Ten years after the first Global Consensus on Keratoconus and Ectatic Diseases, we convened its second edition, published this year in Cornea (1). The first consensus, in 2014, brought together three panels drawn from four supranational corneal societies (2). The scale of the update reflects how much the field has grown in the decade since: 12 ophthalmologists nominated by 12 national and supranational societies across five continents, organized into seven panels covering everything from diagnosis and staging to cross-linking (CXL), keratoplasty, refractive surgery, and – for the first time – cataract surgery in keratoconic eyes. The machinery matters, so briefly: four structured Delphi rounds, anonymized for the first three to stop seniority and strong personalities from holding disproportionate influence over the voting, an independent statistics team holding the results until every response was in, a face-to-face meeting in Chicago, and a two-thirds majority required for any statement to stand.

The document delivers what a reference document should: a revised definition of keratoconus, a new mild-moderate-severe staging system built on the Belin ABCD grading (3), confirmation of CXL as the standard of care for stabilization at every stage, DALK as the preferred keratoplasty, and the field’s first consensus guidance on cataract surgery in keratoconus. Those recommendations will be quoted for the next decade, and rightly so.

But we would direct you to the parts of a consensus paper that usually go unread: the disagreements. When 128 people who have spent their careers focusing on one disease cannot agree on something, that is not noise. It is one of the most precise measurements we have of where the science actually stands.

The Kmax paradox

Let’s start with the most revealing finding. Asked about maximum keratometry – the parameter that has defined “progression” in virtually every CXL study since the Dresden protocol – 91.7 percent of the diagnostic panel agreed that Kmax has low repeatability and reproducibility for evaluating progression. In the same breath, 75 percent admitted they still use it in daily practice, and 67 percent still consider it a relevant progression parameter.

Read that again: nine in ten experts agree the ruler is unreliable; three in four keep measuring with it anyway.

This is not hypocrisy – it is what happens when an entire evidence base is built on a single metric. Kmax is what the pivotal trials reported, what regulators accepted, and what every tomographer displays by default. The panel could only agree on one alternative as genuinely superior: the anterior radius of curvature at the thinnest point. It also flagged a subtler problem – measurement variability itself increases as keratoconus worsens, so a fixed cutoff for “progression” means something different in a mild cornea than in a severe one. Hence, a direct request to industry: display zonal keratometry and anterior and posterior radii of curvature on every Scheimpflug and OCT device, so the next consensus can stand on better-behaved numbers.

Words we kept, but could not define

The panel voted to retain “subclinical keratoconus” and “forme fruste keratoconus” – and then could not reach consensus on formal definitions for either. The compromise is honest: both describe early of preclinical stages of disease, with subclinical usually implying progression and forme fruste usually not. What we could define was a new category, “suspicious keratoconus” – deliberately analogous to the glaucoma suspect – for eyes with findings that justify close monitoring but not a diagnosis.

Why such caution over vocabulary? Because the first consensus taught us what happens when a definition outruns its evidence. In 2015, we stated that posterior elevation abnormalities must be present to diagnose mild or subclinical keratoconus. A decade later, Randleman and colleagues reviewed 29 studies and found posterior surface metrics outperformed anterior and thickness metrics in fewer than 14 percent of them – no support, in aggregate, for that “must” (4). Edition 2 quietly reflects the correction: the major criterion is now abnormal steepening of the anterior and/or posterior surface. Consensus statements get cited as settled fact; the humility to write “we could not agree on a definition” is, we would argue, the more scientific act.

Where conviction outruns evidence

A consistent pattern runs through the disagreement tables: new technologies achieve consensus as adjuncts and lose it the moment they would drive an irreversible decision.

Corneal biomechanics is the clearest example – and we say this as authors whose own laboratories have spent the past two decades on corneal biomechanics and multimodal imaging. The panel agreed that biomechanical assessment may help detect ectasia susceptibility at a preclinical stage, and 71 percent endorsed it as a valuable add-on in refractive surgery screening. But should abnormal biomechanics, by itself, justify performing CXL when topography is stable? Only 53 percent said yes – no consensus. Epithelial thickness mapping followed the same arc: they agreed it was useful for diagnosis, but quantitative epithelial data were judged insufficient to diagnose, classify, or monitor progression, and epithelial thinning as a treatment trigger stalled at 60 percent. Genetic testing fared starkest of all: just 7 percent considered it essential for refractive surgery screening.

The same discipline was applied to pharmacology. Nearly three-quarters of the CXL panel judged the current evidence for adjuvant therapies – oral riboflavin, vitamin D, copper supplementation, biomarker modulation – insufficient, and 93 percent agreed that if such therapies are used at all, they belong alongside CXL, never instead of it. (On one point the panel was unanimous, and it is worth stating because patients encounter the claim online: oral riboflavin plus sunlight is not cross-linking, and cannot replace it.)

None of this means the technologies are wrong. It means the longitudinal studies that would let us act on them – rather than merely look at them – have not been done. That is not a limitation of the consensus; that is its output.

Agreeing on whether, disagreeing on how

Some of the sharpest splits came after agreement in principle. Thin corneas below 400 µm of stromal thickness: 86 percent agreed CXL is a valid option – a genuine shift from the first consensus era, when such corneas were simply ineligible – but no single technique reached consensus, with 60 percent swelling the stroma with hypoosmolar riboflavin and 53 percent using the sub400 individualized-fluence approach (5). Pediatric keratoconus shows the same shape: strong agreement (86 percent) that confirmed disease in a child justifies CXL at diagnosis without waiting for progression, yet when progression has not been documented, the panel fractured – 46 percent would monitor, 33 percent would perform epi-off CXL, 20 percent epi-on. Protocol choice itself never converged: accelerated epi-off at 9 mW/cm² was the most common regimen at around 40 percent, the original Dresden protocol held 26 percent, epi-on remained a minority technique reserved by 80 percent for milder, less progressive disease, and high-fluence protocols were routine for only around a third of panelists outside the US. Retreatment after continued progression? Over 90 percent agreed it was appropriate – and if the first procedure was epi-on, 86 percent would switch to epi-off rather than repeat it – but on when to retreat, the panel scattered across every interval from under six months to four years.

The surgical panels tell the same story. Everyone agrees intrastromal ring segments regularize the cornea; there is not enough evidence to say whether CXL or the rings should come first when both are indicated. Corneal allogeneic intrastromal ring segments (CAIRS) won agreement on biocompatibility, but not on whether they yet outperform their synthetic predecessors in predictability – and the reuse of myopic SMILE lenticules for stromal addition found no consensus at all. Even mitomycin C, a fixture of surface ablation for two decades, split the room when combined with PRK and CXL: 75 percent called it contraindicated, which is a consensus, but hardly a comfortable one for the quarter of experts still using it.

Even the apparently simple questions resisted standardization. Goggles to prevent eye rubbing reached only 46 percent. Asthma as a risk factor did something Delphi panels rarely show in public: support fell from 69 percent in round one to 38 percent in round two – consensus moving backward as the panel thought harder. And for acute hydrops, beyond topical steroids and pressure lowering, the panel explicitly noted that management varies by country and region – a reminder that “global consensus” must accommodate the fact that keratoconus is treated in Kigali and Kyoto under very different constraints.

What the gaps are for

Reporting disagreement is not new – the first consensus was explicit that it existed to address the field’s controversies, and said plainly where its panels were split (2). Edition 2 adds structure: the disagreements now sit in dedicated tables, panel by panel, with percentages attached, so they can be cited as precisely as the recommendations. That matters because a consensus document has two audiences. Clinicians need the recommendations; researchers need the residuals. The disagreement tables are, in effect, the field’s research agenda for the next ten years, written by 128 of its most experienced practitioners: a repeatable, severity-adjusted metric for progression to retire the Kmax paradox; longitudinal biomechanical and epithelial datasets robust enough to earn those technologies a place in treatment decisions, not just screening; comparative trials in sub-400 µm corneas; evidence to resolve the pediatric watch-or-treat split; nomograms to move CAIRS from investigational (92 percent agreed it still is) to standardized.

We wrote in the paper’s conclusion that consensus does not equate to definitive scientific truth – it is a structured convergence of expert opinion on current evidence, to be revisited, challenged, and refined. We would go further. The 2015 document’s most productive legacy was not any statement the field accepted; it was the posterior-elevation claim the field tested and overturned. If Edition 2 works as intended, the sentences that matter most are the ones that say “no consensus was reached.” Those are the invitations. We look forward to seeing which of them Edition 3 gets to delete.

Farhad Hafezi is Medical Director of the ELZA Institute, Zurich, Switzerland, and Professor of Ophthalmology at the University of Geneva. José A. P. Gomes is Professor of Ophthalmology at the Escola Paulista de Medicina/Federal University of São Paulo (EPM/UNIFESP), Brazil. Renato Ambrósio Jr is Professor of Ophthalmology at the Federal University of the State of Rio de Janeiro (UNIRIO) and of the EPM/UNIFESP, Brazil. Mark Hillen is the Director of Communications at the ELZA Institute. Hafezi, Gomes, and Ambrósio led the Global Consensus on Keratoconus and Ectatic Diseases – Edition 2 as joint first authors.