Background/aims: To evaluate a new autorefractor, the Plusoptix Vision Screener (PVS), as a screening tool to detect risk factors for amblyopia by comparing it with gold-standard orthoptic vision screening in children.
Methods: Community-based screening study including 288 children age 4–7 years who were screened with the PVS and by orthoptic assessment (distance acuity, cover test, extraocular movements, 20 PD prism test, Lang stereotest). Follow-up comprehensive eye examination of screening-positive children included manual cycloplegic retinoscopy.
Results: Testability was high for both methods. Orthoptic screening identified 36 children with reduced vision and/or factors associated with amblyopia (referral rate 12.5%). The PVS identified 16 children with potential vision problems (referral rate 5.6%), indicating only moderate sensitivity (44%; 95% CI 27.9 to 61.9%), but high specificity (100%; 95% CI 98.5 to 100%) to detect factors associated with amblyopia. The PVS underestimated visually significant refractive errors.
Conclusions: Use of the PVS as single screening test in young children may miss a significant number of children with amblyopia or amblyogenic risk factors.
Statistics from Altmetric.com
With a prevalence of 1.6–3.6%, amblyopia is the most common visual deficit in children.1 Amblyopia is a developmental disorder of spatial vision associated with strabismus, anisometropia or form deprivation early in life.2 Recent large studies have shown that treatment implemented within the first years of life is highly effective.3–8 From the point of view of public health, amblyopia treatment in children may be 44 times more cost-effective than cataract surgery in adults.9 Early detection of amblyopia and associated risk factors is therefore desirable,10 11 but the choice of screening methods is controversial. The gold-standard test for vision screening in children of preschool age is assessment by an orthoptist who performs a battery of tests, which in combination achieve high sensitivity and specificity.12 13 Disadvantages are the need for highly trained staff and the variability in children’s cooperation with the tests.
The recent Vision in Preschoolers (VIP) study evaluated several screening tests including orthoptic assessment, photorefraction and isolated visual acuity (VA) measurement, and found that lay screeners can achieve pickup rates similar to trained staff when using appropriate tests.12 There is great interest in developing automated screening methods which can be operated by lay screeners and require less cooperation from children. However, due to limitations in accuracy, to date no device has found widespread use.14–17 Since the conclusion of the VIP study, a new autorefractor, the Plusoptix Vision Screener (Plusoptix GmbH, Nuremberg, Germany), has become available. This device is marketed as a vision screening tool for young children, but no data evaluating its use as a screening instrument are available.
We set out to compare the ability of the Plusoptix Vision Screener (PVS) to detect refractive errors and other amblyogenic risk factors in a community cohort of 4–7-year-old children, compared with the “gold-standard” orthoptic assessment. The results of our study highlight the need for orthoptist-led vision screening in preschool children and may inform health policy makers and vision screening providers about the advantages and limitations of this new device.
MATERIALS AND METHODS
The PVS performs real-time videoretinoscopy of reflected infrared light in three meridians. The refractive data provided as final output are the median of six frames of the acquired video sequence on which the software recognises both pupils. We used the device operated in Screening mode run by software version CR03 V.126.96.36.199.
Based on published literature we assumed a referral rate of 15–20%. We considered screening 400 children to be within the capacity of our department. We calculated that this sample size would be sufficient to estimate sensitivity of 60% with a precision (ie, 95% CI) of at most ±12% and specificity of 80% of at most ±4%.
After obtaining ethics committee approval, we invited 379 children attending reception class or year 1 at six local schools to take part in the study. Parents were asked to provide written informed consent. A total of 288 children took part; the remainder were excluded either because their parents did not give consent or because the children did not attend school on the day we conducted the study. We included children with special educational needs (n = 11): six with autistic spectrum disorder, two with language difficulties, one each with global developmental delay, attention deficit/hyperactivity disorder and learning difficulties.
With the final number of 288 recruited subjects and the final gold-standard referral rate of 12.5% (see below) the estimated precision of the study in determining autorefractor sensitivity and specificity would be expected to be ±16% for sensitivity and ±5% for specificity.
We tested each child with the PVS operated by an ophthalmologist. An orthoptist masked to the outcome of the autorefractor result performed distance acuity testing, cover test, extraocular movements, 20Δ base out prism test, and stereotest (Lang). Two opthalmologists and two orthoptists performed all screening examinations.
Children who fulfilled the referral criteria (table 1) were referred to our hospital-based paediatric eye service and underwent full ophthalmic and orthoptic assessment including manual cycloplegic refraction (MCR). At the hospital, four ophthalmologists and three orthoptists were involved.
We compared the testability (percentage of children in whom the test could be performed) of both methods and evaluated the performance of the PVS in identifying children who were screening positive on the gold standard. We also calculated the spherical equivalent (SE) of the measurements obtained by the PVS and by MCR. To avoid bias from enantiomorphism (mirror symmetry between the two eyes of the same individual), we only analysed right eyes.
For statistical analysis, continuous variables are presented as means (SD) if normally distributed, or medians with interquartile range (IQR) otherwise. Categorical variables are described as frequencies with percentages. The diagnostic ability of the PVS is presented in terms of sensitivity and specificity along with their 95% CIs. All statistical analyses were performed with the software package SPSS, version 14.0.
Of 379 invited children, 288 took part (76%). Their mean (SD) age was 5.6 (0.6) years (range 4.5 to 7.3). One hundred and fifty-one (52.4%) children were female and 137 (47.6%) male.
Testability was high for both methods (PVS 100%, orthoptic assessment 99.6%).
Orthoptists detected potential vision or eye alignment problems in 46 children. The standard referral criteria of our orthoptic screening service include parameters not usually associated with amblyopia: convergence insufficiency, inability to overcome a 20 PD base out prism, and latent strabismus. In the study population, either of these criteria led to referral of 10 children. As the present study aimed to identify factors associated with amblyopia, we based the comparison of the two methods only on those 36 children who were orthoptic screening positive for reduced vision in one or both eyes, manifest strabismus or ptosis (gold-standard referral rate 12.5%).
The PVS recommended referral of 16 children (referral rate 5.6%), all of whom were also gold-standard positives (table 2). Thus, the diagnostic ability of the PVS in predicting the gold-standard assessment was 44.4% (95% CI 27.9 to 61.9%) sensitive and 100% (95% CI 98.5 to 100%) specific. Table 3 shows the frequencies of referral reasons for both methods.
As for refractive errors, the PVS found most children in the study cohort to be emmetropes or low hypermetropes. The mean (SD) spherical equivalent was +0.2 (0.5) D (fig 1).
Examination of children referred to the hospital
Of the 46 children referred for further evaluation, 12 were already known to the hospital eye service, and no further action was taken. We collected data on their diagnoses from the hospital notes, including the results of the last MCR. Of the other 34 for whom referral was recommended, six were not referred by their family physician, and one did not attend the hospital eye service within 4 months of screening, when final data analysis was performed, leaving 12 known and 27 new (total 39) cases for analysis (85% of all recommended referrals).
Gold-standard orthoptic screening and hospital examination were highly congruent. Of those 36 children referred for amblyopia, manifest strabismus or ptosis, two were found to be false positives. One child had a VA of 0.175 (logMAR) at screening but slightly better on hospital examination (0.16), and the other child had a VA difference of one line on screening, but slightly less on hospital follow-up.
We confirmed 24 children to have reduced vision in one or both eyes as defined by the referral criteria. Of these, the PVS had identified 12. Table 3 summarises all findings at hospital examination.
Seventeen children had visually significant refractive errors, of which the PVS identified five and delivered a screening outcome of “no reading possible—refer” in a further five (false negatives 41%). Most false negatives related to hypermetropia of the spherical component >+3.0 D (in negative cylinder notation), of which there were 13 cases. The PVS correctly identified one of these and gave a screening outcome of “no reading possible—refer” in a further five. The SE of children examined at the hospital differed markedly from the PVS measurements at screening, with a median (IQR) difference of 1.1 (0.1 to 2.0) D (n = 30). The PVS measurements were 2.0 to 3.5 D lower than the MCR measurements. The device did detect all four cases of anisometropia present in the study population.
At screening, the PVS failed to obtain a refractive reading in seven children. On hospital examination, all had either lid position anomalies or large-angle strabismus.
Two children referred to the hospital were PVS positives, but manual inspection of the device data revealed that the PVS measurements did not meet the preset referral criteria. This was due to a software error, which the manufacturer states has been corrected in the latest software version.
This is the first study testing the Plusoptix Vision Screener as a stand-alone preschool visual screening instrument and comparing it with the gold-standard orthoptic examination. In the absence of data about the screening performance of this new device, our original sample size calculations were based on published prevalence rates for amblyopia and associated conditions and referral rates of other screening methods. The final number of subjects included was smaller than intended, resulting in wider confidence intervals. Despite these limitations, our study provides a clinically highly informative result: when set to the referral thresholds detailed in table 1, the sensitivity of the PVS to detect amblyopia-associated factors is 44%, with 95% CI from 28 to 62%. Even a sensitivity of 62% would be less than desirable for a standalone screening test. This figure will also be useful as a starting-point for future studies.
In the present study, the number of children who were gold-standard screening positives is relatively small. This means that small differences in true positive referrals would have a greater effect on sensitivity calculations. We found that two of our 36 gold-standard referrals were false positives, which may increase the calculated PVS sensitivity to 47%.
Another methodological limitation of our study is the number of observers involved. Screening was performed by two orthoptists and two ophthalmologists, which may have introduced some interobserver variability. We attempted to reduce interobserver variability for the gold standard by performing the same battery of tests on all subjects and by following referral criteria laid down at the beginning of the study. As for the PVS, we determined intra- and interobserver variability of measurements as part of a separate study into device reliability and accuracy18 and can exclude any significant influence. Lastly, with view to the comparison of refractive data, our results are similar to those reported by other studies.19–21
This study evaluates the performance of the PVS in detecting factors associated with amblyopia. Two other recently published studies investigated the refractive performance of this new device. Ehrt et al reported a sensitivity of 70% in detecting amblyopia-associated factors, but recruited their study cohort in a children’s eye unit.19 The far higher prevalence of eye disease in this study population will affect sensitivity calculations. Clausen and Arnold compared the PVS with acuity testing and physician-interpreted photoscreening in healthy preschool children. They gave a sensitivity “estimate” of 67%, but their study did not compare the PVS with gold-standard orthoptic assessment.22 Our finding of a lower sensitivity is similar to the VIP study report that with sensitivity set to 94%, the PVS precursor model, the PowerRefractorII, has a sensitivity of 57% to detect amblyopia, 34% for strabismus, 42% for refractive error and 27% for reduced VA.12
In the present work, the PVS only identified 12 of 24 children with reduced unaided VA (50%), the primary target group. The use of the PVS as standalone screening test may miss a significant number of children with eye problems. However, an autorefractor might enhance orthoptist-led vision screening by complementing the functional assessment in borderline cases and those with poor compliance. The PVS had excellent testability in our cohort, which included 11 children with behavioural problems. With improved accuracy in the detection of visually significant refractive errors, the PVS may add to the orthoptic test battery. The limitations in refractive accuracy of the PowerRefractor were well known, and the first reports on the new autorefractor indicate similar problems.12 14 15 17 19 20 In our study, both the spherical equivalent (SE) of +0.20 D and the narrow SD of refractive errors found by the PVS were surprising for a population of 4- to 7-year-old mainly Caucasian children. Previous reports using MCR found higher mean refractive errors in similar populations of children, with an SE of +0.73 D and +0.85 D in studies on children from a broad range of ethnic backgrounds to +1.77 D in mainly Caucasian children of 6 years of age.23–27
Comparison with MCR in our study confirmed that the PVS failed to identify a high number of children with significant refractive errors. Of 13 children with a spherical component exceeding +3.0 D, it only correctly identified one, though it flagged up a potential problem by advising referral as “no reading possible” in another five (these five children had lid position anomalies and/or large angle manifest strabismus, which interfere with binocularly simultaneous infrared autorefraction). The small number of cases with anisometropia, myopia and visually significant astigmatism in our study does not allow meaningful interpretation. The refractive performance of the PVS is the topic of a second study into repeatability and accuracy.18
The manufacturer has suggested two methods of improving the screening performance of the PVS: reducing the hypermetropia referral threshold to +1.50 DS and/or placing +3.00 DS lenses in front of children’s eyes during the test. However, Ehrt’s study has demonstrated that both methods only slightly improve sensitivity but adversely affect testability and specificity.19 We reanalysed our data with a hypermetropia referral threshold of +1.50 D. This would have identified an additional three children as PVS screening positives, increasing the sensitivity to 53%, still falling short of desirable performance for a standalone screening test.
In summary, the Plusoptix Vision Screener in its current form seems to have a moderate sensitivity for the detection of visually significant refractive errors in children in the age group at risk for developing amblyopia. It may not be suitable as a standalone screening test. Device refinements may develop it into a useful tool to enhance orthoptist-led vision screening.
We wish to thank AJV’s personal assistant, E Hutchence, who helped us organise this study with great enthusiasm and unfailing support. R Andrews gave invaluable help in seeing the children at their schools.
Competing interests: None.
Ethics approval: Ethics approval was provided by Suffolk Local Research Ethics Committee.
Patient consent: Obtained from the parents.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.