Eyeworld

FALL 2026 | EYEWORLD | 27 Contact Simpson: Rachel.Simpson@hsc.utah.edu Finally, follow-up beyond 5 years is critical. The paper itself cites 10-year CXL data, em- phasizing that the field is now mature enough to ask not only whether CXL works but how durable one treatment is over decades, especial- ly when performed in children who may have 60–70 years of life remaining after treatment. Repeat CXL rates, predictors of late progression, corneal transplant incidence, long-term contact lens dependence, patient-reported visual func- tion, and cost effectiveness would all represent meaningful endpoints. Conclusion This is a clinically useful and important con- tribution to the CXL literature, particularly because of its relatively large pediatric cohort, racial and ethnic diversity, and follow-up ex- tending to 5 years in some patients. Its stron- gest finding is that epithelium-off CXL is asso- ciated with favorable visual and tomographic outcomes across age groups, with most struc- tural change occurring during the first several postoperative years. The study's statistical design, however, makes the short- and intermediate-term findings considerably more convincing than the 5-year conclusions. Severe attrition, non-consecutive follow-up, lack of adjustment for correlation between fellow eyes, extensive multiple testing, and the absence of formal interaction testing weaken conclusions regarding long-term du- rability and superiority of response in younger patients. The lack of an untreated comparator and selection of patients based on recent Kmax progression also limit causal interpretation. Perhaps the most interesting contribution of this paper is therefore not simply another demonstration that CXL works. It identifies the questions the field should now be asking: How long does the effect last? Who fails? When is apparent postoperative progression truly fail- ure? Does age independently predict treatment response? Which patients ultimately require repeat CXL or keratoplasty? Answering those questions will require a shift from repeated cross-sectional comparisons of cohort means toward patient-level longitudinal modeling and clinically meaningful failure endpoints. Editors' note: To read the full article analysis, including thoughts on statistical weaknesses and points for discussion, go to www.EyeWorld.org. Opportunities for future exploration The most valuable next step would be a pro- spective multicenter longitudinal cohort with standardized postoperative assessments and substantially better patient retention. Given that untreated observation is increasingly difficult to justify in clearly progressive pediatric keratoco- nus, an RCT comparing CXL with no treatment may be neither necessary nor ethically practical for many patients. A carefully designed pro- spective registry may therefore provide more realistic long-term evidence. Future analyses should use mixed-effects longitudinal models capable of simultaneously accounting for repeated measurements, bilat- eral eyes, variable follow-up, baseline Kmax, age, corneal thickness, race/ethnicity, and other relevant covariates. Sensitivity analyses using inverse-probability weighting or multiple impu- tation could investigate how strongly conclu- sions depend on assumptions regarding loss to follow-up. A particularly important research question is who fails CXL. Rather than concentrating pri- marily on cohort means, future studies should examine patient-level failure using prespecified definitions such as ≥1 D or ≥1.5 D Kmax pro- gression, worsening posterior elevation or other tomographic indices, progressive thinning, loss of CDVA, retreatment, contact lens intolerance, or eventual keratoplasty. Time-to-event analysis could then identify predictors of failure. The 21 eyes showing >1 D Kmax worsening at 12 months are especially interesting. A ded- icated analysis of this subgroup could address whether apparent early failure represents true CXL failure or ongoing remodeling. The obser- vation that several eyes subsequently improved raises an important clinical question: At what point following CXL should increasing Kmax actually be labeled treatment failure and trigger repeat treatment? The age finding also warrants formal testing. A future study could model age contin- uously rather than using somewhat arbitrary categories of <18, 18–22, 23–30, and >30 years and test an age-by-treatment-time interac- tion. This could clarify whether younger corneas truly demonstrate a different response or simply begin with more active disease and therefore have more opportunity to flatten.

• Cover