Apparent cone displacement after epithelium-off corneal cross-linking: radial coordinate change, two-dimensional variability and regression to the mean in a retrospective two-centre cohort study

Background: After corneal cross-linking (CXL), the cone sometimes appears to move on follow-up maps. Whether this reflects systematic displacement or measurement-related variation is unknown. We quantified radial and two-dimensional change in four vertex-referenced tomographic landmarks after epithelium-off CXL and tested whether apparent centralisation reflects regression to the mean.Methods: Retrospective two-centre cohort of 378 eyes (245 patients) with Sirius Scheimpflug-Placido tomography before CXL and at 18 months; 294, 331, 334, 161 and 50 eyes also had 3-, 6-, 12-, 24- and 36-month examinations. The primary outcome was change in radial distance from the corneal vertex at 18 months (Δr) for the Kmax apex, thinnest point and anterior and posterior elevation apexes, estimated with patient-clustered generalised estimating equations (GEE) and Holm correction. Secondary outcomes were two-dimensional displacement and its direction. Regression to the mean was assessed with an independent baseline (change from month 3 to 18, classified by month 0). Same-session repeatability was estimated from three repeat 18-month scans in 73 eyes.Results: Kmax decreased by 1.45 ± 1.18 D at 18 months. Mean Δr ranged from +0.001 to +0.020 mm; no landmark changed significantly after Holm correction (adjusted P ≥ 0.156). Median two-dimensional displacement ranged from 0.085 mm (thinnest point) to 0.288 mm (Kmax apex), without a preferred direction towards the vertex. Median distances between same-session repeat scans (0.091–0.278 mm) were close to these values, and displacement from 18 to 24 months was similar to that from 0 to 18 months. Conventional analysis suggested centralisation of peripheral cones (Kmax apex ρ = −0.28), but this association was not observed with an independent baseline (ρ = −0.07, 95% CI −0.18 to +0.04).Conclusions: Kmax decreased after epithelium-off CXL without a statistically detectable group-level radial change in the vertex-referenced coordinates of four tomographic landmarks. Individual coordinate changes of up to several tenths of a millimetre were common, with no consistent direction towards the vertex. Apparent centralisation of peripheral cones was compatible with regression to the mean. A single follow-up map should not be interpreted as cone migration.Keywords: keratoconus; corneal cross-linking; cone location; thinnest point; Scheimpflug tomography; regression to the mean.