Profile of Refractive Errors in Cerebral Palsy: Impact of Severity of Motor Impairment (GMFCS) and CP Subtype on Refractive Outcome

Abstract

Purpose.: To describe refractive status in children and young adults with cerebral palsy (CP) and relate refractive error to standardized measures of type and severity of CP impairment and to ocular dimensions.

Methods.: A population-based sample of 118 participants aged 4 to 23 years with CP (mean 11.64 ± 4.06) and an age-appropriate control group (n = 128; age, 4–16 years; mean, 9.33 ± 3.52) were recruited. Motor impairment was described with the Gross Motor Function Classification Scale (GMFCS), and subtype was allocated with the Surveillance of Cerebral Palsy in Europe (SCPE). Measures of refractive error were obtained from all participants and ocular biometry from a subgroup with CP.

Results.: A significantly higher prevalence and magnitude of refractive error was found in the CP group compared to the control group. Axial length and spherical refractive error were strongly related. This relation did not improve with inclusion of corneal data. There was no relation between the presence or magnitude of spherical refractive errors in CP and the level of motor impairment, intellectual impairment, or the presence of communication difficulties. Higher spherical refractive errors were significantly associated with the nonspastic CP subtype. The presence and magnitude of astigmatism were greater when intellectual impairment was more severe, and astigmatic errors were explained by corneal dimensions.

Conclusions.: High refractive errors are common in CP, pointing to impairment of the emmetropization process. Biometric data support this conclusion. In contrast to other functional vision measures, spherical refractive error is unrelated to CP severity, but those with nonspastic CP tend to demonstrate the most extreme errors in refraction.

Cerebral palsy (CP) is a collective term for a group of nonprogressive disorders that affect movement and posture resulting from damage to the immature, developing brain. The resulting motor and intellectual impairments in people with CP are dependent on the area and extent of damage to the brain. Cerebral palsy is known to be associated with disorders of the visual system including high refractive errors, poor visual acuity, a high prevalence of reduced accommodative function, strabismus, and nystagmus. ^1–9 Ghasia et al. ² have shown that the level of motor impairment in CP is related to the extent of the visual acuity deficit. 

Studies in which visual function in CP has been examined have included subjects from selected populations, such as individuals who attended a school, support group, or hospital clinic specifically for children with CP. ^2–4,7,8 Examining selected CP cohorts whose motor characteristics do not reflect the full extent of the condition (including the most mildly affected) may not accurately describe the visual difficulties associated with CP. For the first time, the present study uses a population-based approach to investigate refractive error with respect to CP subtype and the severity of intellectual and motor impairment, and the presence of communication difficulties. 

Participants were recruited from the Northern Ireland Cerebral Palsy Register (NICPR). This register was designed to provide a systematic approach to monitoring and surveillance of CP in a geographically defined population. ¹⁰ Each person on the register receives a confirmed diagnosis of CP after a pediatric assessment at 4 to 5 years of age. The assessment by a pediatrician follows a standard protocol and includes a record of the type and severity of motor impairment and the presence and severity of associated impairments including intellect and communication. 

The NICPR uses both the Gross Motor Function Classification Scale (GMFCS) ¹¹ and the case definition and classification scheme described by the Surveillance of Cerebral Palsy in Europe project (SCPE). ^12,13 The GMFCS categorizes individuals with CP according to self-initiated movements and uses five levels ranging from level I, walks without limitations, to level V, transported in a manual wheelchair. It has been shown to be a valid and reliable method of classification. ^14,15 Intellectual impairment is considered present where the intelligence quotient (IQ) is <70, moderate intellectual impairment is between 70 and 50, and severe intellectual impairment is <50. Communication difficulties are recorded as either present or absent. 

In addition to clinical characteristics of the individual, the NICPR also provides information regarding the subtype of CP according to the SCPE scale. The condition may be classified into three main subtypes, each primarily affecting different areas of the developing brain. ^16,17 Spastic CP affects ∼75% to 94% of people with CP and results in tightness and stiffness of the muscles; dyskinetic CP affects 3% to 5% of people with CP, causing uncontrolled, slow, writhing movements affecting hands, feet, legs, face, and tongue; ataxic CP, the least common form accounts for ∼1% to 4% of cases and results in balance and co-ordination problems. ^10,16–22  

In the present study, a novel population-based approach was used to examine whether refractive errors are influenced by the type and severity of CP and presents, for the first time, ocular biometry data to explore the relation between eye size and shape and refractive status in CP. 

Methods

Participants

Five pediatricians from different regions of Northern Ireland participated in the study and sent out an invitation to all those with CP under their care who fulfilled the study's criteria: a confirmed diagnosis of CP and age between 4 and 23 years. Ethical approval was obtained from the University of Ulster Research Ethics Committee (ORECNI), the Belfast Health and Social Care Trust, and the five Education and Library Boards involved in the study. The research adhered to the tenets of the Declaration of Helsinki. Informed consent was obtained from the participants and/or their parents after an information sheet had been received and an explanation of the nature and possible consequences of the study was given. 

One hundred eighteen subjects, aged 4 to 23 years (mean, 11.64 ± 4.06 years; 69 male), participated in the study. According to GMFCS classification, 8 participants were level I, 38 level II, 27 level III, 14 level IV, and 26 level V. Five participants were unclassified by GMFCS. Ninety-six participants had the spastic subtype of CP, 11 the dyskinetic form, and 6 the ataxic subtype. Five participants had CP of unclassified subtype. Intellectual impairments varied from no (n = 64) to severe learning difficulties (n = 28). Fifty-six participants (47%) were classified as having communication difficulties; the remainder did not. The clinical characteristics of the participants accurately reflected the overall Northern Ireland CP population. ¹⁰ Thirty-six (40%) children attended mainstream schools, and 78 attended schools for children with special educational needs. The remaining four participants had left school. Forty-seven schools were visited in total, which included 13 schools for children with special needs and 34 mainstream schools. 

A group of 128 children without CP, aged 4 to 16 years (mean age, 9.33 ± 3.52 years; 64 boys), attending one primary and one postprimary mainstream school, were included in the study, to provide comparative refractive error data. 

Procedure

In all participants with CP, refractive error status was examined by either cycloplegic or distance static retinoscopy. When possible, cycloplegic refractions were performed (1 drop 1% cyclopentolate hydrochloride each eye, retinoscopy performed after 30 minutes). When permission to instill cycloplegic eyedrops was refused, distance static retinoscopy was performed. In all the control participants, distance static retinoscopy was used for the refractions. 

Additional measures of ocular biometry were also attempted on a subgroup of participants with CP, when equipment for recording of axial length and corneal curvatures was available. This subgroup contained 44 participants, selected solely on the basis that the test equipment was available to the researchers during their participation. Their clinical and visual characteristics did not differ significantly from the group as a whole. Axial length was measured with an ocular biometer (IOLMaster; Carl Zeiss Meditec, Inc., Oberkochen, Germany) and corneal curvature was measured with a handheld autokeratometer (KM 500; Nidek, Tokyo, Japan) after cycloplegia. 

Results

Success Rates

Measures of refractive status were obtained successfully from both eyes of all participants in the CP (55% cycloplegic retinoscopy, 45% distance static retinoscopy) and control groups. In the biometry subgroup (n = 44, all with CP), axial length measurements were successfully recorded from 36 (82%) participants and corneal curvature measurements from 35 (80%) participants. Lack of success was attributable to physical limitations. 

Classification

The most ametropic meridian (MAM) was derived for each eye, and refractive errors were categorized according to the following classifications to allow for comparison with the data of Ghasia et al. ² MAM was used to characterize refractive error, rather than the commonly used mean spherical equivalent, because the latter method for calculating spherical error is contaminated by high levels of astigmatism, such as those prevalent in CP.

•  
  Emmetropia: MAM > −0.75 to +1.00 D
•  
  Low to moderate hypermetropia: MAM > +1.00 to +4.00 D
•  
  High hypermetropia: MAM > +4.00 D
•  
  Low to moderate myopia: MAM −4.00 to > −0.50 D
•  
  High myopia: MAM > −4.00 D

Astigmatism and anisometropia were defined as follows:

•  
  Significant astigmatism: ≥1.00 DC
•  
  Significant anisometropia: ≥1.00 D between the corresponding meridians of the right and left eyes.

The type of astigmatism was described as against the rule (ATR), with the rule (WTR), or oblique, ²³ according to the following classifications:

•  
  WTR, axis of negative cylinder 180° ± 15°
•  
  ATR, axis of negative cylinder 90° ± 15°
•  
  Oblique, all other cases

MAM did not vary significantly with age in the CP group (linear regression analysis, r = −0.11, P = 0.3). A paired t-test demonstrated that there was no statistically significant difference between the refractive errors from the right and left eyes (t = 1.07; P = 0.29). Therefore, for further analyses, only right eye data will be considered. MAM refractive error data are presented for each participant group in Figure 1. 

Figure 1.

 

Frequency distribution of MAM refractive error for CP (▧) and control group ().

The right eye MAM in the CP group ranged from −15.00 to +14.00 D (mean, +1.03 ± 3.54 D; SD). The MAM of the right eyes of the control group ranged from −3.50 to +8.00 D (mean, +0.53 ± 1.13 D). Kurtosis for the CP data presented in Figure 1 is 4.8 for the CP data and 17.4 for the control data, indicating less leptokurtosis in the CP data. 

More individuals with CP had high myopia, low-moderate hypermetropia, high hyperopia, astigmatism, and anisometropia (χ² P < 0.05) compared with the control group (Table 1). The prevalence of emmetropia was significantly higher in the control group (χ² P < 0.0001). 

Table 1.

 

Distribution of Refractive Error Type in Participants with and Those without CP

Refractive Error Type	CP		Control
	n	%	n	%
Myopia
Low-moderate	10	8.5	7	5.5
High*	11	9.3	0	0
Emmetropia*	33	28.0	108	84.4
Hypermetropia
Low-moderate*	50	42.4	11	8.6
High*	14	11.9	2	1.6
Astigmatism*	43	36.2	4	3.1
Anisometropia*	21	17.8	9	7.0

 CP group, n = 118; control group, n = 128. Some participants had astigmatism and/or anisometropia in addition to a spherical refractive error and are therefore represented in more than one category.

*  Significant difference between CP and non-CP groups.

Of those CP participants with significant astigmatism, 25 (58.1%) were classified as having WTR, 9 (20.9%) as having ATR, and 9 (20.9%) as having oblique astigmatism. 

Participants who were not classified by the GMFCS, SPCE, or intellectual impairment were excluded from the following analyses. 

Refractive Error and Clinical Characteristics

To examine the associations between clinical CP variables (severity of motor impairment, severity of intellectual impairment, presence of communication difficulties and CP subtype) and refractive error (MAM, astigmatism and anisometropia), refractive error data were examined as continuous variables (i.e., the magnitude of the MAM or anisometropic or astigmatic error) by MANOVA, and categorically (i.e., type of refractive error—high hyperopia, low-moderate hyperopia, emmetropia, low-moderate myopia, high myopia—presence of astigmatism, and presence of anisometropia), by χ² analysis. Because of the population-based nature of the data, there were few participants classified with dyskinetic (n = 11) or ataxic (n = 6) CP, making formal statistical analysis using all CP subtypes inappropriate. Instead, data were analyzed by pooling ataxic and dyskinetic CP and comparing the data against the spastic subtype. Other studies have also grouped CP into spastic subtype and other subtypes. ^1,2  

Spherical Refractive Error.

A multivariate analysis revealed no significant association between the magnitude of the MAM and CP variables (MANOVA: Wilks lambda F _{(4 , 113)}=1.25, P = 0.29; Fig. 2). 

Figure 2.

 

The absolute median (central line) and interquartile range (box) of absolute MAM across different levels of (A) motor impairment (GMFCS levels I–V) and (B) intellectual impairment. All box-and-whisker plots in this publication use a small square symbol to depict the mean and the whiskers indicate the 5% and 95% percentiles.

However, postestimation comparisons between groups demonstrated a significant association between CP subtype and MAM refractive error (P < 0.05), with nonspastic CP being associated with significantly greater magnitudes of spherical error. Figure 3 plots the absolute magnitude of MAM refractive error for the spastic and nonspastic subtypes of CP. A one-way ANOVA on the absolute magnitude MAM refractive errors confirmed that the nonspastic CP group had a significantly higher mean than did the spastic CP group (F _{(1 , 111)} = 8.2, P < 0.01). 

Figure 3.

 

The absolute median (central line) and interquartile range (box) of absolute MAM refractive error across spastic and nonspastic CP subgroups.

When considering type of spherical refractive error, no significant associations were found (χ² P > 0.05 in all cases). 

Astigmatism.

Multivariate analysis relating the magnitude of astigmatic error to clinical variables revealed a significant association across CP variables (MANOVA: Wilks λ F_(4,113) = 2.82, P = 0.03). Postestimation comparisons revealed the association between the magnitude of astigmatism and the level of intellectual impairment to be significant (P < 0.05), with those with the mildest intellectual impairment demonstrating smaller amounts of astigmatic error (Fig. 4). A χ² analysis demonstrated a significant relation between the presence of astigmatism and the most severe intellectual impairment in CP (P < 0.001). No other clinical variables demonstrated a significant relation with presence or absence of astigmatism (P > 0.05). However, where significant astigmatism was present, the type of astigmatism varied with the type of CP. None of those with significant astigmatism in the nonspastic subgroup (n = 6) had oblique astigmatism, whereas nine (26%) of those from the spastic subgroup demonstrated this type of astigmatic error. 

Figure 4.

 

The median (central line) and interquartile range (box) of refractive astigmatism across levels of intellectual impairment in CP group.

Anisometropia.

No significant relations were demonstrated between the magnitude (MANOVA: Wilks λ F _{(4 , 113)} = 0.22, P = 0.92) or presence (χ² P > 0.05) of anisometropia and clinical variables. 

Ocular Biometry and Refractive Error

Axial length was a strong correlate of refractive error, with 81% of the variance in refractive error attributable to axial length. Linear regression (Fig. 5) revealed a significant relationship between AL and MAM refractive error (r = −0.90, P < 0.0001). Axial length ranged from 18.94 to 27.47 mm (mean, 22.8 ± 1.59). 

Figure 5.

 

Axial length against MAM refractive error for CP participants (n = 36).

Corneal curvature ranged from 7.28 to 8.45 mm (mean, 7.75 ± 0.24). The inclusion of corneal dimensions in the analysis between axial length and refractive error did not strengthen the relationship. Linear regression of the axial length/corneal radius ratio to MAM refractive error was also significant, but the Pearson correlation coefficient remained unchanged (r = −0.90, P < 0.0001). 

The prevalence of astigmatism was high in the CP group, and evaluation of corneal characteristics in a subgroup (n = 36) demonstrated that astigmatic errors could be explained by astigmatic corneas (linear regression analysis, r = 0.72; ANOVA F _(1,32) = 35.01, P < 0.0001). 

There was an insufficient number of anisometropic individuals from whom biometric data were collected (n = 3) to allow an investigation of the contribution of biometric parameters to interocular refractive error differences. 

Discussion

The present study presents refractive error data for a larger, nonselected group of children and young adults with CP than has previously been reported. In agreement with previous studies of CP the present study found a wide range of refractive errors in the CP group, ^2,4,16,24,25 and a significantly higher proportion of low-moderate and high refractive errors compared to the control group. ^1,2 The study had four main findings. 

Refractive error is more common in CP than in typically developing children. Published data from cohorts of non-CP newborns describe a normal distribution of refractive errors, whereas older children and adult refractive errors demonstrate a markedly skewed (leptokurtic) distribution with a peak at the attainment of emmetropia. ²⁶ The distribution of our CP refractive error data is quite different from that of the developmentally normal population (and our control group), with a less leptokurtic profile. Sobrado et al. ¹ suggest that there is a failure to emmetropize in CP. The distribution of refractive errors in our data points to an individual failure to emmetropize, but prospective data would be needed to confirm this hypothesis. 

Axial length is a stronger predictor of refractive error in CP than in typically developing children. Our biometric data suggest that, if emmetropization is impaired or delayed in CP, it may be due to a failure in compensatory feedback mechanisms controlling the growth of axial length. Axial length is a stronger predictor of refractive error in CP than in the typically developing population. ^27–29 In other studies of typically developing children and adults, correlation coefficients for refractive error and axial length are r = 0.44 (Ojaimi et al. ²⁹ ), r = 0.47 (European Caucasians, Ip et al. ²⁸ ), and r = 0.76 (Grosvenor and Scott ²⁷ ). These data are compared with r = 0.9 in the present study. In typically developing children, the correlation between refractive error and biometric measures is strengthened by the inclusion of corneal curvature. Grosvenor and Scott ²⁷ stated that “the axial length/corneal radius ratio is the most significant determinant of the refractive state of the eye.” Their conclusion is not true of the current CP data. In the developmentally normal population, crystalline lens thinning has been shown to compensate for increasing axial length, to promote and maintain emmetropia. ^30,31 It is interesting to speculate that this feedback mechanism is impaired in CP, resulting in frequent, high refractive errors. 

Spherical refractive error is not related to the severity of CP but subtype impacts on refractive outcome. CP has an impact on refractive development, resulting in an increased prevalence of significant errors. However, our data suggest that this impact is independent of the severity of the motor deficit in CP. This finding contrasts starkly with data relating to visual functions such as visual acuity, binocularity, and accommodation, which have been shown to deteriorate significantly as the level of motor impairment increases. ^2,6  

To date, the only other study to explore the relationship between clinical characteristics of CP and refractive error is that of Ghasia et al. ² They described and compared visual and motor deficits in a group of 50 children and young adults with CP, according to the GMFCS. They deliberately recruited in such a way as to obtain an equal number of participants (n = 10) at each level of GMFCS. In agreement with the present study, Ghasia et al. ² reported that moderate to high refractive errors are common across all severities of motor impairment in CP, with low-moderate hyperopia (+1.00 to +4.00 D) being most prevalent. They suggested that children with the highest level of motor impairment are most at risk of high myopia (> −4.00 D MSE). However, only eight children in their study were highly myopic, and only three of these were in the most impaired classification, making strong conclusions problematic. The present study featured 11 highly myopic participants, 3 of whom were graded level V (most severe motor impairment) by the GMFCS, 5 at level II, and 3 at levels III and IV. Our data do not support the conclusion that more severely physically impaired individuals with CP are more likely to be highly myopic—rather, that CP is associated with moderate and high refractive error across all severities. 

Although the severity of the motor impairment did not influence the type or level of refractive error found in the present study, our data suggest that the type of CP (spastic or nonspastic) has a differential effect on the refractive outcome. Although the number of nonspastic individuals in the present study was small (n = 17), reflective of the underlying sample population, nonspastic CP was associated with higher spherical refractive errors than was spastic CP. Ghasia et al. ² also have reported high refractive errors among their sample of nine nonspastic individuals. 

Intellectual impairment is associated with astigmatism. Previous reports have not investigated how CP subtype or severity relates to astigmatism. In the present study, the least intellectually impaired individuals had significantly less astigmatism on average than their more impaired peers, and those with spastic CP were more likely to have oblique astigmatism, which is generally considered less common in the neurologically normal population. ³² Measures of corneal curvature in a subgroup of participants demonstrated that these errors were attributable to astigmatic corneal shape and may be due to a failure of the normal process of eye growth during which corneal astigmatism decreases as the corneal shape flattens with age. 

The inclusion of both cycloplegic and noncycloplegic refractive error data in the present study may be criticized. However, previous published works on refractive status in comparable cohorts ^3,4,9 have used similar methods for pragmatic reasons. Cycloplegic retinoscopy is primarily used to ensure that latent hyperopic refractive errors are elicited by the retinoscopist. This problem is particularly evident when testing infants and young children with high levels of accommodative facility. The known limitations of accommodative function in the CP population, the age group tested, the use of distance static retinoscopy, and the experience of the refractionists in the present study mitigate the inclusion of noncycloplegic data. Yeotikar et al. ³³ demonstrated that noncycloplegic distance static retinoscopy is as effective as cycloplegic retinoscopy in healthy children aged 7 to 16 years of age. 

Researchers have described refractive error in CP using mean spherical equivalent (MSE). ^1,2 This approach can be problematic when astigmatism is prevalent, and in the present study we chose to present refractive data in terms of the MAM. However, our findings remained unchanged when analyses were replicated using MSE, except to reveal a significant relation between high hyperopia and the nonspastic form of CP (χ² P = 0.015). 

Conclusions

The results in the present study support those in reports that high levels of refractive error in CP indicates a failure or impairment of the emmetropization process. Biometric data, not previously available for a CP cohort, support this conclusion. 

In contrast to other functional vision measures such as visual acuity, spherical refractive status is not related to the severity of CP. All children and young adults with CP, regardless of their neurologic status, should be considered at risk of significant refractive errors which, if uncorrected, may compound existing visual difficulties relating to their condition. In particular, the frequency with which significant levels of hyperopia are present among individuals with CP is noteworthy. It is known that accommodative function is commonly impaired in CP, ⁶ exacerbating the visual deficit and blur caused by uncorrected hyperopic errors. Furthermore, those with nonspastic types of CP demonstrate the most extreme errors of refraction. 

Footnotes

 Supported by The College of Optometrists, UK Research Scholarship (JFM); The Department of Health, Social Services, and Public Safety (DHSSPS) R&D Office, NI RRG4.4 Project 2 (J-AL), and Nuffield Foundation Grant SCI/180/96/41/G (KJS). The NICPR is funded by Department of Health and Social Care in Northern Ireland (DHSSPS/HSC). Its role is to facilitate standardization of CP data to comply with recognized classifications of CP subtype, gross motor function, and intellectual impairment.

Footnotes

 Disclosure: K.J. Saunders, None; J.-A. Little, None; J.F. McClelland, None; A.J. Jackson, None

The authors thank Nan Hill, Moyra Stewart (Consultant Pediatricians), and Jackie Parkes (NICPR) for help with recruitment of participants for the study and all the children and young adults and their parents for their participation.