Finding a Keratoconus Consensus

One of the primary hurdles in managing keratoconus lies in defining progression reliably. Unlike many other progressive conditions, keratoconus lacks a universally accepted definition of ectasia progression, leading to inconsistencies in treatment approaches and clinical trials. Factors such as changes in corneal shape, visual acuity, and corneal thickness are usually considered, yet reaching a consensus remains challenging despite advances in imaging technologies. 

“With better instruments, we have tighter measurements, but the current definition is still somewhat vague and less specific than most of us would like,” conceded José Güell MD, PhD. “Generally speaking, we are still following the guidelines that came from the first Delphi Panel, which was published in 2015.¹ The guidelines need updating, and this is what we are just starting to work on.” 

The progression of keratoconus is usually determined by comparing the results of corneal tomographic measurements at different time points. Kmax (maximum anterior sagittal curvature) is the most commonly used parameter to detect or document ectatic progression and is also regularly used as an indicator for cross-linking’s efficacy. The Delphi Panel advised combining parameters from both the anterior and posterior cornea and corneal thickness, where the magnitude of the change is above the normal noise of the testing system. The panel acknowledged, however, that specific quantitative data to define progression is lacking. 

In the absence of clear parameters, the discernment and experience of the treating clinician take on greater importance, Professor Güell noted. 

“From my perspective, any increases in mean corneal pow­er higher than 1.0 D or corneal astigmatism higher than 1.0 D are usually strong clues for deciding that keratoconus is progressing,” he said. “While there might be different opinions and nuances worldwide, these two main measurements are still broadly used to determine if a patient is progressing.”

Real world impact

Establishing a consensus on what constitutes progression is paramount, as it not only guides treatment decisions but also facilitates the development of uniform protocols for monitoring and managing the disease, Ingemar Gustafsson MD, PhD pointed out. 

“The Global Consensus from 2015 is still valid, in that it tells us which anatomical changes we must consider. It also specifies that we need to consider what it calls the ‘noise’ or measurement error of the measurement system. However, we consistently fail at that aspect, which makes it so troublesome,” he said. “We have become fixated on detecting an increase of Kmax at 1.0 D, which is not so helpful in scientific and clinical practice. Many adhere to the classic increase of Kmax of 1.0 D—in part by tradition, in part because that is what many other clinicians use. I would be glad to review more articles justifying the use of the selected parameters for the diagnosis of progressive keratoconus and the evaluation of the treatment efficacy.” 

With better technology available to clinicians, the critical question now is less about diagnosing a patient than determining the optimal intervention time to perform cross-linking, Dr Gustafsson believes.

“The diagnostic aspect is generally not difficult with modern Scheimpflug-based equipment and optical coherence tomography (OCT). For patients aged 18 or [younger], we proceed with cross-linking (CXL) upon diagnosis as we know these patients can progress rapidly. Otherwise, for patients 18 or [older], we make four initial measurements at one visit, and then we follow up after 4 to 6 months to see if there is any progression,” he said. “We include Kmax and the A (anterior), B (posterior), and C (minimal corneal thickness) parameters based on inter-day repeatability and stratified according to the disease severity. If there is no progression, we follow them for another 4 to 6 months [before making] another set of measurements after another year. If nothing happens, then we tell them to come back if they have any visual disturbances.

Close follow up

Close follow-up of younger patients is advised given the risk of progression is much higher in this cohort, advised Prof Güell. 

“For our young patients, a significant number of them will probably have progression in the 10 years following diagnosis. For those older than 18, if the proper treatment was performed and there is no subsequent trauma—basically no eye rubbing—then the number of cases with progression is extremely low,” he explained. “In cases of progression, we first need to check if the cross-linking treatment performed was the correct one. With the new modalities of epi-on, shorter treatments, new devices, et cetera, many patients have not received a standard protocol for treatment. The rate of retreatments in this particular population may be moderately high.” 

Prof Güell noted the threshold to administer cross-linking has evolved since the Delphi Panel consensus, with surgeons now less concerned about possible complications associated with cross-linking. 

“There is definitely less fear to indicate cross-linking despite not having objective information on the progression. Cross-linking has some potential complications, but the rates are quite low, and it is not always possible or practical to ask patients to undergo new examinations every few months,” he said. “For certain cases, we may decide to do cross-linking as a prophylactic treatment and not as a direct treatment for confirmed progression.” 

He added a decision to treat may also be motivated by low patient compliance in the case of confirmed keratoconus. 

“This is crucial because even with the best tools and technologies available, if the patient continues rubbing their eyes and does not respect the visit intervals for follow-up, then you have a major problem,” he said. “Although it is not my standard, if you think the compliance of your patient is going to be low, perhaps you will be more inclined to decide to perform cross-linking.” 

Dr Gustafsson said the policy in his clinic is to perform cross-linking only if deemed necessary. 

“We need to reflect on what is best for the patient and make the best use of our resources. We cross-link fewer patients in our clinic than before, even though we have many more new referrals. This is due to our more precise detection limits, which have reduced the number of unnecessary cross-linking procedures,” he said. “The CXL resources have thus been directed to those having the most to gain from cross-linking, providing a timely treatment which is essential in terms of preserving vision. Although cross-linking is safe in general, we prefer not to expose patients to a painful treatment and possible side effects unless we are sure it is appropriate.” 

Data deficit

While advances in corneal imaging devices and artificial intelligence-driven algorithms should facilitate more accurate and objective assessments, the effectiveness of these tools ultimately hinges on the availability of high-quality data for training algorithms, noted Olivia Li MD, a cornea consultant based in London. 

“I believe we should be collecting quality data through routine practice in a way that can be used to inform and train algorithms and contribute to our knowledge of keratoconus in the future,” she said. “There needs to be leadership and engagement from all parties really to work towards that goal.” 

Dr Li points to the medical retina domain, and diabetic retinopathy in particular, as an example of what can be achieved using the right approach. 

“What we have seen is how diabetic retinopathy databases across the world have been the foundation of clinically useful algorithms that are now functioning quicker, better, and more efficiently than humans in terms of diagnosing classified diabetic retinopathy,” she said. “It is no coincidence that the first FDA-approved AI algorithm in health was for diabetic retinopathy screening.” 

The key reason for that, Dr Li explained, is the longitudinal data set for diabetic retinopathy uses a standardized method to collect quality data from fundus images which are accessible, usable, and transferable. 

“All of these factors currently limit many aspects of what we do in the field of cornea. Most corneal databases, even internally within hospitals, are gathered across a number of imaging devices. And not all of them adopt the DICOM format—the usual means of storing and transferring medical images in a hospital’s database,” she explained. “That limits what you can do with the data. If you can’t get the raw image, you get derived parameters and human-generated numbers as a surrogate.”

Risk calculator

Despite these drawbacks, Dr Li and colleagues at Moorfields Eye Hospital have drawn on more than a decade’s worth of longitudinal data of keratoconus patients to build a risk calculator capable of predicting progression and time to cross-linking.² 

“It’s not perfect but it is certainly helpful in discussing various options with the patient and planning their treatment,” she said. “It’s particularly helpful for young people who have been otherwise completely healthy to understand the disease and its implications more because often these patients haven’t had to deal with a significant and life-changing diagnosis.” 

Dr Li and co-workers have also focused on the variability of same-day measurements using the Pentacam HR (Oculus).³ 

“For the key markers commonly used to define keratoconus progression, we generated limits of agreement, so the clinician knows exactly what has been defined. The definition for progression is when the change is greater than the 95% limit of agreement,” she said. “Essentially this means there’s a 95% chance any change won’t be just from measurement variability.” 

Moreover, the limit of agreement will change depending on the severity of disease present. 

“Measurements are usually more repeatable for early disease and less repeatable for advanced disease,” Dr Li explained. “What we have done is to put numbers to the different levels of Kmax, K2, and pachymetry, and we have done that for every dioptre to refine the definition of progression based on severity.” 

Although progress is being made to avoid both under- and overdiagnosis using such adaptive thresholds, there is still room for further improvement. 

“The problem is we still don’t really know what constitutes true progression. We have to draw the bar somewhere. A 1.0 D change in Kmax has worked relatively well for the last couple of decades as a decision for cross-linking—but we don’t know what the ground truth is as we move towards a more holistic assessment of progression instead of relying on single parameters,” she said. “As we start to use raw images, we will probably get closer to a tighter definition of progression because the richness of the data offers a much greater dimension of studying the disease, enabling us to get tighter progression thresholds.” 

With the latest Delphi Panel expected to start work later this year, the hope for everyone in keratoconus research is that a new consensus will finally bring clarity to existing grey areas. As Dr Gustafsson sees it, the key to making progress on all these issues lies in collective endeavour. 

“I would encourage colleagues to publish data in open-access repositories, collaborate, and create dedicated registries, as big data permits greater leaps in scientific progress,” he concluded. “And don’t be afraid of using more recent detection limits while we await the outcomes of the next global consensus.” 

José Güell MD, PhD is Director of the Cornea and Refractive Surgery Department at IMO Grupo Miranza and Associate Professor of Ophthalmology at Autònoma University, Barcelona, Spain. guell@imo.es

Ingemar Gustafsson MD, PhD is senior consultant of the cornea service at Sk åne University Hospital, Lund, Sweden. gustafsson.ingemar@gmail.com

Olivia Li MD is a cornea consultant based at Chelsea and Westminster Hospital, London, UK. olivialieyes@gmail.com

1. Gomes JA, et al. “Global consensus on keratoconus and ectatic diseases,” Cornea, 2015 Apr; 34(4): 359–69.

2. Maile Howard P, Li Ji-Peng Olivia, et al. “Personalized Model to Predict Keratoconus Progression from Demographic, Topographic, and Genetic Data,” Amer J of Ophthalmol, 2022 Apr 22; 240: 321–329. DOI: 10.1016/j. ajo.2022.04.004

3. Li Ji-Peng Olivia, Maile Howard P, Bunce Catey, et al. “A Comparison of Keratoconus Progression Following Collagen Cross-Linkage Using Standard or Personalised Keratometry Thresholds,” Eye, 2024 Feb 26: 1–6. DOI: 10.1038/ s41433-024-02994-6