Long-term outcomes of orthokeratology in eyes with myopic regression after photorefractive keratectomy (PRK): a retrospective clinical study of 12 eyes

Abstract

Background

To evaluate long-term (5-year) refractive and corneal tomographic changes induced by overnight orthokeratology (OK) in eyes with myopic regression after photorefractive keratectomy (PRK), with a focus on visual efficacy, mid-peripheral alignment-based lens-fitting methodology, and corneal safety.

Methods

This single-center retrospective study, based on data collected between January 2016 and March 2024, included 12 eyes (nine women and three men; mean age 38.6 ± 2.9 years) with myopic regression after PRK, representing the consecutive eligible patients who continued treatment out of 15 fitted (3 discontinued for compliance-related reasons), treated with specially designed overnight OK lenses for oblate corneal profiles. Sphere, cylinder, spherical equivalent (SE), uncorrected visual acuity (UCVA), anterior and posterior keratometry at 3 mm, Mean Pupil Power (MPP), and central and thinnest corneal thickness (CCT, TCT) were assessed with a Sirius tomographer (CSO, Italy). Fitting was guided by mid-peripheral alignment zone analysis, and safety was monitored at each visit with a zero-tolerance criterion for corneal fluorescein staining. Paired differences were analysed with a Shapiro-Wilk normality test followed by a paired t-test or Wilcoxon signed-rank test as appropriate; bootstrap BCa 95% confidence intervals (B = 50,000) were computed for posterior keratometry, and effect sizes were reported as Hedges’ g.

Results

Mean sphere improved from −2.65 ± 0.82 D to −0.43 ± 0.28 D (p < 0.001; g = 2.57). SE improved from −3.09 ± 0.65 D to −0.76 ± 0.17 D (p < 0.001; g = 3.15). UCVA improved from 0.67 ± 0.18 to 0.05 ± 0.03 logMAR (p < 0.001; g = 3.31); 100% of patients achieved UCVA ≤ 0.10 logMAR at five years. Three parameters (cylinder, posterior K, MPP) departed from normality; paired t-test and Wilcoxon results were fully concordant for all nine parameters. Posterior K showed a mean change of +0.032 mm (Wilcoxon p = 0.021; bootstrap BCa 95% CI +0.012 to +0.068 mm), well below the ectasia threshold of 0.10 mm. No contact lens-induced corneal complications were observed during follow-up.

Conclusions

In this small retrospective case series, overnight orthokeratology was associated with sustained refractive and visual improvement over five years in eyes with myopic regression after PRK, and no serious adverse events were observed in these 12 patients. Effect sizes for the principal refractive and visual outcomes were large to very large; the change in cylinder was small and non-significant, and the posterior keratometry change was of medium magnitude and not clinically concerning. Given the small, uncontrolled cohort, these exploratory findings require confirmation in larger, prospective studies.

Background

Myopic regression is a common and clinically significant complication of corneal photorefractive surgery. Despite technological advances in excimer lasers and improvements in the biological management of the corneal wound-healing response, the prevalence of myopic regression after photorefractive keratectomy (PRK) and laser in situ keratomileusis (LASIK) is estimated to be between 5.5 and 28% of treated cases, with incidence increasing as the preoperative refractive correction increases [1–3]. Regression was defined as a myopic shift of ≥0.50 D relative to stable postoperative refraction.

The underlying pathophysiological mechanisms are multifactorial: central compensatory epithelial hyperplasia, stromal biomechanical remodelling with anterior corneal protrusion, redistribution of lamellar collagen, and, in patients over 40 years of age, a lenticular contribution from the progressive loss of accommodative tone [4, 5]. Chayet et al. [1] attributed a central role to the progressive thickening of the epithelium at the corneal apex, which, on a profile already flattened by surgery, produces a relative corneal steepening effect with a consequent myopic shift. Subsequent tomographic studies have confirmed that this remodelling follows a lenticular distribution, with central epithelial thickening and relative thinning in the mid-peripheral zone [4, 6–8].

Managing postsurgical myopic regression poses specific challenges for clinicians. Laser re-treatment requires adequate residual stromal thickness and causes further biomechanical weakening of the cornea. Topical IOP-lowering agents, such as timolol 0.5%, produce only a modest reduction of the myopic shift—largely confined to its spherical component—and the supporting evidence comes from small trials with short follow-up [9]. In addition, ophthalmic β-blockers undergo systemic absorption and have been associated with bronchospasm and cardiovascular effects such as bradycardia, which constrains their chronic use [10]. Optical correction alone with spectacles or soft contact lenses addresses the symptoms without acting on the underlying corneal geometry.

Orthokeratology (OK) uses reverse-geometry rigid gas-permeable contact lenses worn overnight to redistribute corneal epithelial cells and reversibly modify the anterior corneal profile [11]. It represents a non-invasive, reversible, and surgery-free therapeutic option, whose efficacy in myopia control in non-operated patients is well established [12–15]. Its application to postsurgical corneas presents additional complexity related to the atypical oblate corneal profile generated by myopic ablation, which requires fitting approaches specifically adapted from those designed for naïve prolate corneas.

Park et al. [16] were the first to document the efficacy of OK in patients with myopic regression after LASIK/LASEK at 12 months. More recently, Zhang et al. [17] confirmed the safety and effectiveness of OK for myopic regression after LASIK. Both reports, however, concern eyes treated with flap-based LASIK, in which a lamellar stromal flap is created before excimer ablation; the present cohort consists exclusively of eyes treated with surface photorefractive keratectomy (PRK), in which the epithelium is removed and the anterior stroma ablated without a flap. The two techniques differ in their biomechanical and wound-healing behaviour—flap creation severs the anterior stromal lamellae and corneal sub-basal nerves, whereas surface ablation preserves the deeper stromal architecture but heals through more pronounced epithelial and stromal remodelling—so evidence obtained after LASIK cannot be assumed to apply directly to PRK-treated corneas. Frisani et al. [18] subsequently presented 5-year preliminary results in eight eyes, opening the prospect of extended long-term evaluation. The present study extends that cohort to 12 eyes, details the fitting methodology based on mid-peripheral alignment zone analysis, introduces a structured evaluation of long-term corneal safety, and applies a mixed inferential approach—combining paired t-test, Wilcoxon signed-rank test, bootstrap BCa confidence intervals, and Hedges’ g effect sizes—guided by formal normality testing with the Shapiro-Wilk statistic.

Methods

This single-center retrospective study included 12 right eyes from 12 patients (9 females, 3 males; mean age 38.6 ± 2.9 years; range 35–43 years) with documented myopic regression after primary photorefractive keratectomy (PRK) treated with overnight orthokeratology (OK). Only the right eye of each patient was analysed. Measurements obtained from the two eyes of the same subject are correlated, and combining both eyes without accounting for this inter-eye correlation violates the assumption of independence that underlies standard statistical tests, yielding falsely narrow confidence intervals and falsely small p-values [19, 20]. Restricting the analysis to one eye per subject removes this dependency; when either eye may be equally affected, analysing the right eye only, the left eye only, or a randomly selected eye is statistically equivalent [20]. We therefore selected the right eye of every patient, so that the 12 observations entering the analysis are mutually independent and no inter-eye summation bias is introduced. Clinical data were collected retrospectively over the period from January 2016 to March 2024. All data were drawn from the clinical database of Centro CL72, a private centre specialised in contact lenses (Turin, Italy), where lens fitting and all follow-up examinations were carried out. Because every eye contributed both a baseline examination and a five-year follow-up examination, this eight-year window is what made a complete five-year observation period documentable for each eye: lens fitting and baseline assessment necessarily fell in the earlier part of the window, with the corresponding five-year assessments occurring within it. During this period, 15 consecutive patients were fitted with orthokeratology lenses for myopic regression after PRK and met the eligibility criteria. Three patients (20%) discontinued treatment and were therefore not available for the five-year assessment: one because of visual dissatisfaction arising from irregular, discontinuous lens wear; one who abandoned treatment after repeated episodes of lens loss and breakage attributable to unreliable handling; and one in whom discontinuation was clinically advised because of inadequate adherence to lens and case maintenance procedures, with improperly stored lenses and storage case documented at several consecutive visits. No patient discontinued because of a corneal complication, an adverse event, or loss of efficacy of the treatment: all three discontinuations were related to compliance and lens handling. The selection of the study sample is summarised in Fig. 1. The 12 eyes reported here therefore represent all consecutive eligible patients who continued treatment, rather than a subset selected because they completed five years. All three discontinuations occurred within the first six months of treatment, during the initial adaptation and lens-refinement phase and therefore well before the five-year endpoint. The baseline characteristics of these three patients are available and are reported here for the right eye, the eye that would have entered the analysis: a 34-year-old with −1.75 −0.75 × 35° (SE − 2.13 D), a 37-year-old with −1.25 −1.00 × 5° (SE − 1.75 D), and a 31-year-old with −1.00 −1.50 × 175° (SE − 1.75 D). Compared with the 12 patients who completed follow-up, the three discontinuers were younger (mean 34.0 ± 3.0 versus 38.6 ± 2.9 years; Mann-Whitney p = 0.060) and had a significantly less myopic baseline spherical equivalent (mean −1.88 ± 0.22 versus −3.09 ± 0.65 D; Mann-Whitney p = 0.017), while baseline cylinder did not differ (−1.08 ± 0.38 versus −0.88 ± 0.73 D; p = 0.180). With only three discontinuers these comparisons are indicative rather than conclusive, but they show that the two groups were not equivalent at baseline. Eligibility and fitting were determined prospectively at the start of treatment and no eye was excluded on the basis of its five-year outcome; nevertheless, the analysed cohort consists of the patients who proved able and willing to continue treatment, and the efficacy and safety estimates reported here may be enriched by the exclusion of those who did not. Although the reasons for discontinuation were related to compliance and lens handling rather than to documented corneal complications, these three cases represent clinically relevant limitations of the treatment from a real-world effectiveness perspective. This is addressed among the study limitations. All patients received preliminary ophthalmic medical clearance before OK lens fitting and underwent at least one ophthalmologist visit annually throughout the study. During follow-up, each eye was reviewed at least every six months. At these visits the orthokeratology lens fit was reassessed and, when required, minor adjustments to lens power or geometry were made to accommodate physiological refractive changes over time or refractive re-adaptation; accordingly, the present analysis compares only the baseline and the five-year time points and does not model the intermediate visits. Throughout follow-up no patient used supplementary spectacle correction: refractive compensation was provided entirely by orthokeratology, and all eyes maintained satisfactory unaided visual function with the OK optical effect (UCVA ≤ 0.10 logMAR in every eye at five years), although a small residual myopic spherical equivalent persisted (mean −0.76 ± 0.17 D at five years). Best-corrected visual acuity over time was not recorded, because the therapeutic goal and the outcome of interest were the uncorrected visual performance achieved with OK; this is acknowledged among the study limitations. The inclusion criteria were as follows: (a) documented myopic regression (shift ≥ 0.50 D relative to stable postoperative refraction detected at least 6 months after surgery); (b) absence of post-surgical complications, including ectasia, corneal haze above grade 1, or stromal opacities; and (c) residual corneal thickness compatible with safe lens use. The exclusion criteria were suspected ectasia on topography or tomography, uncontrolled severe dry eye, and opacity of the ocular media.

Fig. 1.

Flow of patient selection. Of 15 consecutive patients fitted with overnight orthokeratology for myopic regression after PRK between January 2016 and March 2024, 3 discontinued treatment for reasons related to compliance and lens handling and contributed no five-year data; the right eye of each of the 12 patients who continued treatment was analysed. All three discontinuations occurred within the first six months of treatment, and none was due to a corneal complication, an adverse event, or loss of treatment efficacy

Before lens fitting, the following data were collected for each patient: subjective and objective refraction expressed as spherical equivalent (SE = sphere + cylinder/2), uncorrected visual acuity (UCVA) in logMAR, full corneal tomography with Sirius (CSO, Florence, Italy). From the Sirius tomography, the following parameters were extracted: anterior and posterior keratometry K1, K2, and mean K at 3 mm; Mean Pupil Power (MPP, D); central corneal thickness (CCT, µm); and thinnest corneal thickness (TCT, µm). Anterior and posterior elevation maps and differential maps were used to characterize the oblate corneal profile and for longitudinal monitoring of the corneal shape. Individual pre-PRK data (preoperative refraction, pre-surgical corneal thickness, ablation profile, and the exact timing of refractive regression and interval between PRK and initiation of OK) were not available for the eyes in this series. The patients had undergone PRK at different centres and at different times, and were subsequently referred to our contact-lens service only after myopic regression had become clinically manifest; the original surgical records were therefore not retrievable for the present retrospective analysis. This is acknowledged as a limitation (see Discussion). Each patient had, however, been individually evaluated by an ophthalmologist who judged overnight orthokeratology to be the most appropriate option for that eye, on the basis of one or more of the following clinical considerations: (i) a further PRK re-treatment was contraindicated (e.g. insufficient residual stromal bed or an unfavourable risk profile for additional ablation); (ii) the patient was unwilling to undergo a second surgical procedure because of apprehension about re-operation; (iii) spectacle correction was not a viable option because of high anisometropia or associated binocular-vision disturbances; and (iv) the patient was intolerant to, or unable to use, daytime soft or rigid gas-permeable contact lenses. Orthokeratology was thus selected as a non-surgical, reversible means of managing post-PRK myopic regression in eyes for which the alternative options were unsuitable or had failed.

The oblate profile of the post-myopic cornea renders standard fitting methods inapplicable. Central keratometry (SimK) in surgically treated eyes is artificially flattened by ablation and does not reflect the natural tissue curvature [21–23]. The fitting method adopted in this study relied on the analysis of the mid-peripheral corneal alignment zone, defined as the annular region between 3.5 mm and 5.0 mm from the corneal center. This zone lies beyond the laser ablation area in most standard myopic treatments and therefore retains the original presurgical curvature. The corneal meridian was modelled as a conic section described by Baker’s equation [24], y² = 2 r₀x − px², where y is the semichord (perpendicular distance of a corneal point from the axis of symmetry), x is the sagitta (sag) of the section at that semichord, r₀ is the apical radius of curvature (mm), and p is the asphericity (flattening/steepening) parameter, related to the corneal eccentricity e by p = 1 − e²; a value of e = 0.50 (hence p = 0.75) was used for the post-PRK oblate profile. Peripheral keratometric values at 4 mm from the center (semichord y = 2 mm) were averaged from the sagittal map of the Sirius tomographer across the four principal half-meridians to obtain the local sagitta ×=S. The apical radius that defines the starting base curve (the “Estimated K”, expressed as a radius in mm) was then recovered by inverting Baker’s equation for r₀: r₀ = (y² + pS²)/(2S). As an illustrative example, a measured sagitta of S = 0.26 mm at y = 2 mm with p = 0.75 yields r₀ = (2.00² + 0.75 × 0.26²)/(2 × 0.26) = 7.79 mm (≈43.3 D at n = 1.3375). This estimated K served as the starting point for trial-lens base curve (BC) selection. The final BC was confirmed using fluorescein pattern analysis, targeting a uniform mid-peripheral alignment band (3.5–5.0 mm), central pooling, and peripheral edge lift. The final prescribed BC was systematically flatter than the estimated K by +0.10 to +0.15 mm in all study eyes. Lens power was determined using the clinical history method with an over-refraction target of +0.25 to +0.75 D. All lenses were manufactured by MedLac (Avellino, Italy). Three high-Dk rigid gas-permeable materials were used, assigned without a predefined selection criterion: Boston XO (hexafocon A, Dk = 100 × 10^{− 1 1}), Optimum Infinite (tisilfocon A, Dk = 180 × 10^{− 1 1}), and Optimum Extra (Dk = 100 × 10^{− 1 1}) (Dk in cm² ·mL O₂ · s^{− 1} ·mL^{− 1} ·mmHg^{− 1}, ISO/Fatt method). For a mean central lens thickness of approximately 0.15 mm, the corresponding oxygen transmissibility (Dk/t) was ≈67 × 10^{− 9} for the Dk-100 materials and ≈120 × 10^{− 9} for Optimum Infinite. The reverse-geometry design comprised five zones of variable asphericity and width: a central optic (base-curve) zone of 6.0 mm; a second, reverse-curve zone of 1.0 mm width; third and fourth alignment (landing) curves; and a narrower peripheral curve at the lens edge providing tear-film clearance and lacrimal exchange beneath the lens.

The lenses were worn overnight for 8–10 h, and follow-up visits were scheduled at 1 day, 1 week, 3 weeks, and every 6 months over the 5-year study period. At each visit, UCVA, subjective refraction, biomicroscopic examination under white and cobalt blue light with sodium fluorescein 2%, corneal tomography, and fitting evaluation using the fluorescein pattern were performed. Safety was assessed using the CCLRU/EFRON grading scale with a zero-tolerance criterion: any staining of grade ≥ 1 was an immediate indication for fitting adjustment or temporary lens suspension.

Statistical analysis

The primary efficacy endpoint was the change in spherical equivalent (SE) from baseline (t0) to five years (t5); the primary safety endpoint was the change in posterior mean keratometry at 3 mm over the same interval. Secondary endpoints comprised the remaining refractive, visual, keratometric, and pachymetric parameters. The normality of paired differences (Δ = t5 − t0) was formally assessed for each of the nine study parameters using the Shapiro-Wilk (SW) test, which is considered the most powerful omnibus normality test for small samples (n < 50) [25]. The SW statistic W ranges from 0 to 1, with values close to 1 indicating conformity to a normal distribution and values below approximately 0.90 indicating a departure. The significance threshold was set at α = 0.05. We recognise that a two-stage procedure in which a preliminary normality test selects the subsequent test is open to criticism at this sample size: the Shapiro-Wilk test has low power at n = 12, so failure to reject normality is weak evidence of normality, and conditioning the choice of test on a preliminary test applied to the same data does not preserve the nominal type I error rate of the second stage. For this reason the normality screen was not used to discard either test: both the paired t-test and the Wilcoxon signed-rank test were computed for every parameter and both are reported in Table 1, and distribution-free bootstrap BCa intervals were computed for the primary safety parameter, so that no conclusion rests on the outcome of the normality screen alone. Separately from this paired analysis, the baseline characteristics of the three patients who discontinued were compared with those of the 12 completers using the Mann-Whitney U test for independent samples, reported as a descriptive check on the comparability of the two groups rather than as a hypothesis test.

Table 1.

Main refractive and corneal tomographic outcomes at baseline (t0) and after 5 years of orthokeratology (t5)

Parameter	t0 (mean±SD)	t5 (mean±SD)	Δ mean	p (t-test)	p (Wilcoxon)	|Hedges g|
Sphere (D)	−2.65±0.82	−0.43±0.28	2.22 ± 0.80	<0.001	0.001	2.57 vl
Cylinder (D)	−0.88±0.73	−0.65±0.30	0.23 ± 0.56	0.194 ns	0.204 ns	0.37 sm
SE (D)	−3.09±0.65	−0.76±0.17	2.33 ± 0.69	<0.001	0.001	3.15 vl
UCVA (logMAR)	0.67 ± 0.18	0.05 ± 0.03	−0.62±0.18	<0.001	0.001	3.31 vl
Ant. K avg 3 mm (mm)	8.44 ± 0.29	8.76 ± 0.35	0.33 ± 0.14	<0.001	0.001	2.14 vl
Post. K avg 3 mm (mm)†	6.41 ± 0.18	6.44 ± 0.17	0.03 ± 0.05	0.041*	0.021*	0.62 md
MPP (D)†	39.19 ± 1.43	37.98 ± 1.65	−1.21±0.55	<0.001	0.001	2.07 vl
CCT (µm)	478.6 ± 33.8	471.4 ± 32.5	−7.24±2.99	<0.001	0.001	2.25 vl
TCT (µm)	477.4 ± 33.8	469.9 ± 32.6	−7.57±2.86	<0.001	0.001	2.46 vl

Mean ± SD, n = 12. Both paired t-test and Wilcoxon signed-rank test p-values are reported; Wilcoxon was the primary test for † parameters (non-normal per Shapiro-Wilk). Hedges’ g: vl = very large (≥1.20); md = medium (0.50–0.79); sm = small (0.20–0.49). ns = not significant; * p < 0.05; all others p < 0.001. BCa 95% CI for posterior K: [+0.012, +0.068] mm

For parameters whose differences satisfied the normality assumption (SW p ≥ 0.05), a paired t-test was applied. For parameters whose differences were non-normally distributed (SW p < 0.05), the Wilcoxon signed-rank test was used, with exact p-values computed when feasible (n = 12). We note that the Wilcoxon test evaluates the paired distribution difference, whereas the bootstrap BCa interval estimates the mean paired difference; the two are reported as complementary rather than equivalent. For the posterior keratometry parameter, which is the safety-critical measurement for ectasia assessment and for which non-normality was confirmed, an additional bootstrap bias-corrected and accelerated (BCa) 95% confidence interval for the mean difference was computed with B = 50,000 resamples to provide a robust, distribution-free estimate of the uncertainty around the key safety conclusion.

Effect sizes were reported as Hedges’ g for paired data, computed from the paired differences as g = (meanΔ/SDΔ) × J, where J = 1 − 3/(4(n − 1) − 1) is the small-sample bias correction (i.e. the average-standardised-difference form based on the SD of the differences, appropriate for a within-subject paired design). Hedges’ g was interpreted as negligible (<0.20), small (0.20–0.49), medium (0.50–0.79), large (0.80–1.19), and very large (≥1.20) [26]. Effect sizes are reported together with the direction of change, since a large effect does not by itself indicate a beneficial change. All analyses were performed using JASP (JASP Team, 2026; JASP, Version 0.98.0). Figure 2 shows the complete panel of statistical diagnostics.

Fig. 2.

Statistical analysis panel, n = 12. A: shapiro-Wilk p-values; dashed line at α = 0.05; red = non-normal, green = normal. B: paired t-test (dark blue) vs Wilcoxon (orange) p-values; shading = non-normal parameters; ✓/✕ = concordance. C: Hedges’ g effect sizes with threshold lines. D: bootstrap BCa distribution (B = 50,000) for posterior K Δ; BCa 95% CI [+0.012, +0.068] mm vs ectasia threshold 0.10 mm. E1–E3: Q-Q plots for the three non-normal parameters (cylinder, posterior K, MPP)

Results

Normality testing

The Shapiro-Wilk test on the nine paired difference distributions identified three parameters with statistically significant departures from normality (Table 2; Fig. 2, panels A and E). Cylinder differences showed W = 0.836 (p = 0.025), with a skewness of +1.41 and excess kurtosis of +2.63, attributable primarily to the outlier in patient 7 (Δ = +1.73 D). Posterior keratometry differences showed W = 0.814 (p = 0.014), with a skewness of +1.22 and excess kurtosis of +2.61, reflecting the asymmetric distribution of small positive values with one negative outlier (patient 6, Δ = −0.051 mm). The MPP differences showed W = 0.858 (p = 0.046) with a skewness of −1.07. The remaining six parameters (sphere, SE, UCVA, anterior keratometry, CCT, and TCT) satisfied the normality assumption, with SW p-values ranging from 0.132 to 0.674.

Table 2.

Shapiro-Wilk normality test results for paired differences Δ (t5 − t0), n = 12

Parameter	SW W	SW p	Normal?	Recommended test	Notes
Sphere (D)	0.9171	0.263	YES	Paired t-test	—
Cylinder (D)	0.8361	0.025	NO	Wilcoxon	Skewness +1.41; outlier Pt.7 (Δ=+1.73 D)
SE (D)	0.8987	0.153	YES	Paired t-test	—
UCVA (logMAR)	0.8939	0.132	YES	Paired t-test	—
Ant. K avg 3 mm (mm)	0.9055	0.187	YES	Paired t-test	—
Post. K avg 3 mm (mm)	0.8142	0.014	NO	Wilcoxon + Bootstrap BCa	Skewness +1.22; safety-critical
MPP (D)	0.8580	0.046	NO	Wilcoxon	Skewness −1.07; borderline
CCT (µm)	0.9525	0.674	YES	Paired t-test	—
TCT (µm)	0.9288	0.368	YES	Paired t-test	—

W closer to 1 indicates a more normal distribution; p < 0.05 indicates rejection of normality. The recommended inferential test is listed for each parameter, with notes on the nature of any departure from normality

Refractive parameters and visual acuity

After 5 years of overnight orthokeratology treatment, the spherical components of refraction showed statistically significant improvements, whereas the cylinder did not change significantly (Table 1; Figs. 3, 4 and 5). The mean sphere improved from −2.65 ± 0.82 D at baseline to −0.43 ± 0.28 D at 5 years (paired t-test: t = −9.56, p < 0.001; Wilcoxon: p = 0.001; Hedges g = 2.57). The spherical equivalent decreased from −3.09 ± 0.65 D to −0.76 ± 0.17 D (paired t-test: p < 0.001; g = 3.15). The cylinder showed no statistically significant change in either test (paired t-test: p = 0.194; Wilcoxon: p = 0.204; g = 0.37, small effect), with full concordance between the two methods despite the non-normal distribution of the differences.

Fig. 3.

Box-and-whisker plot of UCVA (logMAR) at baseline (t0, dark blue) and after 5 years (t5, orange), n = 12. Y-axis inverted (lower = better). Boxes = IQR; white lines = medians; whiskers = 1.5 × IQR; white circles = individual data (jittered); gray lines = paired observations. Red dashed line: 0.10 logMAR (20/25). Paired t-test p < 0.001; Wilcoxon p = 0.001; Hedges g = 3.31. 100% of patients achieved UCVA ≤ 0.10 logMAR at t5

Fig. 4.

Corneal curvature parameters at t0 and t5 (individual patients: gray lines; group means: black). Left: anterior K avg 3 mm—significant flattening (p < 0.001; g = 2.14). Center: posterior K avg 3 mm—a small but statistically significant increase in radius, i.e. slight flattening rather than the steepening seen in ectasia; Wilcoxon p = 0.021; BCa 95% CI [+0.012, +0.068] mm; g = 0.62, with the upper bound well below the 0.10 mm ectasia threshold. Right: MPP—Wilcoxon p < 0.001; g = 2.07

Fig. 5.

Corneal thickness at t0 and t5. Left: CCT (p < 0.001; g = 2.25). Right: TCT (p < 0.001; g = 2.46). Reductions are consistent with centrifugal epithelial redistribution; the absence of posterior steepening is consistent with an epithelial rather than an ectatic mechanism

The mean UCVA improved from 0.67 ± 0.18 logMAR to 0.05 ± 0.03 logMAR (paired t-test: t = 12.32, p < 0.001; Wilcoxon: p = 0.001; g = 3.31, very large effect size). All 12 patients (100%) achieved UCVA ≤ 0.10 logMAR at 5 years. The box-and-whisker plot in Fig. 3 illustrates the shift in the distribution: the interquartile range at 5 years spans only 0.02–0.07 logMAR, compared with 0.51–0.83 logMAR at baseline, demonstrating a homogeneous treatment response across the cohort.

Keratometric and corneal thickness parameters

The anterior mean keratometry at 3 mm flattened from 8.44 ± 0.29 mm to 8.76 ± 0.35 mm (paired t-test: p < 0.001; g = 2.14). The MPP decreased from 39.19 ± 1.43 D to 37.98 ± 1.65 D. For MPP, whose differences were non-normally distributed, the Wilcoxon test (p < 0.001) was concordant with the paired t-test (p < 0.001), and the very large effect size (g = 2.07) confirmed the clinical relevance of this result. CCT decreased by 7.24 ± 2.99 µm (paired t-test: p < 0.001; g = 2.25) and TCT by 7.57 ± 2.86 µm (p < 0.001; g = 2.46). Both findings are consistent with centrifugal epithelial redistribution [27–30] (Figs. 4 and 5; Tables 3 and 4).

Table 3.

Individual patient data: sphere (Sph), spherical equivalent (SE), and UCVA (logMAR) at baseline (t0) and 5-year follow-up (t5)

Pat.	Sex	Age	Sph t0 (D)	Sph t5 (D)	SE t0 (D)	SE t5 (D)	UCVA t0	UCVA t5
1	F	35	−2.66	−0.68	−3.03	−0.90	0.46	0.06
2	F	36	−3.68	−0.50	−3.94	−0.68	0.82	0.02
3	M	43	−2.77	−0.15	−3.38	−0.56	0.74	0.04
4	F	40	−3.53	−0.76	−4.02	−1.00	0.84	0.08
5	F	35	−3.42	−0.24	−3.53	−0.66	0.80	0.02
6	F	36	−2.38	−0.85	−2.50	−1.06	0.90	0.10
7	M	43	−0.74	−0.12	−2.23	−0.75	0.34	0.06
8	F	40	−2.52	+0.04	−3.00	−0.62	0.48	0.00
9	F	38	−2.10	−0.45	−2.41	−0.70	0.56	0.04
10	M	37	−3.15	−0.55	−3.59	−0.86	0.70	0.06
11	F	41	−1.88	−0.62	−2.11	−0.81	0.62	0.08
12	F	39	−2.95	−0.30	−3.31	−0.58	0.78	0.02

Patients 1–8: original cohort; Patients 9–12: extended cohort

Table 4.

Individual patient data: mean anterior and posterior keratometry at 3 mm (mm), mean Pupil power (MPP, D), central corneal thickness (CCT, µm), thinnest corneal thickness (TCT, µm), and refractive cylinder (cyl, D) at baseline (t0) and at 5-year follow-up (t5)

Pat.	K ant t0 (mm)	K ant t5 (mm)	K post t0 (mm)	K post t5 (mm)	MPP t0 (D)	MPP t5 (D)	CCT t0 (µm)	CCT t5 (µm)	TCT t0 (µm)	TCT t5 (µm)	Cyl t0 (D)	Cyl t5 (D)
1	8.187	8.379	6.283	6.322	40.28	39.71	506.3	501.2	505.7	498.9	−0.74	−0.44
2	8.968	9.275	6.549	6.593	36.66	35.81	429.7	422.8	428.4	419.8	−0.51	−0.36
3	8.644	9.259	6.635	6.692	37.84	35.66	510.2	499.5	510.2	498.5	−1.22	−0.82
4	8.449	8.774	6.161	6.171	39.08	38.05	407.4	404.6	407.1	404.2	−0.98	−0.48
5	8.001	8.362	6.306	6.327	41.43	40.11	497.6	494.7	497.1	494.2	−0.21	−0.84
6	8.700	8.835	6.611	6.560	38.21	37.33	457.9	449.9	455.3	447.0	−0.23	−0.41
7	8.646	9.211	6.581	6.625	37.91	35.57	518.6	505.7	516.8	504.7	−2.98	−1.25
8	8.013	8.379	6.074	6.233	41.09	39.77	473.2	463.0	472.8	463.0	−0.96	−1.16
9	8.332	8.567	6.382	6.395	39.84	38.93	490.1	483.2	488.2	481.4	−0.62	−0.50
10	8.610	8.941	6.549	6.566	38.52	37.21	503.9	496.3	502.3	494.6	−0.88	−0.62
11	8.450	8.666	6.461	6.477	39.34	38.61	467.5	461.1	465.8	459.2	−0.46	−0.38
12	8.245	8.513	6.322	6.340	40.13	39.02	481.3	474.6	479.6	472.8	−0.72	−0.55

The posterior mean keratometry at 3 mm showed a mean change of +0.032 ± 0.048 mm. Because posterior keratometry is expressed here as a radius of curvature, a positive change denotes flattening of the posterior surface, whereas ectatic progression is characteristically associated with posterior steepening (a decrease in radius); the observed direction of change is therefore opposite to that expected in ectasia. Because the SW test identified a non-normal distribution for this parameter (W = 0.814, p = 0.014), the Wilcoxon signed-rank test was used to test the paired distribution difference (W-statistic = 10.0, p = 0.021), and the bootstrap BCa 95% confidence interval was used separately to estimate the mean difference (B = 50,000 resamples; z₀ = 0.067, acceleration = 0.059), yielding [+0.012, +0.068] mm (Fig. 2, panel D). Both bounds are below 0.10 mm, a value we adopt as an indicative reference for a clinically relevant posterior-radius change [31]. This 0.10 mm value is used as a pragmatic reference rather than a validated ectasia threshold: it comfortably exceeds the measurement noise of posterior corneal radius obtained with dual Scheimpflug–Placido tomography of the type used here, for which the reported intrasubject repeatability (within-subject standard deviation) of the posterior radius is of the order of 0.05–0.06 mm [31], so a change below 0.10 mm cannot be reliably distinguished from test–retest variability. One eye (patient 8) showed a larger posterior-radius increase, from 6.074 to 6.233 mm (Δ = +0.159 mm); this represents flattening rather than steepening, and the corresponding anterior and posterior elevation/tomography maps for this eye showed no sign of instability—no localised posterior elevation, no inferior steepening, and no progressive thinning were detected over the follow-up. Given its magnitude and direction (a flattening of the posterior radius), and the absence of any corresponding tomographic change, this isolated variation is most plausibly attributable to measurement (instrumental) variability rather than to a true structural change of the posterior surface: it is of the same order as the intrasubject repeatability of posterior-radius measurement with Scheimpflug–Placido tomography [31] and is opposite in direction to the posterior steepening that would characterise ectatic progression. Taken together, and given the medium effect size (Hedges g = 0.62), we found no clinical or tomographic evidence of ectasia in this cohort over five years, although the small sample precludes any firmer conclusion.

Concordance between paired t-test and Wilcoxon

The two statistical methods yielded fully concordant conclusions for all nine parameters; in every case, both testseither rejected or failed to reject the null hypothesis at α = 0.05 (Fig. 2, panel B). This concordance is reassuring as it indicates that the paired t-test conclusions reported in previous versions of this analysis were not materially affected by the departures from normality identified for the three parameters. The Wilcoxon p-values were consistently at least as small as the corresponding t-test p-values for parameters with large effect sizes and were comparable for the two non-significant parameters (cylinder: Wilcoxon p = 0.204 vs. t-test p = 0.194; posterior K: Wilcoxon p = 0.021 vs. t-test p = 0.041).

Effect sizes

Hedges’ g values ranged from 0.37 (cylinder, small) to 3.31 (UCVA, very large) (Fig. 2, panel C). Seven of the nine parameters showed very large effect sizes (|g| ≥1.20): sphere (2.57), SE (3.15), UCVA (3.31), anterior keratometry (2.14), MPP (2.07), CCT (2.25), and TCT (2.46). Posterior keratometry showed a medium effect (g = 0.62), which is clinically informative, indicating a reliably detectable but small-magnitude change, consistent with its role as a safety marker, rather than a therapeutic target. The cylinder showed a small effect (g = 0.37), consistent with the lack of statistical significance.

Safety and corneal tolerance

No patient developed contact lens-induced corneal complications during the 5-year follow-up. No episodes of microbial or Acanthamoeba keratitis, sterile or symptomatic corneal infiltrates, stromal edema, peripheral corneal neovascularization, permanent corneal profile deformation (warpage), lens-induced ectasia, mechanical intolerance or abrasion, or conjunctival changes were observed. The zero-tolerance criterion was operationally effective, and all instances of borderline fluorescein pooling in the alignment zone resolved at the subsequent visit following fitting modifications, without permanent lens discontinuation.

Corneal topography and lens-fitting findings

Figure 6 shows the anterior corneal elevation map of a representative study eye before OK lens application. The map displays the elevation in micrometres relative to the best-fit reference sphere (Rbf = 8.13 mm, 41.54 D; e = 0.00). An oblate post-PRK profile was visible, with a central low elevation within the ablation zone surrounded by a prominent mid-peripheral elevation ring. The two red circles delineate the mid-peripheral alignment zone (3.5–5.0 mm), which lies beyond the ablation area, retains the original curvature, provides Ra for the estimated K calculation, and is where the OK lens landing curve must achieve uniform alignment.

Fig. 6.

Anterior corneal elevation map before orthokeratology (Sirius, CSO; µm scale; rbf = 8.13 mm, 41.54 D; e = 0.00). Oblate post-PRK profile: central low-elevation zone (ablation area) surrounded by a mid-peripheral elevation ring. Red circles: mid-peripheral alignment zone (3.5–5.0 mm) used for Ra measurement and estimated K calculation, which is the anatomical basis for base curve selection

Figure 7 shows the fluorescein pattern of the OK lens in the same eye after BC adjustment. The three-zone pattern of correct oblate fitting was present: central fluorescein pooling (lacrimal reservoir generating hydrodynamic pressure for epithelial redistribution), uniform mid-peripheral alignment band (narrow, regular, consistent across all meridians, confirming correct sagittal depth), and peripheral edge lift (permitting tear exchange). The absence of fluorescein staining, bubbles, and lens decentration were mandatory fitting endpoints in all patients.

Fig. 7.

Fluorescein pattern of the OK lens on a post-PRK oblate cornea (cobalt blue slit lamp; sodium fluorescein 2%). Three zones of correct fitting: (1) central pooling (0–4 mm)—lacrimal reservoir; (2) uniform mid-peripheral alignment band (3.5–5.0 mm)—correct sagittal depth, no staining, no bubbles; (3) peripheral edge lift (>5 mm). Final BC ≈0.12 mm flatter than estimated K

Figure 8 presents the tangential curvature maps before and after 5 years of OK and the differential map (t5 − t0). The bull’s - eye pattern on the differential map—a central blue-gray zone of radius increase (flattening) surrounded by an orange-red annular ring of radius decrease (steepening from epithelial accumulation)—is the tomographic signature of the OK treatment effect and provides visual confirmation that the refractive and keratometric improvements measured at 5 years reflect a stable, reversible epithelial process confined to the anterior surface.

Fig. 8.

Tangential curvature maps and differential map (sirius, CSO; n0 = 1, n1 = 1.3375; mm scale). Upper left: before OK—oblate post-PRK profile. Upper right: after 5 years—further central flattening. Bottom (t5 − t0): bull’s eye pattern—central zone of radius increase (flattening) surrounded by orange-red ring (radius decrease, epithelial accumulation). Inner boundary ≈4–5 mm, corresponding to the OK treatment zone edge. No posterior surface involvement detected

Discussion

The results of this 12-eye case series are consistent with and extend our 8-eye preliminary data [18], and suggest that overnight orthokeratology with specially designed lenses for oblate corneal profiles may provide effective and stable management of myopic regression after PRK. The mean spherical equivalent reduction of 2.33 D and the improvement in UCVA to 0.05 logMAR, with 100% of these patients below the 0.10 logMAR threshold at 5 years, represent clinically relevant outcomes in this selected cohort, comparable to those reported by Park et al. [16] at 12 months and here observed over a 5-year period. Notably, both Park et al. [16] and the recent series by Zhang et al. [17] evaluated OK in eyes with regression after LASIK, whereas the present cohort comprises exclusively eyes treated with surface PRK. Because flap-based LASIK and flapless surface ablation differ in corneal biomechanics, residual stromal architecture, and the epithelial–stromal wound-healing response, the outcomes observed here in PRK-treated eyes add evidence in a surgically distinct population and over a longer follow-up than previously reported, although the small, uncontrolled design limits generalisability.

The adoption of a mixed inferential strategy, guided by formal Shapiro-Wilk normality testing rather than assuming normality a priori, represents a methodological strengthening relative to the use of the paired t-test alone. With n = 12, the central limit theorem cannot guarantee the approximate normality of the sampling distribution of the mean, making distributional assumptions nontrivial. The finding that three of the nine difference distributions were significantly non-normal (cylinder, posterior K, MPP) validates this caution. The full concordance between the t-test and Wilcoxon results across all nine parameters indicates that the conclusions do not depend on which of the two tests is applied, and this concordance, rather than the normality screen itself, is what supports them. The two-stage procedure has recognised limitations at this sample size, as noted in the Methods: the Shapiro-Wilk test has low power at n = 12, and selecting a test on the basis of a preliminary test applied to the same data distorts the nominal error rate of the second stage. Reporting both tests for every parameter, together with distribution-free bootstrap intervals for the primary safety parameter, is therefore the more defensible reading of these data, and the p-values should be interpreted as descriptive of an exploratory analysis rather than as confirmatory inference.

Bootstrap BCa analysis of the posterior keratometry parameter is the most important addition from a clinical perspective. A p-value of 0.021 from the Wilcoxon test alone would not directly answer the clinical question of whether the observed change was below the ectasia threshold. The BCa confidence interval [+0.012, +0.068 mm] does answer this directly: even the upper bound of our best estimate of the true mean change is 32% below the 0.10 mm threshold for ectasia. This provides a level of evidence for the safety conclusion that neither a point estimate nor a p-value alone could deliver and is robust to the non-normality of the posterior K-distribution.

Effect size analysis using Hedges’ g adds a dimension of clinical interpretation that is independent of sample size. The very large effect sizes for the refractive, visual, and keratometric parameters (g = 2.07–3.31) indicate that the observed improvements are not merely statistically detectable artifacts of the paired design but represent changes of substantial magnitude relative to within-subject variability. The medium effect for posterior keratometry (g = 0.62) is equally informative; it shows that the posterior K change, while real and detectable, is of limited magnitude, which is exactly the expected pattern for a safe treatment with no ectatic effect.

The fitting method based on mid-peripheral alignment zone analysis remains the central technical innovation that makes OK applicable to postsurgical oblate corneal treatment. The systematic prescription of a BC flatter than the estimated K is a non-trivial detail that, if neglected, results in mechanical compression, corneal staining, and treatment failure. The zero-tolerance safety protocol demonstrated that this standard is achievable and effective in postsurgical eyes over a 5-year period.

This study has several important limitations. It is retrospective, uncontrolled, and based on a very small cohort (n = 12, one eye per patient), which limits statistical power and generalisability and precludes causal inference. The 12 analysed eyes were the consecutive eligible patients who continued treatment out of 15 fitted during the study period; the 3 patients who discontinued did so for reasons related to compliance and lens handling rather than to complications or loss of efficacy, and contributed no five-year data. All three withdrew within the first six months, so the analysed cohort is properly described as the subset of patients who tolerated and completed the initial adaptation phase. Their baseline data show that the two groups were not comparable: the discontinuers were younger and had significantly less myopic regression at baseline (SE − 1.88 ± 0.22 versus −3.09 ± 0.65 D; Mann-Whitney p = 0.017). The direction of this imbalance matters for how the results should be read. The patients who withdrew were those with the least regression to correct, and therefore the least to gain from nightly lens wear, which is a plausible reason for the compliance-related withdrawals; the consequence is that the analysed cohort is enriched for more severe regression rather than for mild, easily corrected eyes, so the efficacy estimate is unlikely to be inflated by retention of the easiest cases. What remains true is that every efficacy and safety estimate reported here derives from patients who tolerated treatment, and that these estimates therefore describe continuing wearers rather than all patients who begin OK for post-PRK regression. Although no eye was excluded on the basis of its outcome, the absence of complications and the finding that 100% achieved UCVA ≤ 0.10 logMAR are derived from continuing wearers, and may therefore overestimate the efficacy and safety that would be observed across all patients who begin OK for post-PRK regression, in whom an approximately 20% discontinuation rate should be anticipated. There was no non-treated control group. In addition, individual pre-PRK surgical data (preoperative refraction, pre-surgical corneal thickness, ablation profile, and the precise timing of regression) were not available, because patients were referred from different surgical centres only after regression had become manifest; the pre-treatment baseline of each eye could therefore not be characterised. This is a substantial interpretive limitation rather than a minor omission. Without the preoperative refraction, the pre-surgical corneal thickness, and the ablation profile, the depth of tissue removed and the residual stromal bed of each eye are unknown, so neither the starting corneal architecture nor the severity of the original myopia can be described. In consequence, the between-patient variability observed in refractive response and in corneal thickness cannot be attributed to the orthokeratology treatment as distinct from differences in the original surgery, and the cohort cannot be compared with series in which pre-surgical data are available. Any interpretation of individual variability in this series should be read with that constraint in mind. Optical quality was not assessed with best-corrected visual acuity, higher-order aberrometry, contrast sensitivity, or patient-reported outcomes, so no claim of improved optical quality is made. Direct epithelial thickness was not measured with anterior-segment OCT, and corneal biomechanics were not measured directly. Prospective, controlled studies with larger samples, epithelial mapping, biomechanical assessment, and higher-order aberrometry are needed to confirm these exploratory findings.

Conclusions

In this exploratory retrospective case series, overnight orthokeratology with lenses specifically designed for oblate corneal profiles was associated with sustained refractive and visual improvement over five years in eyes with myopic regression after PRK, and no serious adverse events were documented in these 12 patients. The main observations were: (1) a mean SE reduction of 2.33 D, with 100% of this cohort achieving UCVA ≤ 0.10 logMAR at five years; (2) a mid-peripheral alignment-based fitting approach using an estimated K derived from peripheral keratometry, which allowed reproducible adaptation to the post-surgical oblate profile; (3) a small but statistically significant increase in posterior corneal radius (+0.032 ± 0.048 mm), that is, a slight flattening rather than the steepening that characterises ectatic progression, with no clinical or tomographic sign of ectasia in this cohort; and (4) a zero-tolerance policy towards corneal fluorescein staining, associated with a favourable safety profile over five years. Given the small, uncontrolled sample, these findings should be regarded as preliminary and require confirmation in larger, prospective, controlled studies.

Abbreviations

OK

Orthokeratology

PRK

Photorefractive keratectomy

LASIK

Laser in situ keratomileusis

UCVA

Uncorrected visual acuity

logMAR

Logarithm of the minimum angle of resolution

SE

Spherical equivalent

K

Keratometry

MPP

Mean Pupil Power

CCT

Central corneal thickness

TCT

Thinnest corneal thickness

BC

Base curve

SW

Shapiro-Wilk

BCa

Bias-corrected and accelerated

CI

Confidence interval

CSO

Costruzione Strumenti Oftalmici

Author contributions

MF and AR conceived and designed the study. MF and CS collected the clinical data. MF and CC performed the statistical analysis and interpreted the data. MF and AR drafted the manuscript. CS and CC critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

https://docs.google.com/spreadsheets/d/1D4YiCglXf21A1DcFXcfcp9fJZCoDLmnK/edit

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the tenets of the Declaration of Helsinki (1964) and its later amendments. It was a retrospective analysis of anonymized clinical records drawn from the clinical database of a private optics and optometry centre specialised in contact lenses (Centro CL72, Turin, Italy). The data were generated in the course of routine clinical (not academic) activity: each patient was referred for and underwent overnight orthokeratology on the basis of a medical/ophthalmological indication, and was then followed at the centre with at least one contact-lens control every six months and at least one ophthalmological examination per year throughout the study period. The study was not submitted to, and did not require approval by, an institutional review board or ethics committee, and no approval or exemption reference number was therefore issued. The specific legal basis is as follows. In Italy, mandatory ethics-committee review is established for clinical trials of medicinal products by Legislative Decree 24 June 2003, no. 211 (implementing Directive 2001/20/EC) and, since 31 January 2022, by Regulation (EU) 536/2014, and for clinical investigations of medical devices by Articles 62 and 74 of Regulation (EU) 2017/745. The present study falls outside all of these instruments: no medicinal product was administered, and the orthokeratology lenses, which are CE-marked medical devices, were prescribed exclusively within their authorised intended purpose as part of routine clinical care, the decision to fit each patient having been taken on clinical grounds and independently of any research question, so that no clinical investigation within the meaning of Regulation (EU) 2017/745 was performed. The Italian Medicines Agency guidelines on observational studies (AIFA Determination of 20 March 2008, Official Gazette no. 76 of 31 March 2008, superseded by the AIFA guideline in force since 21 August 2024), which require notification of retrospective observational studies to the competent ethics committee, apply to observational studies on medicinal products and are therefore not applicable to this study. What does govern the present analysis is data-protection law. The secondary use of already-collected clinical data for scientific research is regulated by Regulation (EU) 2016/679 (Articles 5, 9(2)(a) and 89), by Legislative Decree 30 June 2003, no. 196 as amended by Legislative Decree 10 August 2018, no. 101, in particular Article 110 on processing for purposes of medical, biomedical and epidemiological research, and by the deontological rules for processing for statistical or scientific research purposes adopted by the Italian Data Protection Authority (Garante per la protezione dei dati personali, Resolution no. 515 of 19 December 2018, Official Gazette no. 11 of 14 January 2019, Annex A.5 to the Code). Every patient had signed, at the time of care, a written consent to the orthokeratology treatment together with an authorization to use their clinical data for scientific research purposes; the condition of Article 9(2)(a) of the Regulation is therefore met by explicit consent, and the records were anonymized before analysis, with full protection of personal data in accordance with the applicable Italian privacy legislation.

Consent for publication

Not applicable in the sense of identifiable personal data: the manuscript does not contain any individual participant’s identifiable information, and the corneal topography, elevation, and fluorescein images shown do not permit identification of individual participants. As stated above, each patient had authorized the use of their anonymized clinical data for scientific research purposes.

Competing interests

MF and CS are affiliated with Centro CL72, the contact lens centre where the patients were treated and from whose clinical records the data analysed here were drawn. This affiliation is disclosed because the centre provides orthokeratology as a clinical service. The centre had no role in the design of the analysis, in the interpretation of the data, or in the decision to submit the manuscript, and no author received any payment for the preparation of this work. The authors declare no other competing interests.