The statistical analysis software SAS version 9.3 (SAS Inc., Cary, NC) was used for data analyses in this study. Only right eye data were used as the correlations in axial length and corneal curvature/radius between the two eyes were high (r>0.95).

Only subjects with at least four visits were included (n=462). Corneal curvature including the steepest and flattest meridians was plotted by visit, as well as by gender, ethnicity and baseline age group across all visits. Linear mixed models (LMM) were used to evaluate longitudinal changes of corneal curvature and compare across subcategories based on all available data adjusting for covariates (gender, ethnicity, lens type, baseline age and baseline refraction). The relationship between AL and corneal curvature was evaluated at baseline, 3, 6, 9, 12 and 14 year visits using Pearson correlation coefficients (PCC). At each selected visit, PCCs were compared between two subcategories (e.g. males vs. females) using Fisher's transformation. To compare PCCs across five ethnic groups, chi-square tests of the Fisher transformed correlations were applied. PCCs between myopia and AL were compared to PCCs between myopia and AL/CR ratio using Meng's macro28 (comparing correlated correlations) at each selected visit. p-Values for the comparisons of correlation coefficients at each visit were adjusted for multiple comparisons using the Bonferroni correction. To test for a linear trend in PCCs across the visits, the AL and corneal curvature measurements were standardized at each visit and then the differences in slopes across the visits were tested using linear mixed models, since the standardized regression coefficients were very close to the PCCs.

The baseline characteristics of the COMET cohort in Table 1 illustrate that the sample was ethnically diverse, with an average age of 9.3 years and low to moderate myopia. The mean axial length was 24.14mm, with an average corneal curvature of 44.11D and an AL/CR ratio of 3.15. At 14 years the mean axial length increased to 25.40mm, the average corneal curvature decreased to 43.98D, and the AL/CR ratio increased to 3.31. The mean amount of myopia increased from −2.38D at baseline to −5.17D at 14 years.

The average axial length to corneal radius ratio (AL/CR) was 3.15 at baseline and increased to 3.31 at the 14-year visit. Fig. 2 shows that the AL/CR ratio was higher for the steepest than the flattest meridian and that the largest changes occurred during the first 5 years of the study (p<0.0001).

Figure 2.

AL/CR ratio by visit.

(0.09MB).

Table 3 compares correlations between the magnitude of spherical equivalent myopia and AL vs. the magnitude of myopia and AL/CR ratio. All correlations were lower at baseline (−0.32 AL alone, −0.55 AL/CR ratio) and in the early years than in the later years and at the 14-year visit (−0.65 AL alone, −0.79 AL/CR ratio).

Fig. 3 compares the baseline AL/CR ratio to myopia progression after 5 years. We chose 5 years for this comparison because the randomized clinical trial ended at this point and most of the progression occurred by the end of year 5. As shown in the figure, there was no significant relationship between the AL/CR ratio and myopia progression (r=0.07, p=0.17).

Figure 3.

Baseline AL/CR ratio vs. myopia progression at 5-year visit.

(0.14MB).

We completed an analysis of all subjects who switched to contact lenses at the end of the clinical trial to determine if there were any effects of contact lens wear on the AL/CR ratio. This analysis showed that there was no difference among those who remained in the original treatment assignment compared to those that switched to contact lenses (data not shown).

The results of this study represent the first long-term (14 years), longitudinal data evaluating changes in corneal curvature, its relationship to axial length, and the AL/CR ratio over time in a cohort consisting exclusively of children/adolescents with myopia. We found that the average corneal curvature in the COMET cohort was relatively stable over 14 years of observation, although there were significant differences between the flattest and steepest meridians as well as significant gender and ethnicity differences in average corneal curvature over this time period. In addition, our findings show a high correlation between corneal curvature and axial length during the early years of myopia progression, but a gradual decrease in this correlation with slowed progression and eventual stabilization of myopia. We also evaluated the relationship between corneal curvature and axial length using the AL/CR ratio and found a high ratio at baseline, which increased significantly over the first 5 years of the study when myopia progressed most and then showed minimal change through year 14. In addition, the magnitude of myopia was better correlated with the AL/CR ratio than with AL alone.

The small and differential changes in the flattest and steepest meridians found in this study are consistent with data from the 3-year longitudinal SCORM study in Singapore. In that study the authors suggested that corneal power appears to have a minor role in the progression of myopia,29 and that suggestion is supported by our data. The flattest corneal meridian (horizontal) became significantly flatter (p<0.0001) over the first 5 years of the study, while the steepest meridian did not change significantly. This differential change would suggest an increase in with-the-rule astigmatism as myopia progresses and agrees with previous reports showing this trend.30–32

We also found gender and ethnicity differences in corneal curvature. Beginning at baseline, females had significantly steeper corneal curvature than males, along with correspondingly shorter axial lengths with the same amount of myopia at all visits, an observation that did not change over time (or as the cohort became young adults). Steeper corneal curvature in females has also been reported by previous authors in young adult myopes,3,33 and in studies of children in the CLEERE study34 and the Sydney Myopia Study.13,35 Even though in CLEERE and the Sydney Myopia Study only 10–12% of the subjects were myopic and the authors did not provide data separately by refraction, the results from these two studies and COMET are consistent with respect to gender differences in corneal curvature in children.

In regard to ethnicity, our longitudinal dataset showed that (1) the pattern of slight cornea flattening over the 14 years of follow up was similar across all ethnicity groups, and (2) at all visits the average corneal curvature in Hispanic children was the flattest of all ethnic groups, and that compared to the White group, who had the steepest corneas, there was a clinically meaningful12,36 difference of ≥0.50D. We are not aware of any other longitudinal studies with an ethnically diverse, myopic only population, for comparison. However, the CLEERE study (using cross-sectional data) included corneal curvature data by ethnic group and the ethnic groups studied were similar to the groups studied in COMET. CLEERE investigators reported that of all ethnicities Whites had the steepest corneas in the horizontal meridian, an observation similar to our finding, and Native Americans had the flattest in both the horizontal and vertical meridians.12

We found that the mean corneal curvature was steeper than what was reported in two other clinical trials in myopic Asian children,37,38 which may be because half of COMET's participants are White, and Whites have the steepest corneas; also Asian children represent only a small percentage (9%) of the COMET cohort. The mean baseline corneal curvature of Asian children in COMET was 44.00D, still slightly steeper than in the two studies. In a 2-year trial in Hong Kong of progressive addition lenses to slow the progression of myopia in children 9–12 years old, Leung and Brown37 reported a mean corneal curvature of approximately 43.30D in three lens groups. In another trial of progressive lenses, Hasebe et al.38 found a mean corneal curvature of 43.60D at baseline in 95 myopic, Japanese children ages 6–12 years old.

We evaluated correlations between corneal curvature and axial length by visit and found a high correlation during the early years of the study, when the range of axial lengths was relatively small, and a gradual decrease in this correlation as myopia continued to progress and the eyes grew longer. These data lend support to the findings of Hoffmann and Hütz6 who reported that the correlation between axial length and corneal curvature decreased significantly with longer eyes. Guggenheim et al.24 also reported that corneal curvature and axial length were correlated (r=0.54) in a group of European teenagers with a range of refractions, a correlation similar to that found in the current study of only myopes (r=0.57). They concluded that the coordinated scaling of corneal curvature and axial length is determined by many common genetic variants, and that environmental factors may play a role in changing the coordinated growth of these components in ametropic eyes.

Our study is the first to provide longitudinal correlations over 14 years from the same subjects for AL/CR ratio vs. magnitude of myopia and for AL vs. magnitude of myopia. The same pattern was found at all visits, even as the correlations grew larger as the eyes grew longer and myopia progressed. Others have reported that the AL/CR ratio is more highly correlated than AL alone with spherical refractive error.10,15,16,39 In an early report of the AL/CR ratio, Stenstrom39 studied 1000 young adult patients and found a correlation of −0.76 for refractive error and AL compared to −0.84 for refractive error and AL/CR ratio. Grosvenor and Scott16 also reported a high correlation between AL/CR ratio and refractive error (−0.92) compared to −0.76 between refractive error and AL alone, and found that 84% of the variance in refractive error could be accounted for by variance in the AL/CR ratio. Ip et al.10 stated that although greater AL was associated with a greater likelihood of myopia in their sample of schoolchildren, myopic refractions were also found in shorter eyes. They proposed that the AL/CR ratio provides a rough measure of the degree of matching between AL and corneal power and, therefore, represents a more useful marker than AL alone of progress toward myopia. He et al.15 suggested that the AL/CR ratio may be a better method for screening for children with myopia than the traditional uncorrected visual acuity method used in most screenings.

While previous studies have evaluated the AL/CR ratio as a predictor for myopia development, the longitudinal nature of this study allowed us to evaluate the possibility that the AL/CR ratio may also be predictive of myopia progression. Data from the 5-year follow-up of the COMET cohort did not support such a relationship (Fig. 3).

An important strength of this study is the availability of 14-year longitudinal data from an ethnically diverse cohort with excellent long-term retention (93% at the 5-year and 78% at the 14-year visit). Unlike many earlier reports on corneal curvature and axial length, the COMET used a standard protocol with data collected by trained and certified optometrists at four clinical centers and analyzed at a central coordinating center. A limitation of this study is that all subjects were myopic at enrollment so changes in AL, CC or AL/CR that might be associated with the onset of myopia cannot be addressed.