Myopic eyes were clearly different from emmetropic eyes before the onset of myopia in terms of refractive error, axial length, relative peripheral refractive error, and growth rates for these variables. Compared with emmetropes
children were less hyperopic on average when nonmyopic as long as 4 years before the onset of myopia. This finding is in agreement with previous longitudinal analyses in which early refractive error was used as a predictive factor.6,40
Axial length followed a similar course—namely, longer than in emmetropes
, but only as early as −3 years before onset. Peripheral refractive differences were delayed by 1 year, with became-myopic
eyes relatively more peripherally hyperopic than emmetropic eyes by −2 years before onset. These results suggest that the window for predicting onset is limited. Under ideal, yet artificial, conditions where the outcome is known, refractive error, axial length, and peripheral refraction as cross-sectional predictors would only be useful between −4 and −2 years before onset. Real-world performance will certainly be worse when age is the only information available instead of visit relative to onset. This small window presents a major challenge to accurate prediction and timely intervention.
In addition to having cross-sectional differences compared with emmetropes
−2 to −4 years before onset, eyes of became-myopic
children differed in their rate of change between years. One of the striking features of the data was that the fastest interval for change in all three components was during the year before onset of myopia (, , ). The change in refractive error and axial length then slowed the year after onset, although the eye continued to grow and to progress in myopia at an elevated rate compared with emmetropic eyes. The occurrence of the fastest rate in the interval between the year before and the year of onset suggests that there is something special about the onset of myopia beyond reaching a criterion number of diopters. The process of onset does not appear to be a consistent, gradual expansion of the globe and increase in refractive error. There appears to be something of an acceleration over the myopic threshold. This phenomenon has been observed previously.4
Whereas our criterion of −0.75 D in each meridian is more conservative than that used in this previous analysis of components before and after onset of myopia (any minus in each meridian and a spherical equivalent of at least −0.25 D), it is noteworthy that despite this difference in definition, both studies show the fastest rate of change in myopia in the interval between the year before and the year of onset of myopia.4
The choice of criterion level for myopia at onset is somewhat arbitrary. This point may be too late for some children and too soon for others considering 11.5% received a refractive correction 1 year before the 0 visit and 34.5% did not wear a correction even 2 years after. Measurement error is a potential explanation for the faster rate of change before the year 0 visit. The assumption that measurement error is randomly distributed at each visit may not be true between visits −1 and 0. More children may be measured inaccurately as less myopic at visit −1, thereby keeping them from being classified as myopic. These children would then be falsely measured as having a faster rate of progression between visits −1 and 0. This error seems unlikely, however, as this finding of faster change between visits −1 and 0 occurred in independently measured axial length as well as in relative peripheral refraction. Understanding the process underlying this acceleration requires further investigation.
Wearing an optical correction appeared to have no effect on the level of or change in relative peripheral refraction. We have no information on the power of the prescriptions worn by children or the reasons corrections were prescribed. One might reasonably assume that most of the corrections after onset were for myopia rather than for astigmatism, which would be expected to increase the amount of peripheral hyperopia and foveal accommodative lag in children with corrected vision compared with those without correction. Whether this increase was symmetric between periphery and fovea is not known. It may be asked whether correction accelerates myopia progression. Children with corrected vision have shown more lag and more myopia,41
but this result is confounded by the fact that correction is not being randomly applied. More myopic children are more likely to be the ones wearing a correction. Cause and effect of the progression of myopia cannot be distinguished by using these data. The lack of impact of correction on relative peripheral refraction shown in is not expected to be confounded by variation in the strength of refractive corrections needed by children with and those without correction because relative peripheral refraction was largely independent of progression in these early postonset years (with the exception of Hispanics in year +5).
Relative peripheral refractive error displayed a pattern around the time of onset that was similar to refractive error and axial length—namely, more rapid change across visits before onset, with the fastest change occurring at onset. But rather than stabilizing at a moderate rate of change after onset, relative peripheral refractive error showed no change across post-onset visits. This pattern suggests a two-phase process in ocular growth, one leading up to onset and a second after onset, assuming that variation in peripheral refraction is due in large part to variation in local ocular shape.12–17,42
An increasingly prolate shape suggests equatorial restriction, whereas a constant shape suggests an overall, more uniform global expansion.18
The suggestion of two phases for shape change raises the possibility that two mechanisms may be at work. Various mechanisms may be proposed for why ocular shape is prolate or less oblate in myopia. If growth occurs preferentially at the posterior pole, axial expansion will outpace equatorial expansion, to create a relatively more prolate shape. This mechanism would produce a monotonically increasing prolate shape, however. Stability of shape could be achieved by cessation of growth, yet the data clearly show continued axial growth during the time shape was stable after onset. External equatorial restriction from extraocular muscles16
or constraints from orbital size18
have also been mentioned as possible causes of a relatively prolate shape. These external sources of compression most likely would result in a continuously increasing, relatively prolate shape that was not seen after onset, unless some plasticity or adaptation reduced their restrictive force.
Several authors have discussed ocular shape in refractive error as either the source or the consequence of local defocus stimulating eye growth.9,19,20,43
Yet it seems counterintuitive that progress toward myopia should involve an increasingly relatively prolate shape. Local control of eye growth suggests that the eye should maintain a roughly uniform spherical equivalent, as do the most emmetropic eyes.20
Relative peripheral refraction did not even reach an absolute hyperopic state in became-myopic
children until visit −1 (), 2 years after axial length exceeded the average for emmetropes
. It could be argued that the increasingly relatively prolate shape before the onset of myopia adds to hyperopic defocus on a relative basis and drives the accelerating axial growth. However, stability in peripheral refraction happens when the eye is at its most hyperopic peripherally. A sudden loss of sensitivity to peripheral defocus would create stability, but seems unlikely. Another possible scenario is that peripheral defocus is less effective in driving eye growth compared with defocus at the posterior pole. This could accentuate posterior growth at the pole leading to an increasingly prolate or less oblate shape. As shape becomes less oblate, peripheral hyperopic defocus would increase until some point where low amounts of central defocus and high central sensitivity are equally effective in driving eye growth as the higher amount of peripheral defocus and low peripheral sensitivity. Spherical expansion may then occur when central and peripheral signals are equally effective.
Although this model of differential peripheral and foveal sensitivity seems plausible, it seems reasonable to assume that once a shape with sufficient peripheral hyperopia to drive eye growth both centrally and peripherally is attained at onset, axial length should continue to elongate at the highest rate. However, axial elongation is most rapid in the year before onset and slower after onset. In addition, the idea that a relatively prolate shape and peripheral hyperopia stimulates axial elongation in the periphery would seem only as valid as the evidence that hyperopic defocus drives central ocular growth. Recent evidence of only small effects from bifocal clinical trials,2,3,44
that accommodative lag is elevated only several years after ocular growth first becomes excessive,41
and that elevated accommodative lag is not always observed before onset of myopia41,45
argues against a major role for the influence of hyperopic defocus on axial growth and refractive error. Alternatively, the amount of lag may mediate the effectiveness of PALs for slowing the progression of myopia. Wearing PALs resulted in a larger treatment effect in children having more than the median amount of lag (by 0.33 D or 21%), although it should also be noted that the amount of lag did not significantly affect the rate of progression in children wearing single-vision lenses.46
Internal sources of restriction may also produce the patterns observed. Connections between the choroid, ciliary body, and crystalline lens are sufficiently strong to transmit deforming forces to the crystalline lens when the ciliary diameter expands. Lens deformation can occur whether the force applied is ordinary, such as during relaxation of accommodation, or when the force is extreme, such as during induced stretching experiments.47
The lens thins and flattens during stretching, because it is the most elastic tissue (i.e., has the lowest modulus of elasticity, of the structures in this chain).47,48
Two phases for equatorial growth, one where uveal tension increases and one when this tension reaches the limit of stretching, may produce two different patterns for change in ocular shape. These two patterns are suggested by van Alphen's experiment,31
which involved inflating globes with exposed choroids with a ring of sclera left in place near the ciliary body. The first phase of inflation is characterized by anteroposterior expansion, a change to a more prolate shape. The suggestion of a second phase occurs as the globe expands further toward a point where the ciliary body becomes what is described as maximumly flat. Expansion of the globe then becomes more spherical in the third panel of van Alphen's .31
These phases may be analogous to the increase in relative peripheral hyperopia before onset followed by the more symmetric expansion after onset.
Sampling of relative peripheral refractive error is limited in this study, in that it is only measured at one peripheral point. However, recent evidence shows that both horizontal and vertical dimensions are relatively more prolate, or less oblate, in myopic eyes.11
There is some uncertainty as to whether there is symmetry horizontally and vertically in ocular shape based on MRI data. Although the slope values that have been reported for asphericity as a function of refractive error and plane are numerically different (0.028 horizontal and 0.018 vertical), these slopes were not tested for the significance of the difference between them. Inspection of Figure 4b in Atchison et al.11
shows quite a bit of variation and overlap. Within the range of common refractive errors (−1.00 to −5.00 D), the two lines are nearly superimposed.11
The mean horizontal and vertical asphericities for myopes were not significantly different (overlapping 95% CIs of 0.15–0.21 and 0.17–0.23).11
There is evidence, therefore, that vertical–horizontal asymmetry in shape is either small or not significant. General symmetry offers some justification that measurement of one peripheral point may represent the behavior of other parts of the globe.
Asymmetries in peripheral refraction are more apparent. Horizontal–vertical asymmetries have been reported to be between 0.0 to 1.0 D, with vertical quadrants inconsistently either more myopic or more hyperopic than horizontal quadrants.9,19
Atchison et al.21
report horizontal–vertical asymmetries up to approximately 3.0 D with a more consistent relative vertical myopia. Lower visual field (superior retina) myopia has also been found in adult samples,19,21
although this was not found in a sample of children.9
The limitations of data showing asymmetries are that they are not from children, do not examine incident myopes, and are not longitudinal. Therefore they are not conclusive as to whether the relative changes in each quadrant are symmetric during the development of myopia. The implications of these asymmetries are unclear in either a defocus-based or a restriction-based model. No human investigation has attempted to correlate ocular growth with local defocus. Nasal–temporal asymmetries in ciliary muscle anatomy and lens position are well-known,49
suggesting that restriction-based expansion need not be strictly symmetric. Future studies would be strengthened by longitudinal assessment of multiple quadrants.
Ethnic variation in relative peripheral refraction indicates that neither relative peripheral hyperopia nor an increase in relative peripheral hyperopia are universal features of the myopic eye. African-American became-myopic children in year +3 are nearly as myopic on average as Asian-American became-myopic children in year +1 (), yet the African-American group has no significant average relative peripheral hyperopia and the Asian-American group has the largest amount (). This may indicate ethnic variation in the underlying process that leads to excessive axial elongation or merely variation in degree within the same process.
In summary, the current findings suggest that longer eyes, more negative refractive errors, and increased relative peripheral hyperopia occur 2 to 4 years before the onset of myopia and may therefore be potentially useful as predictors of myopia onset. It is problematic, however, that this window of opportunity is brief. Longitudinal data suggest that faster growth, faster progression, and a more rapid change toward peripheral hyperopia are also predictive of the onset of myopia, but again, only within a narrow window of time. Because time is arranged relative to onset rather than to age, the analysis may be better suited to pointing out potential predictive factors for future analyses and for evaluating what these patterns near the time of onset indicate about the process of becoming myopic rather than for making specific predictions. Even if not optimized for prediction, the current analysis suggests that the process of becoming myopic is not one of gradually accumulating an excessively long axial length. The acceleration in axial growth, myopia progression, and peripheral hyperopia before the onset of myopia followed by no change in relative peripheral refraction after onset suggest a two-stage process that is not consistent with simple external restriction. Differential sensitivity to defocus in the periphery compared with the fovea is plausible, but does not seem consistent with the timing of axial ocular growth. The current findings may be consistent with a process of resistance to stretching by ocular tissues during growth followed by failure to stretch when growth is excessive, but this hypothesis remains speculative at this stage.