To examine the relationship of birth weight with ocular measures in a Caucasian twin population.
Cross-sectional study of 1498 twins (308 monozygotic and 441 dizygotic pairs) aged between 5 to 80 years participating in the Australian Twins Eye Study.
All participants underwent ophthalmic examination including bilateral cycloplegic autorefraction, keratometry, interpupillary distance (IPD), central corneal thickness, intraocular pressure (IOP), and retinal photography. Birth weight and gestation were obtained from a self-administered questionnaire. A subset of the twins also participated in the Tasmanian Infant Health Study (288) and the Childhood Blood Pressure Study (184), which collected data on birth parameters allowing for verification of data. Linear mixed models were used for the main analysis.
Both the within-pair (β w 0.27, 95% confidence interval [CI] 0.15, 0.38 mm per kg increase in birth weight, P < .001) and between-pair associations (β B 0.22, 95% CI 0.08, 0.35, P = .002) of birth weight with axial length were significant and of similar magnitude (difference in effect, P = .56), after adjusting for relevant confounders. In contrast, birth weight was negatively associated with corneal curvature (β w −0.82, 95% CI −1.09, −0.55 diopters per kg increase; β B −0.69, 95% CI −0.98, −0.41, both P < .001). These associations remained significant within dizygotic and monozygotic pairs. Refraction, anterior chamber depth, IPD, IOP, and optic disc parameters are unrelated to birth weight.
Consistent with previous studies in singleton children, lower birth weight is associated with shorter axial length and more curved corneas in this twin study. This also adds new insights into the emmetropization process.
Fetal origins of adult disease has been an area of active research in the past 2 decades. Many epidemiologic studies have shown that lower birth weight, shorter birth length, and smaller head circumference, regarded as markers for impaired intrauterine development, are associated with increased risk or earlier onset of cardiovascular diseases, hypertension, and diabetes. Furthermore, new research has produced preliminary evidence that intrauterine environment may also have long-lasting impact on the development of ocular dimensions and retinal microcirculation. For example, recent population-based studies reported that smaller birth size such as low birth weight was associated with ocular traits such as shorter axial length, narrower vitreous chamber, more curved corneas, larger cup-to-disc ratio, and narrower retinal arteriolar caliber. However, birth size seems to have little long-term effect on refractive error such as myopia. Importantly, these investigations suggest that adverse intrauterine growth may confer increased future risk of diseases specifically affecting the eye (eg, glaucomatous optic neuropathy) as well as other parts of the body.
To date, limited population-based studies have explored the relationship between birth parameters and ocular measures. Moreover, it is still unknown whether any potential associations between birth parameters and ocular measures are confounded by shared genes or shared environmental factors (eg, maternal nutrition) that affect both birth size and the development of ocular parameters. Within-pair analysis in a twin population offers an opportunity to disentangle these effects, and we are unaware of any previous study examining this.
The purpose of this report is therefore to investigate the relationship between birth weight and a range of ocular measures (ocular biometry, refraction, and glaucoma-related endophenotypes) in a large Caucasian twin population.
The Australian Twins Eye Study (ATES), involving 2235 twins and nontwin siblings, was designed to investigate the relative influence of genetic and environmental factors on a variety of ocular traits related to glaucoma. The study design and details of sample recruitment are described elsewhere. In brief, the study population comprising predominantly Caucasian twins were ascertained from the Tasmanian Infant Health Study (TIHS) cohort, the Brisbane Adolescent Twin Study, and the Australia Twin Registry; through media appeal; and from a school-recruitment approach during the period 2000 to 2008. All twins and their nontwin siblings in the ATES answered a standardized questionnaire providing details of sociodemographic and medical information and underwent a thorough ophthalmic examination.
A total of 288 children participating in the ATES were recruited from the TIHS, which examined sudden infant death syndrome in infants during the years 1988 to 1995. Among these children, 184 multiplets (born between 1991 and 1993) also participated in the Childhood Blood Pressure Study that examined cardiovascular diseases in 1999.
Of the total population of 2235 twin individuals, we excluded 269 nontwin siblings, triplets, and twins with missing zygosity data; 349 persons who had no birth weight information; and an additional 119 single twins, leaving 1498 twin individuals with known zygosity and birth weight, comprising 308 monozygotic (MZ) and 441 dizygotic (DZ) twin pairs for this analysis.
Birth Parameters Assessment
We collected information on birth weight, birth order, and gestation duration for all participants in the ATES from a detailed self-administered questionnaire. Two hundred eighty-eight of the twins also had data relating to birth weight, birth length, head circumference, and gestational age extracted from medical birth records as part of their involvement in the TIHS, details of which have been published elsewhere. Briefly, last menstrual period was used to estimate gestational duration. These data provided a means by which the self-reported birth parameters in the ATES could be validated.
Reliability assessment of 283 individuals with both birth weight and gestational age data in the ATES and the TIHS showed very high agreement between the 2 studies, with an intraclass correlation coefficient of 0.978 (95% confidence interval [CI], 0.972–0.982) for birth weight, and 0.992 (0.990-0.994) for gestational age. Low birth weight was defined as <2500 g. Prematurity was defined as <37 weeks gestation duration.
Measurement of Optic Disc Parameters
All twins and nontwin siblings had 10-degree stereoscopic optic disc–centered photographs taken (Nidek fundus camera 3-Dx/F; Nidek, Gamagori, Japan) after pupil dilation with tropicamide 1% or cyclopentolate 1% (for children). All fundus photographs were digitalized at high resolution (2102 × 1435 pixels, 2900 dpi, 36-bit color) using a Nikon CoolScan scanner (Nikon Corp, Tokyo, Japan).
Assessments of optic disc parameters, including optic disc, cup, and rim area, were performed by 2 trained graders using a standardized grading program. Re-measurement of 50 randomly selected retinal photographs 3 months apart showed high intragrader reproducibility, with intraclass correlation coefficient (95% CI) of 0.94 (0.87-0.98) for optic disc area. The intergrader reliability was assessed in 73 randomly selected retinal images, and interclass correlation coefficient (95% CI) was 0.75 (0.67–0.86) for optic disc area.
Blood or mouth swab samples were collected for DNA extraction in the ATES. Twin pair zygosity was confirmed by up to 12 highly polymorphic short tandem repeats (STR), with an accuracy of more than 99%.
Measurement of Ocular Biometry and Other Variables
Participants underwent a comprehensive ophthalmic examination. Postcycloplegic interpupillary distance (IPD) measure, central keratometry, and refractive errors for both eyes were measured using a Humphery-598 automatic refractor (Carl Zeiss Meditec, Inc, Miami, Florida, USA) after pupil dilation. To assess refraction status, spherical equivalents were calculated using the standard formula of the algebraic sum of the value of the sphere and half the cylinder value (sphere + 0.5 cylinder). After topical anesthesia, bilateral intraocular pressure (IOP) was measured for each participant using the TONO-PEN XL (Reichert Ophthalmic Instruments, Depew, New York, USA). Measurements of central corneal thickness were obtained from the average reading of the central cornea using the Pachymeter Tomey SP 2000 (Tomey Corp, Nagoya, Japan) or Pachmate DGH 55 (DGH Technology, Inc, Exton, Pennsylvania, USA). The intraocular lens (IOL) Master (Carl Zeiss, Oberkochen, Germany) was used to obtain ocular biometry (axial length, anterior chamber depth, and horizontal and vertical corneal curvature).
Data for current height and weight in the ATES were collected from a self-administered questionnaire. For younger twins with no height and weight data, we used 184 readings available in the concurrent child blood pressure study, which measured child height in centimeters using a wall-mounted measuring tape and weight in kilograms using a digital scale (SECA, model 782 2321009; Vogel & Halke, Hamburg, Germany). Body mass index (BMI) was calculated as kg/m 2 .
A range of ocular measures including axial length, anterior chamber depth, corneal curvature, IPD measure (postcycloplegic), and glaucoma-related endophenotypes such as central corneal thickness, IOP, optic disc area, optic cup area, and optic cup-to-disc ratio were outcomes of interest and were all analyzed as continuous variables. Refraction was measured as spherical equivalent and was analyzed both as a continuous and a categorical trait (1 for spherical equivalent ≥ median and 0 for those < median). Spearman correlations between eye outcomes were obtained with Bonferonni adjustment for multiple comparisons.
Standard linear regression was performed with “twins as individuals,” initially controlling for age at examination and gender. Logistic regression was performed for categorical outcome such as refractive error. Multivariable linear regression models were then constructed for those ocular outcomes with a significant association in the age-gender-adjusted linear models ( P < .05) (for more details, see Supplemental Material at AJO.com ).
The main form of analysis was then conducted by using linear mixed regression models, treating each twin pair as a cluster to account for the correlated nature of the data. This model provided estimates of both within-pair (β w ) and between-pair (β B ) associations. Stratified analysis by zygosity was also performed using linear mixed models. Covariates included in the model were based on statistical and biological consideration of confounding.
We also investigated the associations of birth weight with ocular biometry in preterm twins using multivariable linear regression, accounting for family structure. Finally, we investigated the potential interaction of risk factors with gender. Where interactions were statistically significant ( P < .01), stratified analyses were performed.
All analyses were performed for the right eye (except for IOP, where the average of right and left eye measurements was used because of variation of the measurement) using STATA 11.0 (Stata Corp, College Station, Texas, USA). (See the online Supplemental Material at AJO.com for more details of the statistical analysis.)
Selected characteristics including demographic information, birth parameters, and anthropometric measures of the study sample stratified by zygosity are shown in Table 1 . The median age of the whole study sample was 17 years (range, 5-80 years). MZ twins (n = 616; 308 pairs) were more likely to be female and older, and a higher proportion were of low birth weight, small for gestational age, and premature than DZ twins (n = 882; 441 pairs). MZ and DZ twins had the same median birth length of 47 cm, although MZ twins had a slightly greater but significantly different range (33–51 weeks and 38–53 weeks, respectively, P = .03). Similarly, the median head circumference of 33 cm was the same in both twin types, but the range was significantly larger in DZ twins (27–36 cm) compared to MZ twins (23–37 cm) ( P = .002).
|Characteristic||MZ Twins (n = 616; 308 Pairs)||DZ Twins (n = 882; 441 Pairs)||P Value c|
|Age, years a||19.0||17.0||<.001|
|Male gender, %||31.2||46.4||<.001|
|Birth weight, kg||2.4||2.5||<.001|
|Low birth weight, %||54.4||41.7||<.001|
|Small for gestational age, %||18.6||13.8||.16|
|Birth length, cm a , b||47.0||47.0||.03|
|Head circumference, cm a , b||33.0||33.0||.002|
|Body mass index, kg/m 2||22.8||21.8||.57|
c P < .05, represents statistically significant difference in means or proportions, adjusted for age and gender, except for age and male gender, and rank sum test for birth length and head circumference.
Table 2 shows the correlations among various ocular measures. All the optic disc parameters including disc area, cup area, and cup-to-disc area ratio were significantly correlated ( P < .001). Axial length and horizontal corneal curvature were among those most highly correlated outcomes (r = −0.57, P < .001). There were moderately strong correlations between axial length and anterior chamber depth, IPD measure, anterior chamber depth and refractive error, horizontal corneal curvature and refractive error (r > 0.3, P < .001). However, two other correlations including those between axial length and refractive error and that between axial length and central corneal thickness were weak and nonsignificant.
|Axial Length (mm)||Anterior Chamber Depth (mm)||Corneal Curvature (diopters)||Central Corneal Thickness (μm)||Refractive Error (diopters)||Mean Intraocular Pressure (mm Hg)||Optic Disc Area (mm 2 )||Optic Cup Area (mm 2 )||Area Cup-to-Disc Ratio||IPD (mm)|
|Axial length (mm)||1.00|
|Anterior chamber depth (mm)||.41 a||1.00|
|Corneal curvature (diopters)||−0.57 a||.07||1.00|
|Central corneal thickness (μm)||.06||−.02||−0.04||1.00|
|Refractive error (diopters)||−0.16 b||−0.31 a||−0.43 a||−0.09||1.00|
|Mean intraocular pressure (mm Hg)||0||0||.04||.11||−0.05||1.00|
|Optic disc area (mm 2 )||.03||−0.18 a||−0.25 a||.03||.16 b||−0.09||1.00|
|Optic cup area (mm 2 )||.12||−0.08||−0.20 a||0||.09||.03||.61 a||1.00|
|Area cup-to-disc ratio||.13||−0.03||−0.14 b||0||.05||.07||.37 a||.95 a||1.00|
|IPD measure (mm)||.30 a||.04||−0.30 a||−0.04||.07||.05||.10||.13 b||.14 b||1.00|
Axial length, corneal curvature (horizontal), IPD, central corneal thickness, mean IOP, and optic disc and cup area were approximately normally distributed in the study population, with the mean (SD) being 23.20 (0.88) mm, 43.68 (1.54) diopters (D), 60.55 (3.94) mm, 544.29 (35.09) μm, 15.95 (2.98) mm Hg, 2.06 (0.43) mm 2 , and 0.44 (0.31) mm 2 , respectively. The distribution of refractive error was peaked (Kurtosis 18.79) and skewed to the left (skewness −2.22) and the median in the right eye was 0 D (range, −16.50–6.38). The distribution of anterior chamber depth was peaked (Kurtosis 9.49) and slightly skewed to the left (skewness −1.65) and the median in the right eye was 3.65 mm (range, 1.65–4.46).
Using standard linear regression with each twin treated as an individual, Table 3 shows the associations between birth weight and ocular measures after controlling for age and gender. Birth weight was significantly associated with both axial length and corneal curvature (both horizontal and vertical) in this model. However, no significant relationship between birth weight and central corneal thickness was observed, nor between birth weight and other ocular measures including IPD, mean IOP, optic cup area, or optic cup-to-disc area ratio. Birth weight was only marginally significantly associated with optic disc area ( P < .03), but this association did not persist in further adjustment for other confounders. Neither anterior chamber depth nor refractive error was related to birth weight ( P = .49 and P = .18, respectively, data not shown).
|Ocular Measures a||Birth Weight, kg (n)|
|<2.5 (703)||2.5–2.9 (506)||3.0–3.4 (239)||3.5–3.7 (36)||≥3.8 (14)||P Value b|
|Axial length (mm)||23.13 (0.03)||23.19 (0.04)||23.36 (0.06)||23.38 (0.15)||23.19 (0.25)||.002|
|Corneal curvature (diopters)||43.87 (0.06)||43.69 (0.07)||43.16 (0.10)||42.90 (0.25)||42.86 (0.41)||<.001|
|Central corneal thickness (μm)||544.82 (1.41)||544.39 (1.66)||543.80 (2.39)||536.12 (6.03)||546.74 (10.60)||.42|
|IPD measure (mm)||60.24 (0.16)||60.51 (0.19)||61.05 (0.27)||61.45 (0.74)||62.01 (1.20)||.002|
|Mean IOP (mm Hg)||16.00 (0.11)||15.93 (0.13)||15.82 (0.19)||16.68 (0.50)||15.11 (0.83)||.66|
|Optic disc area (mm 2 )||2.05 (0.02)||2.05 (0.02)||2.10 (0.03)||2.23 (0.08)||2.15 (0.13)||.03|
|Optic cup area (mm 2 )||0.43 (0.01)||0.44 (0.02)||0.45 (0.02)||0.51 (0.06)||0.59 (0.09)||.16|
|Area cup-to-disc ratio||0.20 (0.01)||0.20 (0.01)||0.20 (0.02)||0.22 (0.02)||0.27 (0.04)||.44|
Table 4 presents the results from the linear mixed regression models for the associations between birth weight and axial length fitted for the whole sample of twins (MZ + DZ) and separately by zygosity (MZ or DZ). Both the within-pair (β w 0.27, 95% CI 0.15, 0.38) and between-pair (β B 0.22, 95% CI 0.08, 0.35) associations of birth weight with axial length were significant and of a similar magnitude, after adjusting for age, gender, spherical equivalent, and gestational age. These associations remained significant even after adjustment for current height (β w 0.28, 95% CI 0.15, 0.28; β B 0.20, 95% CI 0.05, 0.34) given current height may be associated with axial length. This implies that the association between lower birth weight and shorter axial length was not merely attributable to people with lower birth weight growing up to be shorter in stature.