CXL has modified the management of keratoconus and has been proven effective in both adults and children. The standard Dresden protocol and accelerated protocols have shown encouraging results in both age groups. More recently, protocols such as the Custom Fast protocol have been developed. This protocol uses software based on a mathematical model derived from the Lambert-Beer law in combination with the temporal law of rate of consumption of riboflavin, allowing efficient customization of the procedure. In the present study, we report the long-term results of CXL on a large cohort of patients affected by keratoconus. As follow-up data was not homogenous, we divided patients into two groups. A total of 886 eyes had completed at least one year of follow-up and were included in the safety analysis, which was the primary objective of the study, whereas 610 eyes with at least 3 years of data were included in the visual outcome analysis.
The average age of the patients was 22.48 years, 682 patients were male and 204 were female, with a male to female ratio of 3.3 to 1, in line with other studies. In our study, both uncorrected and corrected visual acuity minimally improved three years after CXL; however, this finding is of limited clinical significance. Kmax decreased significantly by a mean of 1.30 D after three years, indicating flattening of the conical corneal shape after treatment. These results are in line with previous studies. Pachymetry of the thinnest corneal point decreased from 467.75 microns to 463.45 microns after three years. Corneal thickness was measured using an ultrasound probe in 56 of the 610 eyes, comprising the refractive outcome subgroup, whereas the remaining eyes underwent OCT pachymetry. These findings are in line with other studies that describe a long-term stability of pachymetry, whereas in the early postoperative period, a slight reduction in thickness may be observed, likely related to the compacting effect on collagen fibers and epithelial remodeling. Since pachymetry measurements were performed using different instruments, the significance of data deriving from the present study is limited. Demarcation line values were similar to those reported by other groups and were not significantly correlated with reduction in Kmax, as reported by other studies. The demarcation line depth was lower than average, but comparable to a previous study of our group, perhaps indicating excessive riboflavin imbibition during CXL.
Long term follow-up data with a complete 10 year follow-up was limited, and available for only 35 patients. Data analysis revealed stability of topographic parameters and mean visual acuity, with three patients (8.6 percent) showing signs of progression, which will be discussed in the next paragraphs. These promising findings are comparable to the results of previous studies with a 10 year follow-up, indicating a long-term effectiveness of CXL. Newer data with ten year follow-ups will become available, and continue to add evidence regarding the long-term success of CXL in stopping progressive keratoconus.
Previous studies in different countries have reported infectious keratitis as a rare event, with three large retrospective series stating percentages of 0.17 percent, 0.5 percent, and 1.3 percent. In our series, no infections occurred. This may be determined by the temperate climate in our country, and to the immunosuppressive status of our cases: no subjects were under therapy with chronic topical steroids or oral immunosuppressive therapy. Furthermore, precise instructions were given to every patient, stressing the importance of following the correct postoperative therapeutical regimen, the perils of infectious keratitis, and risk factors related to inappropriate contact lens wear.
In the work by Shetty et al., the four subjects with keratitis had associated risk factors such as vernal conjunctivitis and oral immunosuppressive agents. Maharana et al. reported several cases of keratitis after accelerated CXL and highlighted that four of the cases had vernal keratoconjunctivitis. Our patient population had 3 percent of cases with vernal conjunctivitis and atopy, but ocular surface inflammation was minimal, thus justifying the absence of postoperative keratitis. In some cases, prior to surgery, a short course of steroids was utilized to reduce ocular surface inflammation. No short-term surgical complications were observed, other than presence of sterile infiltrates and temporary dry eye. We reported sterile infiltrates in 10 eyes (1.13 percent), compatible with findings from other studies. The presence of corneal haze differs greatly in different studies: different groups have noted rates of 9 percent, 9.1 percent, and 100 percent. These important variances may be due to inter-observer differences, and may likely be underestimated because of limited impact on vision. Haze after CXL has diverse clinical characteristics compared to haze after other procedures, such as photorefractive keratectomy used for refractive surgery. After CXL, a superficial change in the corneal stroma is observed, whereas after excimer laser ablation, it adopts a reticulated subepithelial pattern, suggesting that the mechanisms involved are likely different. The risk of haze after CXL has been reported to be greater in patients with lower pachymetry and steeper corneas. We reported the presence of mild trace haze in all of our cohort, in line with the study by Wittig-Silva et al. Haze disappeared from 6 to 12 months after treatment, and did not impact on visual acuity results. However, more specific tests, such as contrast sensitivity testing, were not performed, and could have been altered in the first months after CXL.
Eleven eyes (1.24 percent) had a reduction of CDVA of one or two lines, which was a lower rate compared to data reported by Kymionis et al. and Koller et al. In nine cases, a loss of visual acuity was related to the formation of anterior corneal stromal opacity and scarring, whereas in two atopic patients, vision decreased due to inflammation and epithelial irregularity. In all of these cases, there were no clear signs of keratoconus progression. Anterior corneal scarring may be difficult to distinguish from corneal haze in some eyes, contributing to a potential underestimation of this complication.
Five eyes (0.82% 5/610) showed signs of keratoconus progression, which was defined as an increase in Kmax by more than one diopter after treatment. Two patients had documented progression, but were not retreated due to corneal pachymetry with a value less than 400 micrometers. The patients are currently satisfied with their visual acuity, which has remained stable, and are being monitored. Three patients with progression after initial CXL were retreated using the same Dresden protocol technique, without intra- or postoperative complications. All patients that were retreated showed clinical and topographic stability after at least five years of follow-up. When analyzing the possible risk factors for progression in these five patients, we hypothesized that younger age could play a role. Three of the five patients were minors, and all patients were younger than the mean age of the study population. Furthermore, two patients had suboptimal demarcation lines (153 and 174 micrometers), which may be considered as an additional risk factor. The percentage of patients with keratoconus progression in our series was relatively low when compared to other studies. Since our refractive analysis data regards patients with at least three years of follow-up, it is possible that the percentage of CXL failure is underestimated, and failure rates may increase with longer follow-up data. For example, when analyzing the relatively few patients that completed 10 years of follow-up, the rate of progression was greater (8.6%) but likely overestimated when considering that mean age was lower (19.09 ± 4.63), and the adult/minor ratio was lower compared to the initial population (2.09 vs. 5.76). Kmax is the most commonly used parameter to evaluate progression and CXL efficacy, but it has the main limitation of representing a small area of anterior curvature without taking into account thickness and posterior corneal measurements. In our experience, Kmax values can fluctuate significantly after CXL, adding further confusion. Twelve eyes experienced continuous corneal flattening (1.97%, 12/610), which, in this study, was defined as corneal flattening greater than three diopters that persisted after three years of follow-up. The term “standard corneal flattening” has been used to indicate typical flattening which may occur in the first years after the procedure and tends to stabilize after three years. However, in some cases, continuous corneal flattening may occur even after three years, and has been reported as a long-term complication of cross-linking with a prevalence of 6.3%. We identified twelve eyes in which corneal flattening occurred after the three year time point. In all cases, Kmax values were greater than 55 diopters, which is a general risk factor for corneal flattening. This percentage may be significantly affected by the retrospective study design, and only long-term prospective studies will reveal the actual rates of this late complication.
The principal weakness of the study is the lack of full data for all follow-up intervals, which potentially underestimates the number of complications. More than half of the patient population had completed three year post-operative examinations and were included in the visual outcome analysis. All patients (886 selected eyes) had completed at least one year of follow-up and were included in the safety analysis. This subgroup division was chosen because reported complications of CXL usually occur in the first weeks and months after treatment, with the exception of treatment failure and consequent progression, which can occur several years after the procedure. It is also important to highlight that progression after CXL is not readily identifiable, since, in the first years after the procedure, corneal remodeling takes place, and parameters used to assess progression may be temporarily modified. An additional limitation of the study is the variation in pachymetric measurements. In the first patients, most pachymetric measurements were performed with ultrasound, whereas in the later months, OCT was the most used technique as it was less invasive and more reproducible. Furthermore, even though each eye was measured with the same topographer, we compared eyes scanned with different instruments. This limitation could have an impact on keratometry data. To our knowledge, there are no studies evaluating the agreement between measurements of the two topographers used in this study. With regards to safety data, endothelial cell density measurements were not performed, even though it is important to monitor potential endothelial cell damage.
In conclusion, this large retrospective cohort study confirms the success and safety of CXL for the treatment of progressive keratoconus. Furthermore, this study adds important information regarding the long-term results contributing to the limited currently available literature.
