Heritability of Refractive Error and Ocular Biometrics: The Genes in Myopia (GEM) Twin Study

Abstract

purpose. A classic twin study was undertaken to assess the contribution of genes and environment to the development of refractive errors and ocular biometrics in a twin population.

methods. A total of 1224 twins (345 monozygotic [MZ] and 267 dizygotic [DZ] twin pairs) aged between 18 and 88 years were examined. All twins completed a questionnaire consisting of a medical history, education, and zygosity. Objective refraction was measured in all twins, and biometric measurements were obtained using partial coherence interferometry.

results. Intrapair correlations for spherical equivalent and ocular biometrics were significantly higher in the MZ than in the DZ twin pairs (P < 0.05), when refraction was considered as a continuous variable. A significant gender difference in the variation of spherical equivalent and ocular biometrics was found (P < 0.05). A genetic model specifying an additive, dominant, and unique environmental factor that was sex limited was the best fit for all measured variables. Heritability of spherical equivalents of 88% and 75% were found in the men and women, respectively, whereas, that of axial length was 94% and 92%, respectively. Additive genetic effects accounted for a greater proportion of the variance in spherical equivalent, whereas the variance in ocular biometrics, particularly axial length was explained mostly by dominant genetic effects.

conclusions. Genetic factors, both additive and dominant, play a significant role in refractive error (myopia and hypermetropia) as well as in ocular biometrics, particularly axial length. The sex limitation ADE model (additive genetic, nonadditive genetic, and environmental components) provided the best-fit genetic model for all parameters.

Refractive error can be broadly defined as an optical condition wherein distant light rays focus before the retina (myopia) or beyond the retina (hypermetropia), with a blurred image as the consequence. The optical properties of the eye are determined by several components, including corneal curvature (CC), anterior chamber depth (ACD), lens power, and the ocular axial length (AL). ¹ The development of myopia has been strongly associated with structural changes of the eye, in particular AL elongation. ² Astigmatism is also a refractive error that usually accompanies myopia and hypermetropia. Astigmatism, most likely develops as a result of irregularities in the cornea (corneal astigmatism; CA) or lens (lenticular astigmatism). ¹ Epidemiologic studies have collectively shown that the prevalence of refractive errors is rapidly escalating, particularly in urbanized areas of southeast Asia. ^{3 4} As a result, the global initiative to eliminate avoidable blindness by the year 2020 has included refractive error as one of its five priority eye diseases. ⁵  

Most studies on refractive error have concentrated on myopia, ^{6 7 8} with little or no emphasis on hypermetropia or astigmatism. Findings in these studies have indicated that myopia is a complex eye disease resulting from both genetic and environmental risk factors. The principle environmental risk factor consistently identified in numerous epidemiologic studies is near work (typically measured as books read per week) explaining approximately 10% of the total variance in myopia. ⁹ In addition, support for a genetic origin of myopia comes from several sources, including family linkage studies in which several myopia loci (MYP-1 to -12) have been identified for a range of myopic severities, ^{6 7 8 10 11} and family correlation studies in which children with myopic parents are at a greater risk of development of myopia than are children with nonmyopic parents. ^{1 12} Further studies based on the heritability of refractive errors, including myopia in twins, have also consistently shown the likely involvement of a major genetic component to refractive error. ^{13 14 15 16 17}  

Classic twin studies primarily serve to quantify the genetic and environmental components that influence a trait. The twin methodology is based on the comparison of intrapair correlations between genetically identical (monozygotic [MZ]) twin pairs and nonidentical (dizygotic [DZ]) twin pairs who share up to 50% of their genes. A higher intrapair correlation in MZ twin pairs compared with DZ twin pairs generally supports a genetic component to the trait. The largest twin study to date on refractive error ¹⁷ examined 506 female twin pairs aged between 50 and 79 years and reported that the correlation for refractive error in MZ twin pairs (>0.8) was almost double that in DZ twin pairs (>0.4), thus supporting a significant genetic role in the development of refractive errors. The best-fitting model reported in that study was the AE model, which suggests that the variance of refractive error was mainly explained by both additive genetic effects (A) and a unique environmental component (E). ¹⁷ However, other areas not explored in the study but of importance in refractive error include the assessment of ocular biometrics, sex comparisons, and a wider age range. 

In the current report we undertook a classic twin study to assess the heritability of refractive errors and ocular biometrics in adult twins of both genders based on an Australian twin population of mainly white background. 

Materials and Methods

Subjects

Participants (MZ and DZ twin pairs) were recruited by a mailing to the Australian Twin Registry (ATR) in Melbourne, Australia. All twins of either gender, aged 18 years or older who were registered on the ATR living in the state of Victoria were invited to participate in the study. Individuals with ocular disorders or retinal disease that may lead to changes in refractive error such as amblyopia (greater than a 2-line Snellen difference between the eyes), strabismus, lens opacification (Wilmer grading system), glaucoma, retinopathy, or keratoconus were excluded from the main analysis. There were a total of 94 individuals with any form of cataract as defined by the Wilmer grading system. Individuals with visually significant cataract (visual acuity <6/9, n = 20 participants) were excluded from the main analysis. However, individuals with anisometropia (>5-D difference) were included in the main analysis. Individuals with connective tissue disease such as Marfan’s or Stickler syndrome were also excluded from the main analysis. The latter conditions were identified by the individual’s medical history obtained via the standard questionnaire, and no cases were reported in the present study. 

Each twin provided informed consent before examination. The consent adhered to the tenets of the Declaration of Helsinki, and the study was approved by the Royal Victorian Eye and Ear Hospital (RVEEH) Human Research and Ethics Committee and by the Australian Twin Registry. 

All twins underwent a comprehensive eye examination, including a visual acuity assessment, subjective and objective (cycloplegic) refraction, slit lamp examination, and ocular biometry, and a blood sample was collected by venipuncture. Each individual also completed a standard questionnaire consisting of information on demographics, medical and family history, past ocular history, and zygosity. 

Zygosity

Twin zygosity was determined by a series of questions recommended by the ATR. ¹⁸ These questions were validated as having a 95% accuracy in determining correct zygosity. ¹⁹ In the present study, most of the twins themselves were aware of their zygosity, because of their upbringing, physical and psychological similarities, or zygosity testing in other twin studies. Moreover, in cases in which zygosity was uncertain (n = 20 twins), standardized genotyping with a panel of 12 polymorphic markers (Linkage Mapping Set version 2; Applied Biosystems, Foster City, CA) ²⁰ was performed by the Australian Genome Research Facility, Melbourne. The results of this genotyping confirmed the zygosity as previously determined by the examiner based on the series of twin questions and the assessment of physical characteristics in all cases. 

Examination

Visual acuity was assessed with and without correction using the National Vision Research Institute (NVRI, Melbourne, Australia) logMAR chart (logarithm of the minimum angle of resolution) at 3 m in normal room lighting. Subjective refraction was performed using a modified version of the Early Treatment of Diabetic Retinopathy Study protocol. ²¹ The biomicroscope examination was used to assess the integrity of the anterior segment of the eye and to exclude any lens opacities. 

Objective refraction and ocular biometry was performed after dilation (Tropicamide 1%). Dilated autorefraction was measured (model KR 8100; auto-refractor; Device Technologies, Melbourne, Australia). A total of three readings were taken for each eye and the average value recorded. Results for each eye were converted to their spherical equivalent (SE). Myopia was defined as equal to or worse than −0.50 diopter sphere (DS), and hypermetropia was defined as equal to or worse than +0.50 DS. Emmetropia was defined as between −0.499 and +0.499 DS. The intraocular lens (IOL) master (Carl Zeiss Meditec, Oberkochen, Germany) was used to obtain ocular biometry measurements on AL (anteroposterior diameter), keratometry (corneal curvature), and ACD in both eyes. Corneal astigmatism was defined as the difference between the keratometry readings K1 and K2 obtained from the IOL master. At least three readings were taken for each ocular biometry measurement to ensure consistency and reproducibility in the results. 

Variance Component Genetic Modeling

Genetic modeling is primarily used to quantify the proportion of phenotypic variance attributable to either genetic or environmental factors. The phenotypic variance is then separated into additive genetic (all genes contribute equally to the final phenotype) (A), non-additive genetic (one gene contributes more or less to the final phenotype) (D), common environmental (shared environment between siblings reared together) (C), and unique environmental (individual specific effects) (E) components. Genetic modeling is then used to determine the most parsimonious model (a model that explains both the variance and covariance with the least parameters) for each variable. The best-fitting model was determined based on maximum-likelihood and χ² tests. The sex limitation model was applied in the analysis, as the variances for measured variables were significantly different between the men and women. Covariates were adjusted for in the variance component modeling. All measured variables fitted a normal distribution. Astigmatism also resulted in normal distribution after log transformations. 

Statistical Packages

All descriptive statistics were analyzed with commercial software (Statistical Package for the Social Sciences, ver. 12.1; SPSS, Inc., Chicago, IL). Quantitative genetic modeling was achieved by using the Mx statistical program. ²²  

Results

Demographic Characteristics

A total of 345 (56.3%) MZ and 267 (43.7%) DZ twin pairs aged 18 to 88 years were recruited and examined between November 2004 and August 2005. Of all twins (n = 1224), 400 (32.7%) were men and 824 (67.3%) were women. There were significantly more female than male twins within both the MZ (female 473, 68.5%; male 217; 31.5%, P < 0.001) and DZ (female 356, 66.7%; males 178, 33.3%; P < 0.001) twin groups—a common phenomenon in twin studies. The mean age of all twins was 52.36 ± 15.43 years (range, 18–86 years) with no significant difference between mean ages of the MZ (52.11 ± 15.85 years) and DZ (52.63 ± 14.96 years, P = 0.56) twins. No significant differences in mean SE, AL, ACD, and CA were evident between the right and left eyes of all twins (P > 0.05). Analysis in this study was therefore undertaken only for the right eye. Baseline characteristics for both the MZ and DZ groups are shown in Table 1 . 

Heritability Estimates

Intrapair correlations for SE were significantly higher in the MZ (r = 0.82) than in the DZ (r = 0.36; P < 0.001; Fig. 1 ) twin pairs. Ocular biometric measures were also significantly higher in the MZ than in the DZ twin pairs, with AL correlations being r _MZ = 0.90 and r _DZ = 0.38 (P < 0.001), and ACD and CA being r _MZ = 0.70, r _DZ = 0.30 (P < 0.001) and r _MZ = 0.48, r _DZ = 0.13 (P < 0.05), respectively. In addition, the correlations for corneal curvature (CC) were found to be significantly higher in the MZ (r = 0.84) than in the DZ (r = 0.47) twin pairs. The significantly higher correlation observed in the MZ than in the DZ twin pairs for all parameters further supports a genetic basis to the development of refractive error. Correlations for refractive error and ocular biometrics measures were also significantly higher in both the male and female MZ twin pairs than in all the DZ pairs (Table 2) . 

In this twin sample, the DZ intrapair correlations were consistently less than half the MZ intrapair correlation for refractive error and ocular biometrics, suggesting that nonadditive genetic effects provided a greater source of variation than did common environmental effects. A sex limitation ADE model was also found to provide the best-fit model for refractive error and all ocular biometrics (Table 3) . Heritability estimates using this model for all measured variables ranged from 50% in male CA to as high as 94% in male AL (Table 4) . Heritability estimates for SE were 88% and 75% in the men and women, respectively (Table 4) , where additive genetic effects explained the most of the variance (men, 58%; women, 47%) compared with dominant genetic effects (men, 30%; women, 28%). Unique environmental effects explained only 12% and 25% of the variance in SE in the men and women, respectively. In contrast, dominant genetic effects accounted for most of the variance compared with either additive genetic or unique environmental effects for all ocular biometric measures (Table 4) . The only exception to this was in AL in the female twin pairs, for which additive genetic effects explained a slightly higher proportion of the variance (42%) compared with dominant genetic effects (36%; Table 4 ). The heritability estimate for AL was found to be the highest of all parameters—more than 90%, with unique environmental effects explaining less than 10% of the variance in both genders (Table 4) . Overall, heritability estimates supported a major role for a genetic basis in both refractive error and ocular biometric measures. 

Covariates

Several covariates were assessed for their involvement with refractive error and ocular biometrics. Gender was found to be significantly associated with all parameters (P < 0.05). Age was found to be associated only with a higher SE (P < 0.05) and a lower ACD. Higher education was associated with SE (lower) and AL (higher; P < 0.05). Height was associated with a lower SE and higher AL and ACD, whereas weight was associated with AL and CA (P < 0.05). 

Discussion

The purpose of this twin study was to quantify the genetic contribution to refractive error based on assessment of refraction and ocular biometrics in an adult twin population of both sexes. Case-wise concordance, intrapair correlations, and heritability estimates all provided evidence that genes play a major role in refractive error and ocular biometrics. 

The effects of additive and dominant genes explained 88% and 75% of the total variance for refractive error in the men and women, respectively. These high heritability estimates for refractive error strongly concur with the previous findings of Hammond et al., ¹⁷ who reported a heritability estimate for refractive error of 84% to 85% in women. ¹⁷ In their study, all the variance (genetic) was explained by additive genetic factors. However, in the present study, additive genetic effects accounted for only 47% of the variance in the women and 58% of the variance in the men, with the remaining 28% and 30%, respectively, resulting from dominant genetic factors. Thus, the involvement of both additive and dominant genetic effects suggests the involvement of multiple genes in the etiology of refractive error. 

In addition, we were able to show high heritability for AL, with additive and dominant genetic effects accounting for more than 90% of the variance for this trait in both sexes. These findings support previous heritability estimates for AL of 83.3% ¹³ and 94%, ¹⁵ after adjustment for significant covariates, and suggests that AL is strongly associated with the development of refractive error. AL measurement may therefore provide the best measurable predictor of the development of refractive error. Lyhne et al. ¹⁵ examined 114 twin pairs aged between 20 and 45 years and excluded those with myopia greater than −6.00 DS; however, we were able to confirm their heritability findings for AL using a larger sample size (>600 twin pairs) and including a wider refractive error and age range. 

This is the third twin study to have investigated the genetic contribution to astigmatism. An earlier study by Teikari and O’Donnell ²³ showed no significant differences in concordance for astigmatism between MZ and DZ twin pairs, thus suggesting a major environmental role in the development of astigmatism. However, that study had a small sample size (n = 42 MZ and 30 DZ twin pairs) and used a highly selective sample reflecting the method of data collection used in that study. This relied on spectacle wearers’ sending their latest prescription, and thus twins with uncorrected astigmatic errors and those who failed to send their prescription were not represented in the study. In contrast, in their twin study, Hammond et al. ¹⁷ reported the opposite finding with dominant genetic effects accounting for between 42% and 61% of the variance of CA. In the present study, t a similar proportion of the variance (50%) in CA was accounted for by dominant genetic effects, with unique environmental variance contributing 50% and 40% in the men and women, respectively. Although, research on the genetics of astigmatism is limited, our study and that of Hammond et al., ¹⁷ together with a family study, ²⁴ suggest that a dominant genes may account for approximately half of astigmatism. Moreover, the significantly higher MZ intrapair correlations for corneal curvature also indicate a strong role for the cornea in the development of refractive error. 

In the present study, we noted a significant difference in variance between the sexes for refractive error and ocular biometrics. Support for a sex effect on refractive error and other ocular biometrics has also come from two previous genetic studies. ²⁵ In the first, Teikari et al. ²⁶ examined 109 twin pairs (53 MZ and 61 DZ) and found that the heritability of refractive error (myopia) was 0.74 in males and 0.61 in females. A second study by Lyhne et al. ¹⁵ assessed a total of 114 twins pairs (53 MZ, 61 DZ) and also found that refraction, AL, radius of CC, and ACD were all sex dependent (P < 0.01). In the present study, we found a higher heritability estimate in the men for SE compared with the women, corresponding to previous studies. ²⁶ In addition, we were able to confirm the sex effect for ocular biometrics (Al and ACD) reported by Lyhne et al. ¹⁵ In the present study, the sex limitation ADE model was therefore found to be the best-fitting model for calculating heritability estimates for refractive error and also for all ocular biometric measures. 

The heritability estimates for refraction and its determiners (ocular biometrics) suggest that most of the variance in refractive error is due to genetic rather than environmental risk factors. Nevertheless, as previously shown by epidemiologic studies, ²⁷ we found that higher education levels was significantly associated with a more negative refraction. Furthermore, increasing age was associated with a more positive refraction. 

Although this study was sampled from a large volunteer cohort of twins of both sexes over various ages from a twin registry, there are still several limitations to the study. With all registry-based twin studies, ascertainment bias is a problem, as twin registries generally represent a selective sample of twins that are interested in scientific research. As a result, twin pairs are typically more motivated to participate in the study (self-selection) because they have the disease of interest (concordance-dependent bias). In addition, most of the twin pairs examined in this study were of white background and were all English-speaking (Anglo Celtic), thus limiting the analysis of ethnic differences in refractive error. 

In conclusion, this twin study provides evidence to support a major genetic role in the development of myopia, hypermetropia, CA, and ocular biometrics. Both additive and dominant genetic effects explained most of the variance in refractive errors and ocular biometrics, suggesting the likely involvement of several genes in the development of refractive error. This study also suggests a role for a dominant gene effect in CA. Variance component modeling indicated that the sex-limitation ADE model was the most parsimonious one for this twin sample and that this most likely holds true for all twin studies in which there is a large enough sample size of twins of both sexes. Finally, DZ twins from this cohort can be used for future genetic linkage studies to assist in identifying the genes responsible for the development of refractive error, particularly myopia. 

 Supported by the Australian Government Cooperative Research Centre Program.

 Submitted for publication March 13, 2006; revised May 24, June 21, and July 9, 2006; accepted August 24, 2006.

 Disclosure: M. Dirani, None; M. Chamberlain, None; S.N. Shekar, None; A.F.M. Islam, None; P. Garoufalis, None; C.Y. Chen, None; R.H. Guymer, None; P.N. Baird, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Baseline Characteristics of Twin Pairs by Zygosity

	MZ	DZ	P
Twin Pairs (n)	345 (56%)	267 (44%)	—
Age (y)	52.11 ± 15.85	52.63 ± 14.96	0.56
Gender (female/total)	467/690 (68%)	356/534 (67%)	<0.05
SE (DS)	0.051 ± 2.17	−0.015 ± 2.12	0.60
AL (mm)	23.40 ± 1.12	23.45 ± 1.07	0.52
ACD (mm)	3.45 ± 0.41	3.49 ± 0.45	0.14
CA (D)	0.745 ± 0.738	0.819 ± 0.792	0.13

Figure 1.

 

Intrapair correlations for SE refractive error and ocular biometrics in twin pairs.

Table 2.

 

Correlations for Spherical Equivalent and Ocular Biometrics by Gender for Each Twin Zygosity Group

Zygosity	Gender	n (Pairs)	SE (DS)	AL (mm)	ACD (mm)	CA (D)
MZ	(F/F)	234	0.82 (0.76–0.86)	0.88 (0.84–0.92)	0.62 (0.48–0.72)	0.48 (0.33–0.60)
	(M/M)	111	0.92 (0.88–0.95)	0.94 (0.91–0.96)	0.46 (0.25–0.57)	0.44 (0.17–0.64)
DZ	(F/F)	132	0.30 (0.10–0.48)	0.43 (0.24–0.59)	0.26 (0.11–0.42)	0.14 (0.09–0.20)
	(M/M)	43	0.10 (0.05–0.32)	0.34 (0.07–0.56)	0.37 (0.24–0.56)	0.20 (0.11–0.27)
	(F/M)	46	0.34 (0.02–0.52)	0.46 (0.10–0.72)	0.29 (0.11–0.38)	0.08 (−0.31–0.44)
	(M/F)	46	0.39 (0.10–0.62)	0.29 (0.03–0.56)	0.34 (0.21–0.44)	0.21 (−0.14–0.52)

Table 3.

 

Results of Sex Limitation ADE Model Fitting

Variable	Model	Log-Likelihood	df	Ch. Fit	Cd. df	P
SE	Sex lim. ADE*	4564.50	1141
	ADE	4578.87	1144	14.36	3	<0.001
	AE	4580.90	1145	2.02	1	0.16
	E	4936.34	1146	355.50	1	<0.001
AL	Sex lim. ADE*	1766.38	714
	ADE	1776.134	717	9.75	3	0.02
	AE	2106.66	718	330.52	1	<0.001
	E	2114.71	719	8.05	1	0.004
ACD	Sex lim. ADE*	712.75	729
	ADE	772.16	732	59.41	3	<0.001
	AE	846.60	733	74.45	1	<0.001
	E	866.28	734	19.68	1	<0.001
CA	Sex lim. ADE*	1361.11	635
	ADE	1390.86	638	29.75	3	<0.001
	AE	1417.89	639	27.03	1	<0.001
	E	1417.89	640

 A, additive genetics; D, dominant genetics; E, unique environment; df, degrees of freedom; ch. fit, chi square; cd. df, difference in degrees of freedom; sex lim. sex limitation.

*  Best-fit model.

Table 4.

 

Heritability Estimates for Refraction and Ocular Biometrics between Genders

Variable	Gender	A	D	E	h2 (CI)
SE	M	0.58	0.30	0.12	0.88 ± 0.02
	F	0.47	0.28	0.25	0.75 ± 0.03
AL	M	0.16	0.78	0.06	0.94 ± 0.05
	F	0.40	0.52	0.08	0.92 ± 0.04
ACD	M	0.15	0.36	0.49	0.51 ± 0.04
	F	0.42	0.36	0.22	0.78 ± 0.06
CA	M	0.22	0.28	0.50	0.50 ± 0.08
	F	0.13	0.47	0.40	0.60 ± 0.07

 h², heritability.

The authors thank the ATR for acting as the main referral source for twin recruitment.