Peripheral refraction and the development of refractive error: a review

Authors


Correspondence and reprint requests to: Dr Hema Radhakrishnan.
Tel.: +44 (0) 161 306 8763; Fax: +44 (0) 161 306 3887.
E-mail address: hema.radhakrishnan@manchester.ac.uk

Abstract

It has been suggested that emmetropic and low-hyperopic eyes in which the refractive error in the periphery of the visual field is relatively hyperopic with respect to the axial refraction may be at greater risk of developing myopia than eyes with similar refractions but relatively myopic peripheral refractive errors. The animal and human evidence to support this hypothesis is reviewed. The most persuasive studies are those in which emmetropization has been shown to occur in infant rhesus monkeys with ablated foveas but intact peripheral fields, and the demonstration that, in similar animals, lens-induced relative peripheral hyperopia produces central axial myopia. Evidence for emmetropization in animals with severed optic nerves suggests that emmetropization is primarily controlled at the retinal level but that the higher levels of the visual system play a significant role in refining the process: there appear to be no directly equivalent human studies. Since any contribution of the higher centres to the control of refractive development must depend upon the sensitivity to defocus, the results of human studies of the changes in depth-of-focus across the field and of the contribution of the retinal periphery to the accommodation response are discussed. Although peripheral resolution is relatively insensitive to focus, this is not the case for detection. Moreover accommodation occurs to peripheral stimuli out to a field angle of at least 10 deg, and the presence of a peripheral stimulus can influence the accommodation to a central target. Although the basic hypothesis that a relatively hyperopic peripheral refractive error can drive the development of human myopia remains unproven, the available data support the possibility of an interaction between the states of focus on axis and in the periphery.

Introduction

There is general acceptance that, in addition to genetic factors, visual experience after birth affects refractive development and helps most eyes to approximate to emmetropia (see for example Wildsoet (1997), Smith (1998), Norton (1999) and Wallman and Winawer (2004) for reviews). This concept of emmetropization implies that there is an active feedback mechanism involving the state of focus of the retinal image which drives ocular growth to produce a final refractive state close to emmetropia or, if the habitual viewing distances are relatively short, towards an appropriate level of myopia (e.g. Schaeffel and Howland, 1988). Given the importance of foveal vision, most attention has been concentrated on mechanisms relating to the state of focus on the visual axis. However, the foveal area forms only a small fraction of the overall visual field and it is reasonable to suggest that more peripheral retinal areas might also be of at least some importance. The fact that optimal retinal image quality (Jennings and Charman, 1981a; Navarro et al., 1993) and neural spatial resolution (Weymouth, 1958) both decline with field angle need not be an objection to this hypothesis, since emmetropization and ability to detect the sign of defocus has been shown to be possible even when only very low spatial frequency information is available, at least in avian eyes (Schaeffel and Diether, 1999).

Although the concept that the state of image focus in the peripheral retina might influence refractive development goes back nearly 40 years (Hoogerheide et al., 1971), interest in this idea has recently grown markedly (Seidemann et al., 2002; Wallman and Winawer, 2004; Stone and Flitcroft, 2004; Charman, 2005; Mutti et al., 2007), stimulated by the increasing incidence of myopia in many parts of the world (Saw, 2003; Norton et al., 2005; Logan and Gilmartin, 2005). A particularly attractive possibility is that, if it were true that measurement of the pattern of peripheral refraction allowed individuals who were at risk of developing myopia to be identified at an early stage, steps could in principle be taken to check the myopia development, e.g. by using spectacle lenses which deliberately introduced peripheral myopia while not affecting central vision (e.g. Smith et al., 2005a; Tabernero et al., 2009). Thus it is important that the hypothesis that peripheral refraction can influence ocular growth and refractive error in humans be investigated as fully as possible. Evidently, such a mechanism implies that defocus, and its sign, can in some way be detected in the peripheral retina and can generate a signal to control ocular growth rates, although such detection need not occur at the conscious level. As will be discussed in more detail below, animal studies suggest that emmetropization can still occur even when the optic nerve is severed, so that local retinal mechanisms may be sufficient for gross regulation of refractive errors: however, the effectiveness of the emmetropization process is impaired as compared to that found when the optic nerve is intact, implying that higher centres contribute to the refinement of the refractive state (e.g. Troilo and Wallman, 1991; Wildsoet and Wallman, 1995; Wildsoet and Schmid, 2000).

There have been a number of both longitudinal and transverse studies of the links between axial and peripheral refraction. Further valuable insights into the ability of the peripheral visual system to detect focus change can be gained from studies of both depth-of-focus and accommodation response for stimuli falling at different peripheral angles. Therefore we now review the evidence produced by these studies in relation to the suggestion that peripheral refraction may play a role in the development of myopia in at least some individuals.

Evidence from animal studies on the role of peripheral refraction in the regulation of refractive error

As is well known, animal studies have provided substantial evidence that post-natal visual experience and non-foveal areas of the retina can affect refractive development (see, e.g. Wildsoet, 1997; Smith 1998; Norton, 1999, Wallman and Winawer, 2004, for reviews). They show, for example, that if negative lenses are used to place the ocular image behind the retina, axial myopia results. This is because an abnormal increase in the growth of axial length occurs, which acts to keep the images formed by the lens-eye combination sharply focused on the retina, thus making the eye itself myopic. Analogously, positive lenses, which place the image in front of the retina, reduce the ocular growth rate, leading to hyperopia. When the retinal image is degraded, e.g. as a result of wearing a diffusing goggle, myopia also results (form deprivation myopia), again through an increase in axial length (primarily vitreous chamber depth). If form deprivation or lens treatment is discontinued during visual development of the young eye, recovery from the induced myopia can occur. However, if a lens that optically corrects the myopia is placed over the eye, it remains myopic (Diether and Schaeffel, 1997; McBrien et al., 1996; Wildsoet and Schmid, 2000).

Application of cylindrical lenses to the eyes of developing chicks has produced mixed results. Irving et al. (1995) found evidence for partial compensatory astigmatic growth, the effectiveness of the compensation varying with the orientation of the axis of the cylinder. In contrast, Schmid and Wildsoet (1997) found no astigmatic compensation in the growing eye: rather chicks appeared to “emmetropize” to the meridian with the greater myopic defocus, irrespective of the cylinder axis. In monkeys, application of +1.50DS/−3.00DC crossed-cylinder lenses resulted in “emmetropization” to one of the two focal planes associated with the two principal meridians of the astigmatism, rather than the circle of least confusion (Kee et al., 2004): most animals became more hyperopic than controls. In all studies, variation between animals was greater than that found where spherical lenses were used.

The primary mechanisms of developing form-deprivation or lens-induced myopia appear to be regulated locally within the eye. When the optic nerve of an animal eye is severed, development of form-deprivation myopia continues (Norton et al., 1994; Troilo et al., 1987; Wildsoet and Pettigrew, 1988), and lens compensation may still occur, although not to the same extent as would be expected when optic nerve is intact (Troilo and Wallman, 1991; Wildsoet, 2003; Wildsoet and Wallman, 1995; Wildsoet and Schmid, 2000). In particular, Wildsoet and Wallman (1995) state “Optic nerve section in chicks prevents the development of myopia in response to negative lenses, but not to diffusers, suggesting that compensation for hyperopia requires the central nervous system”.

Thus both retinal and central elements are involved in the normal emmetropization process and the retina must, as a result of local defocus, provide some biochemical signal which controls eye growth (e.g. Crewther, 2000; Wallman and Winawer, 2004). For example, it has been shown that, in chicks, relatively low levels of lens-induced defocus can affect the expression of the transcription factor ZENK in glucagon-immuno-reactive amacrine cells of the chicken retina (Bitzer and Schaeffel, 2002). After 40 min of lens wear, the number of ZENK-expressing cells increases with positive-powered lens wear and reduces with negative lenses, with the changes in ZENK expression being related to lens powers by a sigmoidal function that saturates at approximately ±7 D of defocus. The important feature is that this mechanism gives information of the sign of the defocus, implying that the relative hyperopia in the peripheral retina might therefore trigger a signal for eye growth. Zhong et al. (2004) have also shown changes in the activity of bipolar and amacrine cells as a function of focus and Műnch et al. (2009) have described neural circuits in the mouse retina which are sensitive to looming and hence might assist in any focus mechanism.

Glucagon has also been shown to play a significant role in regulating ocular growth (Vessey et al., 2005). Vessey et al. found that in the chick eye glucagon significantly inhibits the development of form deprivation myopia and causes the choroid of treated eyes to thicken. This suggests that glucagon may contribute to visual regulation of the refractive state, reducing axial length by restraining scleral growth and stimulating choroidal thickening. A variety of drugs acting at the retinal level can reduce the axial elongation in experimental chick myopia (e.g. Stone et al., 1991; McBrien et al., 1993; Leech et al., 1995).

Evidently, some of these biochemical processes may be involved in detecting the direction of defocus and therefore could play a crucial part in regulating eye growth. However, the primary mechanism used by the eye to detect the direction of defocus remains unclear (see Wallman and Winawer, 2004 for a review) with several possible mechanisms including trial and error, and chromatic and monochromatic aberrations acting together.

Several studies show that eye growth can be regulated by different local regions of the retina, rather than just by conditions at the fovea. Chicks reared with translucent occluders over the nasal or temporal half of the eye show form-deprivation myopia which is limited to the deprived part of the retina (e.g. Wallman et al. 1987; Diether and Schaeffel, 1997) and, for many species, raising animals in an environment in which objects are at systematically different distances in different parts of the visual field leads to appropriate differences in refraction across the field as a result of asymmetric eye growth (e.g. Fitzke et al., 1985; Hodos & Erichsen, 1990; Miles and Wallman, 1990; Schaeffel et al., 1994; Smith, 1998).

More recently, studies on infant monkeys suggest that the peripheral retina can play an important role in modulating overall eye growth and axial refraction (Smith et al., 2005b, 2007a,b; Hung et al., 2008; Smith et al., 2009; Huang et al., 2009). Monkeys raised with diffusers with central apertures allowing up to 37 degrees of unrestricted central vision developed ametropic axial refractive errors, even though central vision was unrestricted (Smith et al., 2005b). Ablation of the central 10 deg diameter of retina around the fovea while leaving the periphery intact resulted in emmetropia (Smith et al., 2007b) and imposition of a hyperopic defocus in the periphery but not on the central 10 deg of field resulted in myopia (Smith et al., 2005b, 2009). However, in contrast, Schippert and Schaeffel (2006) found that chicks reared with lenses having central apertures remained emmetropic, showing no evidence of failure of emmetropization when peripheral refraction was artificially changed. It may be that these apparently conflicting results reflect inter-species differences: in particular, visual resolution declines much more rapidly with field angle in primates than in birds. Note that the studies of Smith et al. (2005b, 2009) imposed a simple spherical hyperopic defocus on the peripheral retina. In human eyes, although the mean spherical error may be hyperopic, there is also substantial peripheral astigmatism (approximating to oblique astigmatism) with a magnitude which varies approximately with the square of the field angle (Atchison and Smith, 2000).

Overall, although not fully consistent, the findings of these animal studies indicate that peripheral visual image quality may have some effect on emmetropisation.

Peripheral refraction as an indicator of the likelihood of myopia development in humans

As noted earlier, the key study which suggested a link between the pattern of peripheral refraction and the development of at least some cases of human myopia was carried out by Hoogerheide et al. (1971). Their longitudinal study involved over 400 young adults who were being trained as pilots (age at entry to training 18–20 years). Refraction was measured over the central ±60 deg of horizontal visual field, using retinoscopy, and the results were classified into five characteristic patterns of refraction, or skiagrams, on the basis of the patterns of change in refraction across the visual field (see Figure 1, after Rempt et al., 1971). The authors found that the majority of the initially weakly-hyperopic or emmetropic pilots who developed myopia over the next few years had skiagrams in which both sagittal and tangential oblique astigmatic image surfaces in either one or both of the horizontal semi-field meridians were relatively hyperopic in comparison to the axial refraction (i.e. Types I and III): this implies that the image surface corresponding to the mean spherical equivalent or best-sphere correction was also relatively hyperopic. For such individuals, the probability of myopia developing was about 40%. In contrast, if either one or both oblique astigmatic image surfaces showed relative myopia or emmetropia with respect to the axial refraction (skiagrams II, IV and V) the probability of subsequent development of axial myopia was low (ca. 4%), see Table 1. Unfortunately the time interval over which these refractive changes occurred was not specified but presumably it was of the order of the few years required to complete the pilot training. In addition, since the onset of myopia was after 18 years of age, Hoogerheide et al. (1971) only studied late-onset myopia. Early-onset myopes tend to have structural changes in the eye, with the ocular growth curve being significantly different from that of emmetropes (Jones et al. 2005), which might further lead to differences in the peripheral refractive error.

Figure 1.

 The common patterns of peripheral refraction across the horizontal meridian (after Rempt et al., 1971). On each skiagram, the thick lines show schematically the radial and tangential image fields across the horizontal meridian, H and M indicating whether the peripheral fields are hyperopic or myopic with respect to the axial refraction (after Rempt et al., 1971). Any axial astigmatism has been corrected to make the image fields coincide on axis (the image fields have been displaced vertically in type I and III skiagrams for ease of viewing). Axial spherical refractive errors have also been corrected, so that each thin horizontal line represents a state of relative emmetropia across the visual field.

Table 1.   Refractive shifts in a group of hyperopic or emmetropic young pilots as a function of their initial pattern of peripheral refraction (skiagram). The number of pilots in each refractive category are shown (after Hoogerheide et al., 1971)
Shift of refractionType I (%)Type II (%)Type III (%)Type IV (%)Type V (%)Total
Hyp. → myop.4 (45)1 (11)2 (22)2 (2)9
Hyp.→ emm.11 (40)5 (11)6 (22)6 (22)28
Emm.→myop.13 (77)2 (12)1 (6)1 (6)17
Emm./Hyp. (no change)8 (5)35 (22)5 (3)103 (66)9 (5)160
Percentage with this skiagram type who became myopic4772130 

Subsequently, however, Mutti et al. (2007), in a longitudinal study of almost 1000 children aged between 6 and 14 years, measured several ocular parameters, including relative peripheral refractive error (RPRE) defined as the difference between the refraction at 30 deg of temporal visual field and the axial refraction. Myopia was defined as a refractive error of at least −0.75D in all meridians. They found that, in general agreement with the ideas of Hoogerheide et al. (1971), “Children who became myopic had more hyperopic relative peripheral refractive errors than did emmetropes from 2 years before onset through 5 years after onset of myopia (p < 0.002 for each year)”. In addition to RPRE, other factors predictive of the onset of myopia included a more negative refractive error, longer axial lengths and faster rates of change in these variables. The changes in refraction and RPRE as a function of time with respect to the onset of myopia are shown in Figure 2, data for those children who remained emmetropic being included for comparison. The marked increase in relative peripheral hyperopia before myopia onset is obvious.

Figure 2.

 (a) Mean spherical equivalent refractive error as a function of the annual visit relative to the onset of myopia for those children who became myopic and those who remained emmetropic (b) Relative peripheral refractive error (after Mutti et al., 2007).

An initially axially-emmetropic individual with a hyperopic RPRE would have the peripheral image lying behind the retina. If the results of Hoogerheide et al. (1971) and Mutti et al. (2007) are interpreted on the basis of some of the animal experiments (Smith et al., 2005b, 2007a, 2007b, 2009; Hung et al., 2008) and the assumption is made that the state of focus in the periphery at least partly controls refractive development, this hyperopic peripheral state of focus would influence eye growth and the axial length of the eye would tend to increase until the peripheral image was in focus (Figure 3). Constraints on the way the shape of the eye could change, set by such factors as intraocular pressure, the mechanical characteristics of the sclera etc, might then mean that, in correcting the focus for the peripheral retina, growth led to a myopic fovea (Figure 3). Clearly, any model used for this process must make assumptions about the relative importance of axial and peripheral foci, as well as about the mechanical and other constraints on eye shape. In general, it is unlikely to be possible for ocular growth to bring the image simultaneously into focus across the full field. Thus the final increase in axial length is likely to depend on some weighted average of the defocus signals from different parts of the retina. If axial focus is dominant, then the state of peripheral focus has little influence on growth. If, on the other hand, final overall focus is at least partly weighted by the peripheral focus, then the observations of Hoogerheide et al. (1971) and Mutti et al. (2007) can be explained. Note that a feature of this model is that, if the axial myopia is corrected by a spectacle or contact lens, the restored peripheral hyperopia will result in further undesirable increases in axial length and myopia.

Figure 3.

 Schematic diagram of the effects of a hyperopic RPRE on ocular growth in an initially axially-emmetropic eye. The red dashed curve shows the image surface for the foveal best-sphere correction. The eye grows and the axial length increases (dashed black curve) to bring the peripheral image at some non-zero field angle into focus but, in doing so, causes axial myopia.

Although this simple model is qualitatively reasonable, careful examination of the Mutti et al. (2007) data suggests that progression towards myopia (defined by Mutti et al. as a refraction of at least −0.75 D in each meridian) may in fact precede the development of relative peripheral hyperopia. This is shown more explicitly in Figure 4, which combines corresponding axial and peripheral refractive data from Figure 2. Each symbol represents data from an annual visit with time proceeding from left to right. There is at least a strong suggestion that marked relative peripheral hyperopia only develops several years after the commencement of myopic progression. Some caution must be exercised in accepting this suggestion, due to the problems involved in averaging large amounts of individual data, but the possibility that the relative peripheral hyperopia is a consequence rather than a cause of the myopia development cannot be lightly dismissed. Note too that, whereas Hoogerheide et al. (1971, p.214) suggested that in adults “the general appearance of the skiagram is inborn and does not change very much during lifetime, especially with regard to its type” (see also Charman & Jennings, 2006), Mutti et al. (2007) found that in their sample of children the relative peripheral error showed marked changes with time (Figure 2b).

Figure 4.

 Relative peripheral refractive error (RPRE) as a function of axial spherical refractive error in “became myopic” children. Progress towards myopia has already started before the RPRE becomes markedly hyperopic (data from Mutti et al., 2007).

Peripheral refraction in existing ametropes

Although no other longitudinal studies of a similar nature to those of Hoogerheide et al. (1971) and Mutti et al. (2007) appear to have been published, there is little doubt that the pattern of peripheral refraction differs between different adult refractive groups. In a study of groups of existing adult ametropes, Millodot (1981) showed that, at least for field angles up to about 30 deg, both oblique astigmatic image surfaces in hyperopic eyes along the horizontal meridian tended to show relative peripheral myopia with respect to the axial refraction, whereas in myopic eyes there was relative hyperopia: in emmetropes the two astigmatic image surfaces tended to lie on opposite sides of the retina. Figure 5 shows Millodot’s average data for the mean equivalent sphere correction (best sphere) for the three refractive groups (myopes −1.00 to −7.87 D mean spherical equivalent; emmetropes −0.99 to +0.74 D; hyperopes +0.75 to +4.50 D): note the presence of nasal/temporal asymmetries and the tendency for the different peripheral refractions to converge as the field angle gets larger. Any relative peripheral ametropia is small for field angles <10 deg. Somewhat similar results have been obtained by several other authors (Seidemann et al., 2002; Mutti et al., 2000; Love et al., 2000; Atchison et al., 2006), some of whose work also shows differences between the patterns of refraction in different meridians. Tabernero and Schaeffel (2009a,b) have recently used a scanning photorefractor to measure continuous peripheral refraction profiles for the vertical meridian of the pupil and have found that in myopes the peripheral retinal shape is more irregular, even in eyes which are only moderately myopic. Only Calver et al. (2007) found no strong differences in the patterns of peripheral refraction between refractive groups. Their study, however, used custom-made trial lenses to correct myopia. It is of interest that, when myopia is corrected by spectacles, conventional lens designs induce significant relative hyperopia in the periphery (Tabernero et al., 2009).

Figure 5.

 Average values of mean spherical error as a function of retinal eccentricity for different refractive groups. The ranges of axial refractions within each group were: myopes −1.00 to −7.87 D; emmetropes −0.99 to +0.74 D; hyperopes +0.75 to +4.50 D (after Millodot, 1981). N and T represent nasal and temporal retina respectively.

Although Figure 5 shows the general form of the differences in peripheral refraction with ametropia, it depicts these differences in terms of mean equivalent spherical (best sphere) corrections. Since animal experiments suggest that imposition of a cylindrical error can affect refractive development, it is important to also take note of the changes in astigmatism across the field. Although the effects in individual eyes are influenced by the degree of axial astigmatism, the peripheral astigmatism approximates to oblique astigmatism, with the two focal lines always being approximately radial or tangential to the axis. Atchison and Smith (2000) suggest, on the basis of a number of experimental studies, that the magnitude of the astigmatism A(θ) can be approximated for field angles up to 60 deg by:

image

where A(θ) is in dioptres and θ is in degrees. Note that the second-order term dominates over most of the retina, so that astigmatism increases steadily to a value of about 7 D when θ = 60 deg. The important point is that, as one moves away from the axis, astigmatism increases roughly parabolically and there is no unique image surface to which any local “emmetropization” can occur: there seem to be at most only minor changes in the magnitude of the astigmatism with ametropia (Millodot, 1981). If animal studies are relevant in this situation, the limited experiments (Schmid and Wildsoet, 1997; Kee et al., 2004) suggest that it is more probable that growth might favour the more myopic, tangential image surface rather than that for the equivalent sphere, and that the presence of astigmatism might increase variability in the final refractive state. Should growth occur to bring the most rotationally-symmetrical point-spread function onto the retina, it is interesting to note that, due to the ellipticity of the off-axis pupil, this is likely to be achieved at a slightly different image plane from that corresponding to the mean equivalent sphere (Charman and Atchison, 2008).

In practice, there are departures from the symmetry of the refraction about the axis (Figure 6) which vary in different meridians and, perhaps, with refraction and age (Atchison et al., 2006; Atchison and Markwell, 2008). It appears possible to explain many of these if the optical system of the eye is rotationally symmetric about an axis whose orientation differs by a few degrees from that of the visual axis, the axis of symmetry typically intersecting the retina at a position slightly nasal and superior to the fovea (e.g. Seidemann et al., 2002; Charman and Atchison, 2009).

Figure 6.

 Average values for the magnitude of astigmatism as a function of retinal eccentricity along the horizontal meridian. The data shown are from Lotmar and Lotmar (1974), mean of 363 male adult eyes unselected for ametropia; Millodot (1981) mean of 62 adult eyes, unselected for ametropia; Gustafsson et al. (2001) mean of 20 adult emmetropes. The continuous curve corresponds to the fit suggested by Atchison and Smith (2000). Astigmatism values are slightly higher on the temporal retina (nasal field).

The observed differences in peripheral refraction of existing ametropes might either be predictive of future refractive change, as suggested by Hoogerheide et al. (1971), or they might simply be properties of eyes whose refractive error is already fully developed. Knowing that most ametropia is associated with differences in axial length, with relatively short axial lengths in hyperopia and longer lengths in myopia (Stenström, 1948), Charman and Jennings (1982) pointed out that Millodot’s mean results could be largely explained in terms of a very simple model (Figure 7) in which the anterior part of the eye up to around the equator and the associated astigmatic image fields together with the intermediate mean-spherical image shell were essentially independent of the ametropia, but the posterior portion of the globe corresponded to part of an oblate ellipsoid, a sphere and a prolate ellipsoid in the hyperopic, emmetropic and myopic groups respectively (see also Dunne and Barnes, 1990; Dunne et al., 1997). Such a situation would account naturally for the fact that the peripheral refractions for different axial ametropias tend to converge as the field angle increases. This suggestion has received some support from later work using magnetic resonance imaging (MRI) and other techniques (Cheng et al., 1992; Schmid, 2003; Atchison et al., 2004; Logan et al., 2004) although Atchison et al. (2005) suggest on the basis of their MRI work that there may be differences between the horizontal and vertical meridians and that it appears that both emmetropic and myopic retinas are in fact oblate in shape, although to a lesser degree in myopes.

Figure 7.

 Three theoretical eyes having identical anterior refractive components but differing axial lengths. The oblique astigmatic image shells (with the tangential image surface lying generally anterior to the sagittal surface) are the same for all the eyes but their differences in axial length will lead to axial hyperopia, emmetropia and myopia. At large field angles, corresponding to the equatorial region of the equator of the globe where the three retinae converge, the peripheral refractions will tend to become the same (after Charman and Jennings, 1982).

The model of Figure 7 suggests that the differences in peripheral refraction might merely be associated with the ametropia, rather than being causative. Indeed, if a peripheral refraction which was relatively hyperopic were always to cause the development of myopia, then (referring to Figure 3) myopia would continue to increase indefinitely in the axially-corrected myopic eye, or at least into young adulthood. Presumably, then, on the basis of the initial hypothesis, it is primarily in the relatively small fraction of hyperopic or emmetropic eyes which show relative peripheral hyperopia (some 10% of all hyperopic and emmetropic eyes in Hoogerheide et al.’s sample) that a myopic shift occurs. This implies that it is the subset of hyperopic and emmetropic eyes which have a more prolate (or less oblate) shape that might be at risk of becoming myopic and that axial growth slows once they have become myopic. Supposing that we are primarily concerned with the potential role of RPRE in this group, an associated defocus signal must somehow be derived to control eye growth. How sensitive is the peripheral visual system to changes in focus? Unfortunately no studies of human refractive development which are analogous to those carried out on animals with severed optic nerves appear to exist (although it would be of interest to examine patients with congenital optic nerve defects but otherwise normal eyes). Thus at present it is not possible to directly assess the sensitivity of the isolated retina to defocus. If, however, the higher-order visual system plays at least some role in controlling the final refractive state, as shown by the animal studies, it is still of interest to consider the overall sensitivity of the visual system to defocus.

Depth-of-focus and related measurements

The classical way of defining the sensitivity to focus error is the depth-of-focus (DOF), the dioptric change in spherical focus necessary to cause just-detectable blur in an originally in-focus image (or, alternatively, the total DOF is the dioptric interval between the positive and negative focal positions for just-detectable blur). In principle, in the periphery, the existence of substantial astigmatism increasing approximately quadratically with field angle (Atchison & Smith, 2000), should make it possible to detect the sign of the defocus: this, and the influence of other aberrations, does not appear to have been explored experimentally.

Few studies have been made of ocular DOF in the retinal periphery. Ronchi and Molenesi (1975) measured “DOFs” amounting to several dioptres over eccentricities between 7 and 60 deg but their flashed targets appear to be more related to a detection task rather than a judgement of blur. Wang and Ciuffreda (2004) determined the DOF using the edge of a circular aperture in a dark surround, the aperture radius varying between 0.5 and 8 deg. Pupil diameter was 5 mm. Total DOF (i.e. roughly twice the detectable error of focus) varied between about 0.9 D at the fovea and 3.5 D at an eccentricity of 8 deg. In related studies of the central retina, Wang and Ciuffreda (2005) and Wang et al. (2006) further explored blur detection and blur discrimination, i.e. the ability to detect changes in blur, in the central retina, the target again being a circular black/white boundary. Over the range of eccentricities studied (0 to 8 deg), both types of threshold increased by a factor of about 2, with the threshold for blur detection (i.e. half the total DOF) at 8 deg being about 1.25 D. Thus conscious detection of blur of target boundaries appears likely to be poor (threshold >1.25 D) for eccentricities greater than about 10 deg. Figure 8 summarises the DOF results of Ciuffreda and his colleagues.

Figure 8.

 Depth-of-focus (the dioptric interval from the best focus to the focus at which blur is just detectable) as a function of eccentricity. Data from Wang and Ciuffreda (2004) (diamonds); Wang et al., 2006) (squares). Also shown is the threshold for blur discrimination (triangles) (after Wang et al., 2006).

Wang and Ciuffreda’s (2005) finding that the threshold for blur discrimination (i.e. focus change) in a defocused image is lower than that for blur detection (i.e. the focus error which produces just noticeable blur in an initially defocused image) is of particular interest, however, in that it suggests that changes in focus which are smaller than the classic DOF are detectable if the image is already slightly out-of-focus. Others have previously shown similar effects for foveal images (Campbell and Westheimer, 1958; Walsh and Charman, 1988; Jacobs et al., 1989). At the fovea, Walsh and Charman (1988) suggested that sensitivity to focus change might be dominated by quite low spatial frequencies (ca. 5 c deg−1) and it is likely that this is also the case in the near periphery, perhaps with even lower spatial frequencies being of primary importance; Gu and Legge (1987) suggested that these might range down to 0.5 c deg−1 at an eccentricity of 30 deg.

Studies in which the state of focus is varied in the peripheral retina show that, for normal subjects, changes in spherical focus have little effect on resolution tasks for peripheral angles in the range 10–60 deg (Low, 1943; Rempt et al., 1976; Millodot et al., 1975; Anderson, 1996; Wang et al., 1997; Lundström et al., 2007a,b), although they may in patients with central field defects (Lundström et al., 2007b).

While these resolution results suggest relative insensitivity to optical blur in the periphery, and hence that peripheral refraction is unlikely to affect eye growth, a variety of studies show that detection of pattern, movement, and flicker may be markedly affected by changes in focus of as little as 0.5 D, even at eccentricities of 20–30 deg (e.g. Ronchi, 1971; Fankhauser and Enoch, 1962; Leibowitz et al., 1972; Jennings and Charman, 1981b; Artal et al., 1995; Wang et al., 1997; Anderson et al., 2001). Thus in real-world situations focus information for the periphery might potentially be derived from a variety of different stimuli.

Accommodation

In the fovea of the young eye, one response to defocus under photopic conditions is accommodation. It is reasonable to ask, then, whether an accommodation response can be elicited by stimuli which fall in the peripheral visual field. If it is, then focus error must be detectable. Any evidence for such a response would tend to support the hypothesis of Hoogerheide et al. (1971), although absence of response would not negate it.

Most early authors suggested that accommodation is entirely controlled by the central foveal area of about 30 min arc diameter. Thus Campbell (1954), on the basis of a study of the luminance threshold for accommodation, proposed that foveal cones were responsible, with no response being elicited by peripheral stimuli. Fincham (1951) found that subjects did not accommodate to a 10 min arc diameter white disk if they kept their direction of vision away from the object by more than the disk diameter. Only if fixation was closer than this did accommodation occur. These results led Crane (1966) and Toates (1972) to agree that the central area of the fovea was responsible for the response.

Several authors have, however, shown that stimuli falling outside the central fovea can still cause accommodation responses. Earliest of these was Whiteside (1957) who used narrow white annuli in a dark field as stimuli and found that, with fixation at the centre of the annuli, accommodation was stimulated for annular radii up to at least about 2.5 deg. A more comprehensive study was conducted by Bullimore and Gilmartin (1987a,b), who used targets consisting of uniform white disks against a black background: the disk radii subtended angles between 0.5 and 10 deg. Subjects fixated the centre of each disk and the slope of the accommodation response/stimulus curve was determined for the stimulus range 0 to 4 D. Slope varied from 0.9 for the case where the retinal target eccentricity for the disk edge was 0.5 deg, to about 0.25 for the 10 deg eccentricity (see Figure 9). The slope was found to fall approximately linearly with the minimum angle of resolution for the retinal location on which the edge contour fell (Charman, 1986): these results appear to be broadly compatible with DOF and blur detection data (Wang and Ciuffreda, 2004, 2005; Wang et al., 2006). A further study of this general nature was that of Gu and Legge (1987) who used black disks in a uniform white field as targets and varied the accommodation stimulus with negative lenses. Using disk radii of 1, 7, 15 and 30 deg, they found that there was an accommodative response for all the disks, even when the stimulus edge had an eccentricity of 30 deg (Figure 9).

Figure 9.

 Slope of the accommodation response-stimulus curve as a function of the eccentricity of the stimulus on the retina. The continuous curve shows mean data for 7 subjects using white disk stimuli in a dark surround (Bullimore and Gilmartin, 1987a, 1987b) and the dashed curves are for two subjects from Gu and Legge (1987), who used black disk stimuli in a white surround.

The few dynamic measurements made suggest that a rapid response is only initiated if the eccentricity of the target contour is less than about 10 deg (Phillips, 1974).

It is worth remarking that, on axis, an accommodation response can be initiated by changes in stimulus (about 0.1 D) which are smaller than the threshold for just-detectable blur (Ludlam et al., 1968; Kotulak and Schor,1986). Kotulak & Schor found that appropriate accommodation occurred to a stimulus whose vergence was varied sinusoidally at a temporal frequency of 1Hz with an amplitude of only about 0.12 D, whereas, under cycloplegia, the amplitude for perceptual recognition of blur was about 0.18 D. It may be, then, that sensitivity to defocus in the periphery is substantially better than the depths-of-focus that Figure 8 would suggest.

Using a rather different approach to evaluate the contribution of the peripheral retina to accommodation, Hennessy and Leibowitz (1971) found that when a target at a fixed distance was viewed foveally through a circular aperture placed in a dark field at a different distance to provide a potentially conflicting accommodation stimulus, the accommodation level varied with the distance of the aperture. A 1 deg aperture had a greater effect than a 4 deg aperture. In further experiments by Hennessy (1975) the influence of annular surrounds of inner diameters 5 or 8 deg and outer diameters 11.9 deg, on accommodation to a central target at a different distance was studied. The annulus was either dark, or a blue and white checkerboard. The checkered surround had much greater influence, presumably because it provided many more luminance boundaries for focusing purposes. Changing the distance of the checkered surround annulus with respect to the central target altered the accommodation response to the central target by about 1 D when the annular and central accommodation stimuli differed by 3 D.

An alternative approach to the question of the sensitivity of the peripheral retina to defocus, and any consequent accommodation, is given by dynamic measurements of accommodative convergence (Semmlow and Tinor, 1978). These suggest that an appropriate blur stimulus continues to generate an accommodative convergence response up to peripheral angles of at least 6 deg, although the magnitude of the response at this angle is only about 50% of that occurring for a foveal stimulus.

Overall, then, accommodation studies suggest that stimuli falling on the peripheral retina can alter the accommodation response of the eye and, in the presence of an axial accommodation target, can affect the response to the latter. There is, however, considerable disagreement between authors as to the exact nature of the response and no real understanding as to how the stimuli falling on different regions of the retina might summate in their effects.

It is clear, however, that in many normal visual environments a strong contribution to the accommodation response from stimuli at large field angles would be disadvantageous. Consider for example a stimulus at a distance of 0.4 m (2.5 D) on a textured flat surface, such as a desk top, inclined at 45 degree to the visual axis (Figure 10a). It can be seen that, in this vertical section, the accommodation stimulus provided by the surface is changing continually with the field direction. Using the sine rule and denoting the superior and inferior visual field angles as θs and θi respectively, we have ds/sin135 = 0.4/sin(45-θs) and di/sin45 = 0.4/sin(135-θi) so that the corresponding superior and inferior field accommodative stimuli are Ss = 1/ds = sin(45-θs)/0.4sin135 and Si = sin(135-θi)/0.4sin45.

Figure 10.

 (a) Geometry of accommodation stimuli for a near target on a flat surface inclined at 45 degrees to the visual axis (b) Corresponding variation in accommodation stimulus along the vertical and horizontal visual field meridians, S,I,T,N represent the superior, inferior, temporal and nasal sectors.

The resultant changes in stimulus as a function of field angle are shown in Figure 10b, those in the horizontal field meridian being included for comparison. Evidently the peripheral accommodation stimulus shows continuous variation with field angle in not only the vertical meridian but also in the horizontal and other meridians, the precise effects depending upon the geometry of the viewing environment, the working distance and other factors. If accommodation is to be appropriate to the fixated object, and the retina and peripheral refraction possess rotational symmetry about the visual axis, it is evidently necessary that the influence of the central stimulus should normally be dominant. Arguments of this type have, of course, been used previously to justify the development of “ramp” retinas and lower field myopia in ground-feeding birds or animals that maintain an approximately constant posture with respect to their environment and have an acuity which is relatively constant across the field (Fitzke et al., 1985; Hodos and Erichsen, 1990). It is of interest that, insofar as the situation assumed in Figure 10 approximates to many human working environments, it might actually be advantageous to have a refraction which, along the horizontal meridian, was myopic on axis and showed relative peripheral hyperopia (c.f. Figure 4).

Discussion

It appears that, while the studies of Hoogerheide et al. (1971) and Mutti et al. (2007) strongly suggest an association between relative peripheral hyperopia and the development of myopia in humans, a causative role for the former has yet to be confirmed. It may still be that the relative peripheral hyperopia observed in eyes that become myopic is simply associated with the more prolate or less oblate shape of the eye, rather than being causative of the myopia development. It could, for example, be speculated that an initially emmetropic eye with a more prolate shape is, perhaps for mechanical reasons, more vulnerable to excessive growth in axial length and the consequent development of myopia.

However, in contrast to this cautious view, there is good evidence to suggest that accommodation can be induced by stimuli lying several degrees (up to 30 deg in the Gu and Legge (1987) study, see Figure 9) outside the central fovea, although with progressively reducing efficiency as the eccentricity is increased. These changes mirror the changes in DOF and blur discrimination. Although the way in which stimulus defocus effects might be summed over extensive retinal areas is not clear, the checkerboard surround studies of Hennessy (1975) suggest that peripheral regions have greater influence if more luminance contours are available. In normal environments such contours are likely to be present at all eccentricities. Overall, then, the cited studies are compatible with the idea that axial focus might be influenced by the state of focus over at least the central 10 deg radius of visual field.

In fact most studies of peripheral refraction show that refraction changes only rather slowly over the central 10 deg radius of visual field (see e.g., Figures 5 and 6) and it is within this area that accommodation stimuli have their greatest effect. Nevertheless, sensitivity to focus change in detection tasks remains of the order of 0.5D or less out to larger peripheral angles, comparable to the relative peripheral hyperopia found at 30 deg in the study of Mutti et al. (2007) in children who became myopic (see Figure 2b), so that in principle focus information from the more peripheral retina could be used to derive a growth signal. It is striking that, in foveal vision, Ludlam et al. (1968) found that accommodation could be stimulated by stimulus changes as small as 0.1D, well below typical values of depth-of-focus of the eye as judged by a “just detectable blur” criterion (see also Kotulak & Schor, 1986). This may imply that a retinal response occurs at defocus levels significantly below those which result in conscious perception of blur.

Some additional insight into the possible importance of relative peripheral hyperopia comes from recent studies in different ethnic groups (see, for example, Saw (2003), Norton et al. (2005), and Logan and Gilmartin (2005) for reviews). Myopia is increasing in epidemic proportions in the Far East, with 84% of 16–18 year old children in Taiwan being myopic (Lin et al. 1999) and approximately 80% of young adults having myopia in Singapore (Saw 2003). The prevalence of myopia is reported to be over 70% in 15 year old children in urban populations in China, with the prevalence being lower in rural populations (He et al. 2004; He et al., 2009). Myopia prevalence in most other ethnic groups has been shown to range between 11 and 27% (Mutti & Zadnik 2000; Rose et al. 2001; Rose et al. 2002; Morgan et al. 2006). The increased prevalence of myopia in some populations would require that more individuals in these populations should have high levels of peripheral hyperopia. However, a study examining the retinal contour of myopic eyes found that anisomyopic Taiwanese-Chinese eyes showed a greater and more uniform relative expansion of the posterior retinal surface in the more myopic eyes (Logan et al., 2004). Therefore the increased prevalence of myopia in Taiwanese-Chinese populations may not be directly related to their peripheral refractions or eye shape.

In the presence of relative peripheral hyperopia, image quality in the periphery during distance vision can be improved by accommodation and it might be that such accommodation is itself a factor in myopia development. Walker and Mutti (2002) have shown that sustained accommodation causes the relative peripheral refractive error to become more hyperopic, so that in principle a positive feedback loop might be created which remained until accommodation was rendered unnecessary through the development of myopia. It is, perhaps, possible that individuals susceptible to developing myopia are more prone to develop peripheral hyperopia with accommodation. However, evidence on the effects of accommodation on RPRE is mixed.

In a study in which the peripheral refraction was measured for two accommodative states in emmetropic and myopic eyes, Calver et al. (2007) showed that peripheral astigmatism increased with increasing eccentricity, but found that there was no significant difference between refractive error groups except at 30 degrees eccentricity in the temporal retina. The effect of viewing distance on astigmatism or mean spherical equivalent error was not significantly different between the two refractive groups. In a similar study, Lundström et al. (2009) found that, in comparison to emmetropes, myopes had a smaller relative peripheral myopia, a larger asymmetry in defocus over the visual field, and an RPRE, which, either did not change or decreased with accommodation. In contrast, the RPRE of emmetropes increased with accommodation. The differences in RPRE between emmetropes and myopes therefore seem to be even larger during near work. This cannot be explained by simple modeling of different retinal shapes for the two types of eyes. Recently, Tabernero and Schaeffel (2009a) used a fast scanning photoretinoscope to measure peripheral refraction and showed that accommodation changes the refraction evenly across the central 90 degrees of visual field in emmetropic eyes: similar results were found by Davies and Mallen (2009) in both emmetropic and myopic eyes. Whatham et al. (2009) found only minor myopic shifts in RPRE with accommodation in myopes but make the important point that the increase in lag with increase in accommodation (effectively a hyperopic shift) has a major influence on the actual peripheral errors in focus. They found that the “best sphere” image lay behind the retina over the entire central 80 deg horizontal diameter of field studied, at all levels of accommodation (0 to 3.3 D stimulus levels). Thus it may be that the arguments that axial lags in accommodation are implicated in myopia development (e.g. Gwiazda et al., 1993) can be extended into the peripheral field. On the other hand, a longitudinal study has suggested that there is no significant relationship between myopia progression and near accommodative lag in children with mild and progressing myopia (Weizhong et al., 2008). Further studies on changes in peripheral refraction with accommodation in different ethnic groups and on progressing myopes are required to determine whether changes in peripheral refraction with accommodation play a crucial part in myopia progression.

It is interesting to note that, in the periphery, the existence of oblique astigmatism gives an optical odd-error cue to the sign of the defocus (Campbell and Westheimer, 1959; Howland, 1982): further cues may be provided by other monochromatic and chromatic aberrations. However, the possible role of oblique astigmatism in any hypothetical myopisation mechanism involving the retinal periphery remains obscure. Its value has already reached values of around 2 DC at field angles of about 30 deg (Figure 6), meaning that retinal images in the outer retina must always be markedly blurred, whatever the plane of focus. This may reduce the probability that the outer areas of the retina can play any important part in controlling eye growth. There is no evidence to suggest the existence of large differences in the magnitude of oblique astigmatism between refractive groups (Millodot, 1981), although there are known to be links between high axial astigmatism and the development of myopia in children (Gwiazda et al., 2000).

Peripheral refractive error can be measured with several subjective and objective techniques (see Fedtke et al. (2009) for a review) including open-field autorefractors, double-pass optical systems and Hartmann-Shack aberrometers. However, most instruments used for measuring peripheral refraction are not primarily designed for this purpose and therefore the participants often need to change their gaze position for obtaining peripheral refraction measurements. Peripheral refraction measurements with these instruments can be obtained either by turning the eye to fixate a series of horizontally/vertically spaced targets or by turning the head to view the same targets while maintaining central fixation and the eye in its primary position. Open-field autorefractors give comparable results for peripheral refraction irrespective of whether the measurements are made with eye turn or head turn (e.g. Radhakrishnan and Charman, 2008; Mathur et al., 2009). The scanning photorefractor recently developed by Tabernero and Schaeffel (2009a) overcomes the need for changes in head orientation or fixation, although currently measurements can only be made in the vertical meridian of the pupil.

As noted earlier, if it proves that relative hyperopia in the peripheral field drives myopia development in at least some individuals, then its measurement might prove a useful tool for identifying at-risk patients. Further, it will become necessary to reconsider the optical characteristics of various methods of correction. Whereas current spectacle lenses are designed to optimise foveal vision as the eye rotates, consideration will have to be given to peripheral correction with the eye stationary. As Tabernero et al. (2009) point out, many current spectacle designs, while correcting axial errors, produce what may be undesirable relative hyperopia in the periphery. They show that “radial refractive gradient lenses” in which the power changes in the positive direction with increasing distance from the lens centre result in a myopic lens-eye RPRE (see also Smith et al., 2005a): such spectacle lenses do, however, demand that the head rather than the eye must always be turned to view the object of regard, otherwise the foveal correction is incorrect.

It is interesting to note that those corneal methods of refractive correction for myopes which only correct a relatively small optical zone, such as laser assisted keratomileusis (LASIK) or orthokeratology (ortho-K), correct axial vision but leave the peripheral field myopic (Ma et al., 2005; Charman et al., 2006). This could explain why ortho-K reduced growth rates for vitreous chamber depth in children with progressing myopia, in comparison with age-matched controls who were corrected with single-vision spectacles, from 0.27 mm year−1 for the controls to 0.14 mm year−1 for the ortho-K subjects (Cho et al., 2005). A similar study comparing an ortho-K corrected group of children with a matched control group corrected with soft contact lenses again found that ortho-K produced slower increases in vitreous chamber depth (0.13 mm year−1 compared with 0.23 mm year−1) and hence in myopia progression (Walline et al., 2009). Although these results lend support to the idea that the myopic RPRE resulting from ortho-K helps to control myopia progression, it appears that even when younger subjects wear soft contact lenses there may be significant differences in the mean myopia progression, depending upon the lens material and wearing regime (Blacker et al., 2009). The ortho-K findings, while promising, cannot yet be regarded as confirmation of an important role for peripheral refraction in myopia development.

We note finally that the use of pharmacological agents such as atropine and pirenzipene has shown some apparent success in reducing rates of myopia progression in children (e.g. Siatkowski et al., 2004; Chua et al., 2006; Gwiazda, 2009). These agents are thought to act at the retinal level but nothing is known regarding any regional variations in their effects across the retina.

To summarise, the hypothesis that relative peripheral hyperopia is causative for axial myopia whereas relative peripheral myopia inhibits it, so that patterns of peripheral refraction can be used to predict the likelihood of future myopia development, remains inadequately proven. There is, however, substantial evidence to suggest that the state of focus in the peripheral retina might have some influence on foveal focus and refraction. Further exploration of the link between peripheral refraction and the development of axial ametropia is justified. Should a hyperopic RPRE prove to be undesirable, several types of optical correction are available to reduce its effect.

Ancillary