most people, but the formulas we're using right now really came from Dr. Gauss in 1840, and keep in mind that's the first half of the 19th century. What Dr. Gauss published was a basic formula: object vergence, plus lens vergence, equals image vergence. If we fast-forward to current times, if you look at any modern formula, if we have a formula that solves for the IOL power, that's lens vergence, equals image vergence, minus object vergence; it's still Dr. Gauss’ formula.

Probably about 30 or maybe more percent of the accuracy of that formula is the effective lens position, and, as you know, the effective lens position can only be estimated; it can't be calculated. So, a large part of all theoretical formulas is based on something that can only be estimated, and that's kind of the blind spot. If the eye didn't read the textbook, if the eye didn't line up with the assumptions of the formula, that is the refractive surprise you get.

It might be a surprise to people that if you look at a series of patients, and I've looked at about 260,000 implantations as part of the formula optimization database, the vast majority, greater than 90% of surgeons are at about 78% within 0.50 D. Said differently, that's 22% who are outside the 0.50 D accuracy. Only 6% of surgeons are at 84% or better, and less than 1% of surgeons are at 92%. So, using the older formulas, like third-generation two-variable formulas, these are the kind of accuracy results you're going to get. We really need to move up to theoretical formulas from this century, is kind of the joke we like to make, that do a much better job with the effective lens position. Those formulas would be like Olsen's formula, Thomas Olsen's C formula, and then Graham Barrett’s Universal II formula. These are really two gifts to ophthalmology, and they've taken theoretical formulas way, way passed what we thought in the past. As long as you have a theoretical formula, you're still tied to the effect of the lens position.

Gary: Right and it sounds like you've found another unmet need in terms of figuring out how do we move the ball beyond the 78% plus or minus 0.50 D for the vast majority of physicians, and how do we get a formula that is going to perform well, not just for the eyes that read the textbook, and the eyes that are the "normal shape and size eyes." It sounds like the Hill-RBF is maybe our next best advancement in ophthalmology to get it right for eyes across the spectrum. Walk me through a little bit about the genesis of this idea, what made you think of it, and how you decided to tackle such a huge project.

Warren: First, let's start at the beginning; this all began about 7 years ago, and it began with working with MathWorks, with Pete Maloney, who is an engineer at MathWorks. It has evolved into a project with 25 surgeon beta-testers in 14 countries, so this isn't just my work, of course, this is very much a team effort. Right now, our core investigators are Doug Koch and Li Wang at Baylor University (these two need no introduction to anybody in ophthalmology); a young ophthalmologist in Tel Aviv, Israel, his name is Adi Abulafia, someone you're going to be hearing a lot about in the future, this young man has a gift for analysis; and then David Goldblum, who is a professor of ophthalmology at the University of Basel in Switzerland. Our beta-testers are in Europe, the Middle East, Africa, North and South America, Asia, India, and Australia, so we have a really wide range of people involved in this project.

So, the idea was to get rid of the effective lens position using a completely different mathematical model. That mathematical model is an engineering-based statistical model of which there are many different forms. There are things like Gaussian process methods, stochastic process methods, polynomial models, and what we settled on was some thing called a radial basis function. A radial basis function is like a form of artificial neural network, except that it handles the data a little bit differently, and we use it for pattern recognition.

The way our theoretical formula works is you have the measurement points that kind of go into a black box, which is usually a regression-derived algorithm, and then out the other end comes an IOL power for a given spherical equivalent. Well, what this does is we use what are called boundary model pairs, like axial length versus anterior chamber depth, anterior chamber depth versus central corneal power, central corneal power versus spherical equivalent. We create what are called a boundary model around each one of these; in other words, data points within these theoretical boundary models are considered valid, and, if a data point falls outside of it, it's invalid. You can actually flag this for the user to tell them if the lens power that we are going to estimate is at a certain accuracy level.

So, instead of applying preoperative measurement data points to a regression-derived algorithm, instead, we look at it as a pattern and apply a boundary model for internal validation. Right now our boundary model is at about 90% accuracy within 0.50 D. So rather than just run regression line across axial length versus anterior chamber depth, we actually were able to keep track of all the data points simultaneously, so it's a completely different mathematical model. In engineering circles, this is completely normal, this is just a standard tool, but this is something new to ophthalmology.

Gary: It sounds like this is something that is already well-established but also sounds like something that may learn and advance and get better as time goes on with more and more data points, is that correct?

Warren: You bring up really an excellent point: These engineering based statistical models are big data exercises. Right now, we're working with about 3,445 points to create our pairwise boundary models and do our panel recognition. This last week we just finished adding about 10,000 cases to the database for this, including IOLs down as far as -5.00 D, and we specifically concentrated on implantations of IOLs for very, very high hyperopes, which is always one of the more problematic parts of cases that we do. As the size of the database increases, the depth and breadth of the boundary model increases, and the overall accuracy gets better and better and better.

So we are going to be at about 10,000 or 12,000 cases in the next version, which I hope to have finished by the end of the year. Then, at the end of next year, we may be up to 30,000 cases, and eventually our hope is that with this model and the input from physicians around the world who have volunteered to help us, we may not ever have to worry about the spherical equivalent again. Just like when we went from ultrasound to optical biometry, you remember what an epiphany that was for all of us. Going from an applanation A-scan to the IOLMaster basically took the axial length completely out of the picture; we just didn't have to worry about it. Our goal with this project is to make the spherical equivalent mostly a nonissue and then move on to other things.

Gary: I guess one question I would have is, and this requires you to take off your developer and scientist hat and put on the hat of a physician who's busy in practice, we all are creatures of habit, and that can be protective, that in some ways prevents us from doing things that may not be well-established, but it can also be detrimental in some ways, where, as surgeons, were maybe reticent to change from that good old Holladay I formula we've been using for 10 years or longer. In a busy practice, you get this data that says, "All right, we've got a new formula everyone, stop the presses. Change from what you've been doing and move over to this." What do you advise fellow surgeons who this whole concept may be a little bit new to, and maybe the security blanket of that formula is maybe not perfect but they're at least used to the variability, how do you counsel them to take the plunge and feel safe to move over to something that is a new model?

Warren: Gary, this is really an excellent observation. Lets backtrack a little bit to the 1960s. There was a book I read back then by a man named Everett Rogers called The Diffusion of Innovations, and if you don't know the book, if you haven't read or are not familiar with his work, you're certainly familiar with the terminology because we use it everyday.

Gary: Yeah, I'm very familiar; love it.

Warren: What he said was that there are basically five groups. It is distributed like a bell curve. About 34% of any group that's exposed to a technology is termed the early majority, and these are people who try something only after somebody tries it first, and I think that's probably what you're referring to. Another 34% are the late majority, and these are people that kind of border on cynical or are looking for a lot of control, perhaps administrative types that oversee health plans. There is another 16% that are kind of like the laggards, or people who will only change technologies if it is absolutely forced on us. Then there is the 13.5% of people who are the early adapters, and these are people who believe in it and do it for themselves, regardless of what anybody tells us. There is a very, very small group of the innovators—the Graham Barrett’s, the Doug Koch’s, the Thomas Olsen’s—who give us new technology.

Now, to the left of this curve, all of these people are interested in enhancing outcomes; to the right of the curve, they're interested in maintaining familiarity, and a lot of ophthalmologists are in that last group of maintaining familiarity. Where I'm going with this is when we first came out with this and I put up the online calculator, I wasn't sure ophthalmologists would warm up to this until a couple of years. We had 9,000 calculations on the calculator in the first 5 days, and, in a 17-week period, we had about 38,000 calculations.

So what's happened is not only is this an unmet need, but I think ophthalmologists as they become more and more interested and adapt to new technology are more willing to try new things than they were perhaps in the years in my generations years ago. So the acceptance of this is beyond anything I could've imagined.

Gary: So maybe your observation would be that the curve is actually shifted to the left, where there are maybe a few more early adapters than you had anticipated.

Warren: Yeah, judging by the traffic that we're now getting on the website and the sales of the Lenstar, which is where this formula now resides. The sales of the Lenstar have now gone through the ceiling since we put this on it, and this is really what's driving that in some part.

So I think it's two things; one is that the newer generation of ophthalmologists is not quite as concerned about technology as you and I may have been early in our careers, and they're accepting. They're accepting of trying something new, and the use has just been amazing, which is great because the website is going to be used for what's known in the business as crowd-sourcing, where as each calculation is put on that it's captured and then the user gets an email t3 months later asking for results. We're able to build up the database from that, so this is becoming actually a worldwide collaborative effort a lot quicker than anybody could have ever imagined.

Gary: Well, and it starts with leadership, and it starts with someone who has a vision and sees an unmet need and is willing to put in the blood, sweat, and tears and put their name behind something that, at the beginning, you don't know if it's going to work, or if it's going to be accepted, or frankly if you're going to be laughed at if this fails—not that we would ever laugh at you, but I'm saying in general when you innovate it's an exercise that has to be fueled by passion, but it benefits the field, it benefits patients. This is just such an area that we all recognize we could do better, and to be honest that's why I'm doing what I'm doing with trying to develop a lens that could help us predict even better based on some different structural elements within my lens, so I'm kind of on the same side of this situation trying to innovate.

When I go to conferences and I go to trade shows and we walk around and listen to lectures, and I'm trying to think do I need another box? Do I need a better topographer? Do I need dual Scheimpflug? Do I need intraoperative aberrometry? What is the magic box that I need? And what's kind of interesting is it maybe isn't the data that we have, maybe it's the way that we're using the data, and it seems you have approached this in a lateral manner and have said it's not necessarily the data, we have ways to measure the eye; it's how are we applying it to the eye that's going to make the biggest difference. Is that a fair assessment?

Warren: Yeah, actually to take this one step further, I think what we're seeing is really converging technologies. So, the next step for ophthalmology that I think will be as big an epiphany as when we went from applanation ultrasound to optical biometry, will be being able to reach a 90% level for 0.50 D accuracy for everyone. We're seeing three technologies that are actually approaching that all at the same time. One is theoretical formulas and Graham Barrett’s Universal II formula with really good measurements can actually do this, and his formula is just an amazing theoretical formula. The other is intraoperative aberrometry, and one of the newer forms of intraoperative aberrometry, the Holos device, has an algorithm that actually Graham Barrett wrote; instead of doing the preoperative measurements to come up with IOL power, it looks at an aphakic measurement and then adds power to an existing optical system using something called the RX formula that Graham developed, and that can be at 90%. And then now the RBF method is moving us in that direction as well.

Mike Snyder, Steve Scoper, and I did a prospective study at the earlier part of this year, and that's pretty much where we ended up. For all cases, consecutive cases by three surgeons, we ended up at 91%. We're getting to that place where we all want to be, at least at this stage in ophthalmology. It's possible to get there by three different methods, and I see that as wonderful; now it becomes a horse race, and everybody's trying to get a little bit better and a little bit better, and this is a very exciting time. With your technology as well, you have the option of not only changing the IOL if the spherical equivalent is not correct, but changing the type of IOL if the patient wants a multifocal, or a toric, or a drug delivery system. The technologies are just blossoming in all kinds of different directions. This is a really fun time to be an ophthalmologist

Gary: So, as we get sort of toward the end of this conversation, one question I would have for you is: Are we where you thought we would be when you started your career? And maybe there's two parts to that question: What has surprised you, and is there anything that you thought we'd already be at and we're not there yet? So maybe surprised in a positive direction and surprised in a not so positive or in an unmet need direction.

Warren: We'll start with the latter. What still surprises me is that we, as surgeons, are judged by our patients and our peers by our refractive outcomes—no question, that's a fact, undisputed—but yet, so many surgeons hand over the IOL power calculation part of their practice to technicians in a completely unsupervised environment. The old joke is “automate and delegate.” It seems a little bit of a paradox that we all do wonderful surgery and our patients look fantastic the next day, but yet we're using technology from the 1980s or 1990s as far as selecting IOL power for the outcome. None of us would drive a car from 1998, use a cell phone from 1991, or use a telephone from 1993, and yet that's the technology that so many people are using, so that's a little bit of a surprise to me.

Now, on the good part, the good side, I'm encouraged to see people making the move to this artificial intelligence RBF calculator and also using Graham Barrett’s formula. Both of these are available online at no charge to anybody, so there's no cost to use something like this. I am excited about that. Another thing that excites me is the explosion of technology that we're going to be seeing over the next 2 years regarding preoperative measurements. A lot of companies are moving to swept-source OCT. There's a couple of wonderful devices on the market right now, and I think everybody's going to move to that and when we start measuring the cornea using swept-source OCT, that second part of the equation with a lot of variability, which is the anterior cornea, that may go away as well.

Right now, the greatest variability in selecting IOL power is the calculation method and the measuring of the cornea. If we can take the calculation method to very high levels and get rid of the noise from a reflective keratometry to swept-source OTC, then I think it's a whole new ball game and I can say we may not have to worry about the spherical equivalent anymore, now we can worry about the other things that we haven't been concentrating on so much because we've been so worried about getting the power correct.

Gary: Well, Warren, I agree with you. I think those are both areas that, on one hand, we need to strive to improve, and, in other ways, we're just so fortunate and lucky that we have great physicians and also partners in industry that are coming up with technologies that are going to continue to make us better and make us heroes in the eyes of our patients.

So, beyond that, I just want to say a hearty thank you. Personally, there have been a number of times that I've had a significant questions about patients that either required an exchange or other head-scratchers, and as busy as you are taking care of the world's IOL conundrums, you've always made time for me personally, and that goes back to when I was first out of residency and continues to this day. So, I just want to say a hearty thank you from all of the ophthalmologists worldwide for what you've done and also a very personal thank you for being a resource that I know I can count on and trust. So, thank you so much Warren for everything you've done for us.

Warren: Well, I appreciate the kind words, and it was wonderful to spend some time with you this evening.

Gary: Absolutely. Well, thanks again, Warren, and you're always welcome to come back if you ever have anything you'd like to share, so thank you very much.

For more on the Hill RBF calculator visit RBFCalculator.com

This has been Ophthalmology Off the Grid, stay up to date on all of our episodes by subscribing, and if you've got an idea, comment, or opinion you'd like to share, please leave us a review. This is Dr. Gary Wörtz. Thanks for tuning in.

Speaker 1: Ophthalmology off the Grid is an independent podcast supported with advertising by Alcon.