Comparison of current paediatric contrast sensitivity assessments using simulated reduced contrast thresholds

ABSTRACT Purpose There are limited tests of contrast sensitivity (CS) for use in children. The Hiding Heidi (HH) is suitable for all cognitive abilities, but has a ceiling effect. The Double Happy (DH) test has comparable thresholds to the Pelli Robson (PR), however the ability to detect changes in contrast has not been established. This study aims to compare contrast thresholds and agreement between HH and the DH, comparing to the PR chart in normal conditions and under reduced visual and lighting conditions. Methods Tests were repeated under different conditions to reduce the contrast. Room illumination was 20,900{plus minus}2% lux in bright conditions and 2,000{plus minus}2% lux in dim conditions, both conditions were repeated with the addition of simulation spectacles to reduce the clarity of vision. Participants’ CS was measured uniocularly using the PR, HH and DH tests. Results 50 participants, age 18–62 years (mean{plus minus}standard deviation: 24.5{plus minus}7.98), were assessed. On HH 94% (n = 47) reached the maximum score, with the DH it was 18% (n = 9). The difference in reduction between conditions was smaller with HH in comparison to PR and DH, but significantly different from baseline conditions. Under dim conditions the reduction in PR and DH was −0.21 logCS units, but only −0.04 logCS for HH. Conclusion The DH test has better agreement with PR than HH and is better at detecting CS changes, highlighting the advantages of use in clinical practice.


Introduction
Research emphasizes the importance of contrast sensitivity (CS) testing in children with disabilities or low vision in providing an accurate and full assessment of visual function performance. 1For instance, Good et al. exhibited that children with cortical visual impairment showed a significant reduction in CS compared to age-matched control children, demonstrating the value of CS testing. 2 However, currently there is a lack of efficient and reliable CS pediatric tests. 1 CS testing in children needs to overcome certain limitations such as boredom and loss of attention or cooperation, 3 or more importantly, the ability to identify letters.
Clinically, CS thresholds are often measured with the readily available gold standard adult test, the Pelli Robson (PR) chart, 4,5 widely demonstrated to have good repeatability and yield fast and reliable results. 6evertheless, children and other patients with disabilities or limited cognitive abilities are not able to respond to this test design, thus the need for a reliable alternative CS test which is suitable for the pediatric population.One of the commercially available pediatric tests is the Hiding Heidi (HH) 3 which is a simple detection task that can be undertaken by children with limited cognitive abilities.However, when comparing its validity and agreement to the PR chart, Leat and Wegmann 7 evidenced the ceiling effect of the test, with most normally sighted children correctly responding to the lowest contrast face, measuring on average 0.23 logCS units higher than with the PR test.To address this limitation, a novel pediatric CS was developed: the Double Happy (DH). 8The stimulus in the DH test is restricted to a narrow band of spatial frequencies, unlike the multiple spatial frequencies comprising the HH test.Also, it has intervals of 0.15 logCS up to 2.1 logCS compared to the 0.30 logCS intervals of HH just up to 1.9 logCS, making it more comparable to PR. 8 Nonetheless, the comparability of the test has not been reported.In addition, considering the limited number of contrast levels of Hiding Heidi, this may not be sensitive enough to detect subtle changes but has not been evaluated. 8his study aims to compare contrast thresholds and measure the agreement across the current pediatric assessment of contrast sensitivity HH and the novel pediatric CS test DH, with the gold standard PR chart at baseline conditions.By simulating reduced contrast levels under different light and vision conditions, comparisons between results across the assessments can be made to determine whether the Hiding Heidi and Double Happy CS tests can detect subtle changes in contrast.It is proposed that the Double Happy pediatric test will be comparable and in agreement with the PR, in normal lightning conditions, as well as able to detect small changes in contrast between normal testing conditions, reduced visual input and different lightning conditions.Whereas it is anticipated that the Hiding Heidi will present with a ceiling effect.

Participants
The study was conducted at the University of Liverpool from February to April 2022.Ethical approval was obtained from the University of Liverpool ethics committee (10724), the study conformed to the principles of the Declaration of Helsinki, and all participants provided informed consent.Potential participants (staff and students) within the School of Health Sciences at the University of Liverpool were contacted via e-mail with information about the study.Participants were also identified through social contacts.Prior to participating, all volunteers were given a participant information sheet explaining the study.A total of 50 volunteers were planned to be recruited for the study with written consent prior to their participation.Inclusion criteria was a minimum age of 18 years and no known ocular abnormalities.Adult participants were assessed due to the cognitive and concentration demand, to remove cognitive ability as a potential confounding variable.Participants unable to provide consent or with a cognitive impairment were excluded from the study.All testing was performed with participants wearing their habitual correction.

Visual acuity measurement
Participants underwent uniocular visual acuity (VA) measurement using the Thomson Test chart (Thomson Software Solutions) 2020 based ETDRS (Early Treatment of Diabetic Retinopathy Study) logMAR chart.Testing was performed with best corrected visual acuity and participants were seated four meters from the chart.Black out blinds and artificial lightning were used to ensure standardization.
Participants were asked to read the letters starting at 0.20 logMAR.Using a handheld occluder, the right eye was tested first, followed by the left eye, randomizing the optotypes in between to avoid memorization.VA was scored by awarding 0.02 logMAR units for each optotype correctly identified.All participants were encouraged to keep trying to read the smallest optotype even if they thought they would get it wrong, until a complete line of errors was made.
Visual acuity of the participants' weakest eye was also measured with induced reduced clarity of vision (which aims to reduce VA to around 0.48 logMAR) using the VINE (Visual Impairment North-East) simulation spectacles to confirm the reduction in vision. 9

Contrast sensitivity assessment
Participants' CS was measured uniocularly using the gold standard adult test PR chart and the pediatric tests HH and DH.Tests were repeated under different light and vision conditions to reduce the contrast.Maniglia et al. 10 reported a reduction in CS in low illuminance conditions compared to normal ones in young healthy participants, suggesting its importance to consider for the development of more sensitive CS tests.Illumination in the test room was altered to 20,900 ± 2% lux in bright conditions and 2,000 ± 2% lux in dim conditions, achieved by switching the overhead lights to their minimum and maximum settings.The lighting was measured using a digital light meter (Yanmis Luxmeter, LX101BS) to ensure conditions were standardized between participants.CS of the weaker eye, defined by VA measurements, was measured.CS tests were completed in a variable order for each participant, but the conditions were kept in the same order across tests and participants, starting with dim and defocused conditions (with VINE specs), followed by just dim lighting, then bright conditions with spectacles and bright without them.
The PR test consists of a chart with 16 triplets of Sloan letters of reducing contrast and fixed spatial frequency.From top to bottom, each triplet decreases in contrast by 0.15 logCS units, ranging from 0.00 to 2.25 logCS. 4In the present study, two PR charts (Precision Vision) were used to test CS to minimize memorization, swapping when light conditions were changed.Participants were asked to start reading the letters of 100% contrast at one meter distance until two or three letters of the same triple were not identified.CS was then recorded, awarding 0.05 logCS units for each letter correctly identified. 11The letters "C" and "O" were established as interchangeable. 12he HH low contrast "face" test consists of three cards with a face diagram printed on each side, with varying levels of contrast: 1.25%, 2.5%, 5%, 10%, 25% and 100%. 3In this study, a blank and a "face" card were presented side by side and the participant was asked if they were able to see Heidi and if so, to point at her.The testing distance was one meter and the 5% contrast test card was the first one presented, followed by presentation of the lower levels in order, to find the contrast threshold.The participant's contrast threshold was set at the level before the participant unsuccessfully identified Heidi twice.CS percentages readings were converted to logCS units for statistical analysis purposes, with readings of 1.9, 1.6, 1.3, 1, 0.6 and 0, respectively.
The DH test consists of a blank card and 15 one sided cards printed with a face stimulus identical when rotated 180 degrees, with contrasts ranging from 0.05 to 2.1 logCS units, differing by 0.15 logCS units in each card. 8The cards were presented to the participant from highest to lowest contrast starting at 1.35 logCS units, and participants were asked to point at the face stimulus or say if it was on the right or left-hand side of the card.The test distance was 40 cm, measured with the length of the test cards.The CS threshold was set as the lowest contrast card correctly identified by the participant at least twice.

Data and statistical analysis
Data were analyzed using the GraphPad Prism software.The distribution of the results was determined using the Kolmogorov-Smirnov test for normality.The PR CS test scores were compared to the HH and DH scores in best-corrected baseline bright conditions using a Pearson correlation, Bland-Altman agreement test and the Wilcoxon matched-pairs signed rank test.The ability of the tests to detect change was also compared using repeated measures ANOVA and Wilcoxon test.Data were expressed as mean ± standard deviation of logCS units and 95% CI, setting the statistical difference at p < .05.

Demographics
The present study evaluated a total of 50 participants, with no known ocular abnormalities, ranging from 18 to 62 years old (mean ± standard deviation: 24.5 ± 7.98).The mean binocular acuity (with habitual correction) of the participants was −0.04 ± 0.09 logMAR, the mean VA of the stronger eye was −0.06 ± 0.08 while the mean VA of the weaker eye used for further CS testing was −0.01 ± 0.09 logMAR.In defocused conditions with normal bright light, the mean VA of the weaker eye was 0.47 ± 0.09 logMAR.

Comparability
As the PR is the gold standard adult test for CS testing it was compared to the HH and DH tests in baseline habitual corrected bright conditions.The Kolmogorov-Smirnov test for normality demonstrated that the CS results were not normally distributed, thus the results were analyzed with the Spearman correlation and Bland-Altman agreement test, as well as the Wilcoxon matched-pairs signed rank test.
A strong positive correlation was found between the PR and the DH tests (r = 0.38) with a 95% CI ranging from 0.1 to 0.6.On the other hand, no correlation was found between PR and HH (r = 0.27), with a 95% CI ranging from −0.02 to 0.5.The anticipated ceiling effect was present for HH, with 94% (n = 47) of participants reaching the maximum score, evident by the HH points at 1.9 logCS (Figure 1).While DH also seemed to have a ceiling effect, just 18% of participants reached the maximum score of 2.1 logCS, and another 18% scored less than 1.9 logCS.
The Bland-Altman graphical analysis of results showed good agreement of PR with both HH and DH due to the mean difference close to zero.Moreover, most of the data fall within the limits of agreement in both cases (Figure 2).
The Wilcoxon test results showed PR to have significantly different results from HH (p < .0001)and DH (p < .0001).The PR contrast threshold scored gave lower logCS scores than HH, with a mean difference of 0.18 ± 0.1 and a 95% CI ranging from 0.15 to 0.21; and DH, with a mean difference of 0.24 ± 0.11 and 95% CI of 0.21 to 0.28.This discrepancy to the good correlation and agreement found suggests how the different features of the tests might contribute to contrast sensitivity, with the participants being able to score higher on HH and DH compared to PR, averaging 0.18 and 0.24 logCS units higher, respectively.Nevertheless, the Wilcoxon test also showed significant difference between HH and DH (p < .001),with a mean difference of 0.06 ± 0.1 and a 95% CI ranging from 0.03 to 0.09.

Ability to detect change
The mean and standard deviation logCS scores in bright conditions and the reductions derived from each other condition from the baseline condition are recorded in Table 1.Repeated measures ANOVA showed the significant effect of the different light and vision conditions on contrast sensitivity (p < .0001).
Analysis with the Wilcoxon test showed that the effects of lightning and vision conditions were significant on CS (p < .05),with CS being lower in dim, dim with specs and bright with specs conditions than that of bright conditions across CS tests.Notably, the difference in logCS reduction between conditions is smaller with the HH test in comparison to PR and DH, but still significantly different from baseline conditions (PR bright and dim conditions: p < .0001;HH bright and dim conditions: p < .05;DH bright and dim conditions: p < .0001).

Discussion
The study found that the DH is more comparable with PR than the HH in normal lighting conditions.Whist all three contrast assessments shown reduced contrast thresholds with reduced visual input and under different lighting conditions, the pediatric assessments shown less of a reduction, however again the DH was most comparable to the PR.The first point of interest in the study was the comparability of the CS pediatric tests at baseline habitual corrected bright conditions to the PR.As expected, given that the tests are measuring the same function, our findings showed a positive correlation between both tests and PR, with a higher correlation when compared with DH.Nevertheless, the data indicate that HH has a ceiling effect, with almost all participants (94%) reaching the lowest contrast level.In agreement with our results, a study by Leat and Wegmann 7 presented similar findings when testing CS in children between one and eight years old with normal vision, indicating that the HH is not able to measure true contrast thresholds due to the small number of contrast levels.Similarly, Elgohary et al. 13 investigated the age norms for grating acuity and contrast sensitivity in healthy children from three to 36 months old.Their results also suggested the limited use of HH on infants over two years old as the CS threshold increased gradually over the first nine months, but during the third year it stayed constant at the highest CS.Further analyses of the results showed PR to have significantly different results from HH and DH, with participants scoring 0.18 and 0.24 logCS units higher, respectively.These differences may be attributed to the individual characteristics of each test.For instance, the testing distance for DH was 40 cm compared to the one-meter distance for PR or HH, which studies have suggested to be critical.As reported by Njeru et al. 14 the average CS score on the PR chart was 0.398 logCS units higher when participants were tested at a closer distance than one meter based on the scored letter acuity.Other contributing factors might be the differences in spatial frequencies, as HH comprises mixed multiple spatial frequencies, unlike PR and DH, or the differences in logCS intervals across tests.
We also evaluated the three CS tests in different light and vision conditions to measure their ability to detect changes in contrast.Our findings showed that the effect of lighting and vision conditions were significant on the reduction of CS in all tests.However, the difference in logCS reduction between conditions in our study was notably smaller with the HH test in comparison to PR and DH, clearly showing a lack of sensitivity of HH to a change in logCS under dim conditions compared to baseline.Even though the HH shows a significant reduction in contrast, the DH shows a reduction in CS more than HH but not as much as PR.Presumably, a contributing factor along with the difference in test distance, could be the fact that it is a detection task with a 50% guess rate, the same as HH, compared to PR which has a forced-choice design using 10 Sloan letters.Nevertheless, despite the fact that the difference in logCS is statistically significant in HH between conditions, when considering the functional capacity of the test against DH, the results suggest that DH is a better alternative as a pediatric contrast sensitivity test to be used clinically.Moreover, DH is a more objective and unbiased test compared to HH as the examiner does not know where the stimulus is until after testing. 8urther tests would be needed to deduce repeatability as regards to test-retest reliability in child participants to further support their use in a clinical setting, as adult participants were only tested once by the same examiner.Indeed, simulated conditions can be controlled and set as needed, but they are not a perfect representation of actual conditions as they may increase awareness of the participant.Moreover, participants in the study mostly presented with normal VA so there was no range at baseline conditions, which does not represent the diversity of the population.Nevertheless, the present study suggests that the DH is a good pediatric alternative to PR in clinical practice for diagnosis of visual capacities of children and other patients with disabilities or limited cognitive abilities.

Conclusion
The data represent how different lighting and vision conditions affect CS and the ability of the tests to detect those changes.Particularly, our findings highlight how the DH test has better agreement with PR and is better at detecting changes in CS than Hiding Heidi.

Figure 1 .
Figure 1.CS (LogCS) of each participant (n = 50) under baseline habitual correction bright conditions, with the PR, HH and DH tests.

Figure 2 .
Figure 2. Bland-Altman agreement analysis of PR and (a) HH and (b) DH tests in baseline habitual corrected bright conditions.The mean difference (bias) is indicated by the solid line and the upper and lower limits of agreement by the dotted lines.

Table 1 .
Mean and standard deviation of logCS baseline conditions and the logCS reduction by the different lightning and vision conditions in PR, HH and DH (n = 50).