Vision screening at two years does not reduce the prevalence of reduced vision at four and a half years of age

Abstract

Background

There is currently insufficient evidence to recommend vision screening for children < 36 months of age. This study assessed the effect of comprehensive vision screening, as well as the sensitivity of age-appropriate vision tests, at two years of age on habitual visual acuity at 4.5 years of age.

Methods

Children born at risk of neonatal hypoglycaemia (n = 477) underwent vision assessment at 54 ± 2 months of age including measurement of monocular and binocular habitual visual acuity, assessment of binocularity and stereopsis. Of these children, 355 (74.4 per cent) had also received vision screening at two years of age (mean age = 24 ± 1 months), while 122 were not screened.

Results

Eighty (16.8 per cent) children were classified as having reduced vision at 4.5 years of age, but the prevalence of reduced vision did not differ between children who had previously been screened at two years of age and those who had not (15.5 per cent vs 20.5 per cent, p = 0.153). However, children with reduced vision at 4.5 years of age were more like to have had visual abnormalities requiring referral detected at two years of age (p = 0.02). Visual acuity and mean spherical equivalent autorefraction measurements were also worse (higher values) in two-year-old children who were later classified with reduced habitual visual acuity (p = 0.031 and p = 0.001 respectively). Nevertheless, unaided binocular visual acuity, non-cycloplegic refractive error, and stereopsis at two years all showed poor sensitivity and specificity for predicting visual outcomes at 4.5 years of age.

Conclusion

Our findings do not support the adoption of early vision screening in children as current vision tests suitable for use with two-year-old children have poor sensitivity for predicting mild-moderate habitual vision impairment at 4.5 years of age.

Introduction

Children’s vision screening programmes are the primary method for detecting amblyopia, a neurodevelopmental disorder affecting 1–3 per cent of the population,^1–3 typically caused by anisometropia or strabismus that can lead to permanent reduction in acuity and loss of binocular vision. There is clear evidence that the prevalence of amblyopia can be reduced by early detection and treatment^4–6 and international guidelines recommend that children are screened for reduced habitual visual acuity (habitual visual acuity) at 3–6 years of age.⁷ While amblyopia remains treatable up until teenage years,⁸ intervention is more effective for younger patients.⁴ Although vision screening early in life is important to detect amblyopia and its risk factors, there is currently insufficient evidence to recommend screening infants and toddlers younger than 36 months of age.⁹

The current literature on vision screening in children < 36 months of age is largely retrospective and observational.^10,11 The small number of prospective studies provide contradictory results on the benefits of early vision screening, with some programmes showing positive outcomes¹² and others no difference in visual outcomes between screened and unscreened populations.¹³ One possible explanation for the contradictory outcomes between studies may be the different age-appropriate testing protocols utilised in children < 36 months of age. Testing monocular recognition acuity in young children is challenging and only successful in 39 per cent of 30–36 month olds.¹⁴ Therefore, age-appropriate behaviour-based tests to assess resolution acuity in pre-verbal children are commonly used.^15–17

The Cardiff Acuity Test is a vanishing optotype preferential looking task that can maintain a toddler’s attention better than conventional grating targets with testability approaching 100 per cent.^15,16 Nevertheless, vanishing optotypes may be more resistant to optical defocus than recognition acuity tasks and therefore may have reduced sensitivity for detecting refractive errors.^18,19 Consequently, automated preschool vision screening tests, such as autorefraction or the Pediatric Vision Screener²⁰ have been suggested as better screening methods for detecting amblyopia and/or amblyopia risk factors in 12-30 month old children.²¹

The aim of this study was to evaluate the efficacy of a comprehensive vision screening, using a battery of age-appropriate vision tests, at 2 years on visual outcomes at 4.5 years of age.

METHODS

Participants

Participants were part of the Children with Hypoglycaemia and their Later Development (CHYLD) study, a prospective cohort of 614 infants born at risk of neonatal hypoglycaemia.^22,23 Full details of the study design have been reported elsewhere.^22,24 In brief, infants were enrolled before or shortly after birth and underwent regular monitoring of blood glucose concentrations for 24-48 hours, or until there were no ongoing clinical concerns. Follow-up neurodevelopmental assessments were conducted at 24 ± 1 and 54 ± 2 months of age.

The neonatal study and two follow-up studies were approved by the Northern Y Health and Disability Ethics Committee and complied with the tenets of the Declaration of Helsinki. Written informed consent was obtained from a parent or guardian of each child at study entry and at each follow-up visit.

Primary visual outcome measure

Eye care professionals with previous paediatric training conducted examinations at both two and 4.5 years of age. The primary visual outcome measure was the prevalence of reduced vision, defined as reduced habitual visual acuity (habitual visual acuity) of 0.3 logMAR or worse; a difference of more than 0.1 logMAR between the eyes; heterotropia in primary position or ocular pathology at 4.5 years of age (Table 1). Children who did not have any of these findings were classified as visually normal. Children with incomplete or missing data that prevented assignment to either group remained unclassified and were removed from further analysis.

Table 1.

Classification rules used to define children with “normal” or “reduced” vision at 4.5 years of age. Children who failed to achieve visual outcomes in the “normal” range for each of these visual outcomes were classified into the “reduced vision” group.

	Visually normal at 4.5 years of age	Reduced vision at 4.5 years of age
habitual visual acuity	Better than 0.3 logMAR	Worse than or equal to 0.3 logMAR
Interocular difference in habitual visual acuity	More than 0.10 logMAR difference in habitual visual acuity between the eyes	More than 0.15 logMAR difference in habitual visual acuity between the eyes
Ocular misalignment	None	Present
Ocular pathology	None	Present

Visual assessments

Children were assessed in clinical rooms at local hospitals or medical centres, a dedicated vision testing room at the University of Auckland, or in the home of the child if requested by the family. All vision testing rooms met the following standards: ≥ 3 m long, room illuminance of ≥ 400 lux and minimum noise and distractions within or outside the room.

Children were assessed using age-appropriate measures of visual acuity, ocular alignment, autorefraction (at two years of age), stereopsis, and ocular health (Table 2).

Table 2.

Visual assessments and testability definitions

Test/Instrument	Testing Distance	Stopping Rules	Untestable definition
Cardiff Acuity Test	1 m, or 50 cm if child uncooperative	2 cards incorrect at a given optotype size	Unable to observe preferential looking response for largest card at 50 cm/child uncooperative
Single crowded Lea symbols	3 m	2 optotypes incorrect at a given optotype size	Unable to complete the pre-test
Hirschberg Reflex	~ 67 cm		Child uncooperative
Cover-uncover test	40 cm/3 m (4.5 years only)	≥ 3 cover-uncover strokes	Child uncooperative and did not fixate target
Suresight Vision Screener	~ 37 cm	> 6 reliability score or better of 2 measures	Unable to complete both monocular assessments, regardless of reliability score
Lang stereotest	40 cm	Incorrect response	Unable to complete the pre-test
VAC Stereofly test	40 cm	2 successive incorrect responses	Unable to demonstrate appreciation of gross stereoacuity on fly test
Gross ocular health inspection/red reflex test	~ 67 cm	1 measurement	Unable to complete both monocular assessments

Visual assessment at two years of age

Habitual visual acuity was assessed using the Cardiff Acuity Test and preferential looking behaviour.^15,16 Binocular habitual visual acuity was measured before monocular measurements were attempted. When a child refused monocular vision measurements, equal objection to occlusion of the fellow eye was checked to screen for amblyopia or other causes of unequal vision.

The Cardiff Acuity Test cards were presented at the eye level of the child and the examiner judged the eye movement of the child to ascertain whether the child looked at the picture presented on the card. Presentation began at 1 m with the largest optotype and was moved to 50 cm if the child showed a lack of interest or no preferential looking response. habitual visual acuity (in logMAR) was recorded as the smallest optotype size for which a preferential looking response was seen in at least two of three cards of the same size in the series.

The Hirschberg Corneal Reflex test was performed on all two-year-old children. When children were cooperative, eye alignment was also assessed with a cover-uncover test performed with the child looking at a small toy held at 40 cm.

The Suresight™ handheld autorefractor, set to the child calibration mode, was used without cycloplegia to provide an estimate of refractive error. If the reliability index was < 6, the measurement was repeated. If the second measurement also had a reliability index of < 6, then the measurement with the highest reliability was recorded.

The Lang Stereotest (cards I and II) was measured at 40 cm and stereoacuity was recorded as the smallest disparity the child could identify on the cards. Health inspection for gross abnormalities was conducted using a direct ophthalmoscope using an additional +10 D ophthalmoscope lens if required.

Children were referred to a paediatric ophthalmologist if any incidental findings were detected: abnormal pupil reflexes, suspected ocular pathology, strabismus, nystagmus, extraocular muscle restriction or significant refractive error (anisometropia ≥ 1.00 D in any meridian, astigmatism ≥ 1.50 DC, myopia worse than −2.00 DS, hyperopia ≥ +3.50 DS).

Visual assessment at 4.5 years of age

Monocular and binocular habitual visual acuity were measured with a single crowded Lea Symbols chart (Good-lite #259200)^7,25 and recorded in logMAR notation as the smallest correctly identified optotype using an optotype-by-optotype scoring system. The cover-uncover test was performed at both 3 m and 40 cm with the child fixating a single letter one size above threshold habitual visual acuity. Stereoacuity was measured with the VAC Fly Stereotest containing Lea symbols (Stereo Optical Co., Chicago Illinois) using the graded circles (400 to 20″ of arc).

Data analysis

All statistical analyses were performed using Graphpad Prism (Graphpad Software Inc., La Jolla, CA, USA). Data are presented as mean (standard deviation), median (interquartile range), or number (percentage). Distributions were assessed for normality using the Shapiro-Wilk normality test. The association between demographic factors and screening was assessed using the Chi-squared test. Comparisons of sociodemographic characteristics (NZDep Index²⁶) between screened and unscreened populations were assessed with the Mann Whitney U test for nonparametric distributions.

For each visual assessment, testability rates were assessed based on the number of children who successfully completed the measurement relative to the total number of children tested. Comparisons of habitual visual acuity, autorefraction (in power vector form (M, J₀ and J₄₅)²⁷), and stereoacuity between the normal and reduced vision groups were assessed with the Mann Whitney U test for nonparametric distributions. The incidence of visual abnormalities at two years of age and the outcomes of children referred at two years of age were assessed using Fisher’s exact test.

The proportion of children with habitual visual acuity better than 0.3 logMAR and no evidence of ocular misalignment or pathology at 4.5 years of age were used to estimate specificity, while sensitivity was calculated as the proportion of children with reduced vision at 4.5 years of age who failed a screening test at two years of age.

Receiver operating characteristic curves were calculated to measure the predictive value of different levels of habitual visual acuity, autorefraction, and stereoacuity cut-off values at two years of age on visual outcome at 4.5 years of age. Specificity was set at a minimum of 90 per cent, and failure criteria were selected to maximise the overall sensitivity for detecting reduced habitual visual acuity at 4.5 years of age.

RESULTS

Study population

Six-hundred-fourteen children were recruited (Figure 1). A total of 477 participants received a comprehensive vision screening examination at 4.5 years of age (mean age = 54 ± 2 months). Of these children, 355 (74.4 per cent) had also received a vision screening examination at two years of age (mean age = 24 ± 1 months), while 122 (25.6 per cent) were not screened (Figure 1). The population of children screened at two years of age showed a similar distribution of gender, ethnicity, and socio-economic deprivation scores compared to children who were not screened at this age (Table 3).

Figure 1.

Study cohort

Table 3.

Demographics of the study cohort. Data are number (per cent), or median (interquartile range).

Characteristic	Cohort (n = 477)	Screened at 2 years (n=355)	Unscreened at 2 years (n=122)	p
Males	249 (52.2)	185 (52.1)	64 (52.5)	0.947
Ethnicity:				0.281
New Zealand	256 (53.7)	190 (53.5)	60(49.2)
Maori	181 (37.9)	124 (34.9)	47 (38.5)
Pacific Islander	18 (3.8)	13 (3.7)	1 (0.8)
Other	22 (4.6)	28 (7.9)	14 (11.5)
Deprivation index	7 (4-9)	6 (3-9)	6 (3-8)	0.880

Children not assessed at 4.5 years showed a similar distribution of gender (p = 0.766) and ethnicity (p = 0.066), but resided in more deprived areas of New Zealand compared to children who were assessed at 4.5 years of age (NZ Dep: 4.5 years screened = 7 (4-9); unscreened = 8 (6.5-10); p = 0.001).

Vision testing success rates

The testability of habitual visual acuity ranged from 87.5 per cent at two years of age for binocular measurements, to 97.6 per cent at 4.5 years of age for monocular measurements (Figure 2).

Figure 2.

Percentage of children successfully tested using vision screening tests administered at two years and 4.5 years of age. Binocular habitual visual acuity (habitual visual acuity) tested with the Cardiff Acuity Tests at two years of age and monocular habitual visual acuity with Lea Symbols at 4.5 years; stereoacuity tested with Lang stereotest at two years of age and VAC Stereofly at 4.5 years of age. Autorefraction was only assessed at two years of age.

Visual outcomes at 4.5 years

At 4.5 years of age, 75.5 per cent (360) children were classified as visually normal, 16.8 per cent (80) with reduced vision, and 7.8 per cent (37) of children were unable to complete acuity testing and were therefore unclassified (as described in Table 1). Children were classified with reduced vision at 4.5 years of age due to unilateral reduced habitual visual acuity (72.5 per cent), reduced habitual visual acuity in both eyes (37.5 per cent), strabismus (10.0 per cent), and/or abnormal ocular pathology (2.5 per cent). There was no difference in the prevalence of reduced vision at 4.5 years of age between those children screened at two years of age and those not screened (15.5 per cent vs 20.5 per cent, p = 0.153), suggesting early vision screening was not able to accurately identify children who presented with vision problems at 4.5 years of age.

Predictive value of visual acuity screening tests used at two years of age

Children classified with reduced vision at 4.5 years of age had significantly worse habitual visual acuity at two years of age than children later classified as visually normal (Table 4: p = 0.031). The absolute mean sphere equivalent (mean spherical equivalent) autorefraction measurements were also higher at two years of age in the reduced vision group compared to those with normal visual outcomes at 4.5 years (p = 0.001). J₀ and J₄₅ measurements were not different between the two ages (p > 0.05). No differences in stereoacuity were observed at two years of age between the children classified with normal or reduced vision at 4.5 years (p > 0.05). There was a higher prevalence of strabismus observed at two years of age in children later classified with reduced vision (9.3 per cent [five children]) compared to those with normal visual outcomes (1.5 per cent [four children]).

Table 4.

Vision assessments at two years of age comparing children who were classified with normal or reduced vision at 4.5 years of age. Data are number (per cent), or median (interquartile range).

Screening measure at 2 years	Cohort (n = 477)	Visually normal at 4.5 years (n = 360)	Reduced vision at 4.5 years (n = 80)	p
Binocular habitual visual acuity (logMAR)	0.0 (−0.1–0.1)	0.0 (−0.1–0.1)	0.1 (0.0–0.2)	0.031
RE Autorefraction (D)
mean spherical equivalent	1.1 (0.8–1.5)	1.1 (0.8–1.4)	1.4 (1.3–2.0)	0.001
J0	0.0 (−0.2–0.3)	0.0 (−0.2–0.2)	0.1 (−0.2–0.4)	0.098
J45	0.0 (−0.1–0.1)	0.0 (−0.1–0.2)	0.1 (−0.0–0.2)	0.224
Stereoacuity (Lang) Seconds of arc	400 (200–400)	400 (200–400)	400 (200–400)	0.169
Strabismus	9 (2.5)	4 (1.5)	5 (9.6)	0.006
Ocular pathology	3 (0.9)	2 (0.7)	1 (1.9)	0.390

Habitual visual acuity and refractive error measurements at two years of age were poorly predictive of the equivalent measurements at 4.5 years of age (Table 5). The classification of children into reduced vision and normal groups at 4.5 years of age was best predicted by the mean spherical equivalent autorefraction measurements at two years of age (area under the ROC curve = 0.70; 95 per cent CI 0.58 – 0.82). Using mean spherical equivalent values > 2.00 D at two years as the cut-off criteria was the most efficient for detecting children with reduced vision and rejecting visually normal children at 4.5 years of age (sensitivity = 26.9 per cent and specificity = 93.3 per cent). Binocular habitual visual acuity measurements at two years of age were poorly predictive of visual outcome at 4.5 years of age (area under the ROC curve = 0.60; 95 per cent CI 0.51 – 0.70). A habitual visual acuity of worse than 0.3 logMAR was required to achieve greater than 90 per cent specificity, with only 16.28 per cent sensitivity.

Table 5.

Comparison between clinically relevant cut-off scores to predict poor visual outcome at 4.5 years of age. For each test, the cut-off value that achieved the highest sensitivity with at least 90 per cent specificity is shown in bold.

Vision tests at 2 years	Cut-off values	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Visual acuity logMar (Cardiff cards)	0.3	17.8	93.4	33.3	86.0
	0.2	33.3	82.7	26.3	87.0
	0.1	55.5	54.3	18.4	86.8
	0.0	84.4	27.6	17.8	90.5
Autorefraction mean spherical equivalent (D)	3.00	0.00	99.3	0.00	85.1
	2.00	26.9	94.6	46.7	88.2
	1.00	80.8	38.0	18.4	91.9
	0.50	92.3	12.0	15.4	90.0
Stereoacuity (seconds of arc)	300	66.0	41.9	18.6	86.0
	600	12.8	88.9	18.8	83.5
	900	6.4	97.9	37.5	83.9

Are vision problems at two years of age associated with reduced vision at 4.5 years of age?

Thirty-six children (10.1 per cent) seen at 4.5 years of age had been previously referred for further eye care at two years of age. Children were referred for suspected significant refractive error (10), strabismus (7), ptosis (8), or ocular discharge (7) or abnormal red reflex (3). The presence of any of these visual abnormalities at two years of age was associated with reduced vision at 4.5 years of age. Thirty-three percent of children referred for follow up care at two years of age were later classified with reduced vision at 4.5 years of age, compared to 16.3 per cent of those who passed the screening test at two years of age (Fisher’s exact test, p = 0.02). Information as to whether children and their parents attended referral visits or any treatment was provided to these children was unavailable.

DISCUSSION

Comprehensive vision screening at two years of age did not improve visual outcomes at 4.5 years as age-appropriate clinical tests used to measure visual function in two-year-old children had poor sensitivity and specificity for identifying children with reduced vision at 4.5 years. Although children with reduced vision at 4.5 years had worse binocular habitual visual acuity and higher refractive errors at two years of age than those classified as visually normal, the differences between groups were not clinically meaningful as they fell within test-retest variability of the measures (approximately 0.1 logMAR²⁸ and 0.25 D difference²⁹, respectively). Possible explanations of our results include: (i) the inability of vision tests suitable for use with toddlers to adequately assess all aspects of visual function, (ii) the difficulty in achieving monocular visual acuity measurements in two year olds, and/or (iii) the appearance of vision problems only later in childhood.

The poor agreement in visual function measures at each age may be partly due to different tests used at the two ages. The Cardiff Acuity Test is significantly less sensitive than recognition acuity tasks, such as the Lea Symbols, for detecting refractive errors.^18,19,30 Nevertheless, recognition acuity measures are difficult to achieve in children younger than 36 months³¹ and cooperation for monocular visual acuity measures is particularly challenging.³²

The inability to obtain vision measures from each eye separately impedes the diagnosis of monocular refractive error and amblyopia which is typically diagnosed based on interocular differences in acuity.³³ While Adoh and Woodhouse reported obtaining monocular visual acuity measures in 65-73 per cent of children at two years of age using the Cardiff Acuity Test,¹⁵ monocular measurements could only be obtained in 21 per cent of children. This may be due to the multidisciplinary nature of the CHYLD study where children underwent a lengthy developmental assessment including cognitive assessment, neurological evaluation, tests of executive function and global motion testing as well as vision screening.²²

Mean spherical equivalent measurements at two years were found to be the most sensitive for predicting vision outcomes at 4.5 years. Nevertheless, in this population of children autorefraction with the Suresight™ was also one of the least testable procedures (26.9 per cent of children testable) and may be impractical as a screening tool in two-year-old children. There was high testability for the Lang stereovision tests (85.2 per cent testable) and binocular habitual visual acuity measures with the Cardiff Acuity Test (87.5 per cent testable). However, monocular visual acuity measurements, which are required for amblyopia diagnosis, were more challenging (21.0 per cent testable).

Comparison with other studies

Few studies have evaluated vision screening programmes in children < 3 years of age. One prospective study from the Netherlands that showed screening for eye alignment, pupillary reflexes, and ocular motility at 3-9 months of age did not predict visual status at older ages¹³ while other studies have identified some benefit to vision screening in infants and toddlers. Struble et al. retrospectively investigated an eight-part vision screening tool for children enrolled in an early intervention programme due to risk of developmental delay.³⁴ The screening programme consisted of risk factor identification plus eyelid reflex, fixation, tracking, pupil response, corneal light reflex, cover/uncover test, and visual acuity (Lea of HOTV) assessments. The programme detected a high prevalence of vision disorders (39 per cent) with a high specificity and negative predictive value (95 per cent), while sensitivity was less (59.12 per cent) as was positive predictive value (57.25 per cent).

Similarly, the Avon Longitudinal Study of Parents and Children (ALSPAC) found that an intensive preschool vision screening programme where children were evaluated by an orthoptist, with autorefraction, eye alignment and visual acuity, at several different ages significantly reduced the prevalence of amblyopia.¹² At 7.5 years, amblyopia was found less commonly in the intensively screened group (1.45 per cent of children) than in the control group (2.66 per cent) who only received vision screening at 37 months of age.

Study limitations

Complete eye examinations³⁵ were not conducted at either 24 or 54 months of age. Children with reduced vision were categorised based upon habitual visual acuity worse than 0.3 logMAR, the presence of manifest ocular misalignment or suspected ocular pathology detected at 4.5 years of age. This habitual visual acuity cut-off was based on previous studies which have also used acuity worse than 0.3 logMAR to define visual impairment^36,37 and represents the single-sided 95 per cent percentile of VA in children at five years of age.³⁸ Nevertheless, without conducting a complete eye examination the cause of reduced habitual visual acuity could not be identified. Future prospective studies which investigate the causes of reduced vision in a large sample of children are required to address this issue. It is also unknown whether vision problems were present at two years or whether they developed between two and 4.5 years of age and future research should examine the prevalence of eye problems at both ages.

This study was conducted using a population of children at risk for neonatal hypoglycaemia, and may not be applicable to the general population. In this study, 16.4 per cent of children seen at 4.5 years of age were identified as having reduced vision. This is higher than other population-based studies: 6.2 per cent of Australian children,³⁹ 6.2 per cent of African-American and 7.5 per cent of Hispanic preschool children,⁴⁰ and 3.7 per cent of 3-6 year old Chinese children.³⁷ The prevalence of reduced vision in New Zealand children born without risk factors for hypoglycaemia is unknown. Further research is needed to understand the assessment of visual function in children with typical development.

In addition, the visual outcomes of children who did not return for vision screening at 4.5 years of age is not known. Children may not have returned for their 4.5-year vision screen because they were referred for further eye care during their initial screen; therefore, information is lacking on whether referrals resulted in additional assessment, care and improvement in visual outcomes in these children. Furthermore, there are difficulties in comparing children at two and 4.5 years of age without knowledge of the vision care received in the intervening period. This is particularly true for children who were referred at two years and potentially underwent therapy for amblyopia and/or strabismus, and improvements in visual acuity due to surgery or spectacle wear may have limited the predictive value of vision screening at two years of age.

Comparisons between the two ages are also complicated by the low success rate of monocular habitual visual acuity testing at two-years compared with 4.5 years of age. Children with strabismus and/or amblyopia may have achieved normal vision at two years of age when tested binocularly, but poorer visual acuity at 4.5 years under monocular conditions.

Overall, these results suggest that vision screening using age-appropriate testing techniques at two years of age does not predict visual outcomes at 4.5 years of age. Vision screening conducted at 2 years appears to have poor sensitivity and specificity for detecting mild-moderate visual impairment later in childhood. These results are relevant to the recommendations on preschool screening, which support vision screening only for children ≥ 36 months. However, additional research into alternative vision screening tests appropriate to use at two years, for example computerised assessment of acuity using optokinetic nystagmus,⁴¹ may help improve the effectiveness of early vision screening protocols.