Residual amblyopia continues to be a significant problem. Data from randomized trials conducted by the Pediatric Eye Disease Investigator Group show at least 65% of children aged 3 to 6 years with moderate initial amblyopia had visual acuity of 20/32 or worse in the amblyopic eye after 4 months of treatment with either patching1 or atropine eyedrops.2 In older children, Pediatric Eye Disease Investigator Group studies suggest that as many as 36% of children aged 7 to 12 years fail to reach 20/32 or better visual acuity following 24 weeks of maximum treatment with patching and/or drops.3 Furthermore, 78% failed to achieve two or more levels of stereoacuity improvement.3 These high rates of residual amblyopia and subnormal stereoacuity necessitate a search for factors that limit visual improvement.
Anisometropia is present in ∼50% of children with amblyopia who are identified by population sampling.4,5 It is strongly related to amblyopia,6,7 especially when combined with aniseikonia.7 Aniseikonia is a perception of unequal image size, which may be experienced by children with anisometropia. In children, anisometropia usually results from differences in axial length, which can produce differences in retinal image size resulting in aniseikonia. In addition to anisometropia, aniseikonia is associated with reduced stereoacuity,8–10 reduced contrast sensitivity function,11 and reduced fusional vergence ranges.12 It may also be associated with suppression and eccentric fixation.13,14 Any of these factors could limit amblyopia treatment outcomes.
Symptoms of aniseikonia, such as asthenopia, headaches, and reading difficulty, may begin with as little as 2% aniseikonia.15 However, at 5%, aniseikonia has been shown to reduce binocular summation11,16 and, at higher levels (8 to 10%), induce binocular inhibition,16 potentially limiting amblyopia treatment outcomes even further.
No instrument used to test aniseikonia has been shown to be reliable and accurate in children. Furthermore, the most useful test for evaluating aniseikonia in the adult population has been debated in the literature since 1943.17 Some instruments, such as the “original eikonometer” developed in 1932,18 and other “standard” eikonometers require simultaneous binocular perception but not stereopsis. Others, such as the “space eikonometers” do require stereopsis.19 Other means to determine aniseikonia discussed over the last century include the use of iseikonic (or “size”) lenses, the Hess chart,20 estimation using the “Rule of Thumb,”21 the Binocular Aniseikonia Test Chart,22 electrophysiological detection,23,24 and a cheiroscope.25 The Awaya New Aniseikonia Test was developed more recently and may be more readily available and more commonly used, especially among children. It uses direct comparison through anaglyphic dissociation in a hand-held test booklet.26 However, none of these instruments remain a ubiquitously accepted gold standard because they either are not readily available or reportedly27 (and debatably28,29) underestimate aniseikonia.
The Aniseikonia Inspector is a commercially available computer-based test developed by Optical Diagnostics (de Wit)30,31 in 2003. Its administration is quick and easily understood and interpreted. This instrument also uses direct comparison through anaglyphic dissociation using two adjacent semicircles. Measurements in adults were shown to be most accurate and reliable using version 1 in the vertical meridian in the dark. Under these conditions, the mean slope for plots of induced aniseikonia vs. size lens magnification was 1.038.32 The purpose of this study is to evaluate the ability of elementary-aged school children with normal vision to perform computer-based matching and to evaluate the accuracy and reliability of size lens-induced aniseikonia measurement using the Aniseikonia Inspector version 1 with large vertical targets in the dark.
The study was approved by University of Alabama at Birmingham Institutional Review Board. Informed parental consent for younger children or signed parental permission and child assent for older children was obtained from all participants. The research followed the tenets of the Declaration of Helsinki.
For this exploratory study, we planned to recruit a convenience sample including all children attending grades 1, 3, and 5 attending a single elementary school. We hoped to enroll 20 to 25 children per grade. Our goal was to determine the accuracy and reliability of size matches made by children using the Aniseikonia Inspector software.
To characterize the population studied, children completed measures of habitual Snellen monocular distance visual acuity at 6 M, alignment at distance and near using cover test, and stereopsis using Randot global and Wirt Circles. All but one participant, who was 20/50 in each eye, were 20/40 or better in each eye with 95% of participants 20/30 or better (participant average 20/18 distance monocular acuity). All had normal stereopsis (250 informs and 70 in or better Wirt Circles), and all had normal cover test findings at distance (orthophoria) and near (no strabismus and 6 prism diopter or less exophoria and no esophoria). Visual acuity and stereopsis were tested by one experienced optometrist and/or one of two pediatric optometry research technicians. All cover test and aniseikonia testing was performed by a single investigator, the same experienced optometrist. Refractive data were not available for all participants.
The Aniseikonia Inspector (version 1) was used to assess accuracy and reliability of aniseikonia testing in children. The display shows a variable red semicircle on the child's right and a nonvariable green semicircle on the left. To provide binocular dissociation, we always used a green filter in front of the right eye and a red filter in front of the left eye. The child always adjusted the size of the right semicircle while viewing through the green filter until both semicircles appeared to be equal.
Size could be adjusted in 0.5% steps to a maximum minification or magnification of 25%. Testing distance was maintained at 76 cm, which fell within the instrument developer's guidelines (61 to 152 cm recommended testing distance). All semicircle targets were the larger of the software's size options (8.9 cm) and were oriented vertically. All testing was performed in the dark with laptop light emitting diode lights covered and with minimal to no natural lighting to minimize peripheral cues. Software, test distance, target size, and illumination were the same as those shown to be most reliable in adults by Fullard et al.32
There were four test conditions (one associated and three dissociated). First, children adjusted the test target without wearing red/green glasses. This was to discover how accurately children could match the target size without binocular dissociation. Next, children were tested with red/green glasses under three dissociated conditions: no size lens, size lens in front of the right eye (with green filter), and size lens in front of the left eye (with red filter). The order of the last two conditions was randomized according to a predetermined schedule.
Each child completed four adjustments, or trials, per condition. The testing sequence was computer driven and similar for the four test conditions. The starting size of the variable semicircle target was always 25% smaller on the screen compared with the nonvariable semicircle. The next variable semicircle target was 25% larger on the screen compared to the nonvariable semicircle. This sequence was repeated once, for a total of four trials per condition. With the size lens over the right eye, the resultant right target size was roughly 21.5% smaller on trials 1 and 3 and 28.5% larger on trials 2 and 4. With the size lens over the left eye, the resultant right target size was roughly 28.5% smaller on trials 1 and 3 and 21.5% larger on trials 2 and 4.Thus, the child was expected to enlarge the right target on trials 1 and 3 and to reduce the target on trials 2 and 4 across all conditions.
Children were instructed that vertical head movements were allowed, but forward-backward or left-right head movements were not allowed. Software testing was completed in one session and took ∼10 to 15 min per participant.
To assess the repeatability of the four adjustments that are averaged to determine the measure of aniseikonia, the repeatability coefficient was calculated as defined by the British Standards Institution for more than two repeated measurements.33,34 The SD of repeated measurements is calculated as the within-subject SD or sw. In the case of four measures, sw was calculated using a one-way analysis of variance with subjects as the grouping variable. The within-subject variance is the residual mean square in the analysis of variance table, and sw is the square root of the within-subject variance. The repeatability coefficient is defined as 2.77 × sw for a 5% error rate. This means that a difference between two successive adjustments of <2.77 × sw cannot be distinguished from measurement error.
Plots of SDs against the means of the adjustments and factorial analysis of variance were used to determine whether were any systematic changes in the four adjustments over time or interactions with grade and experimental conditions. Accuracy and reliability of aniseikonia measures were determined by examination of distributions of the means determined from the four adjustments as a function of grade and experimental condition.
Fifty-seven children were available on the date of testing and performed study measures. More children attended from grades 5 (n = 26) and 3 (n = 21) vs. grade 1 (n = 10).
Fig. 1 shows the relationships between the SDs and the means of the four adjustments that a subject makes to generate the measure of aniseikonia. The means and SDs cluster very closely in the control condition with no filters in place. When the filters are added, there is still a tight cluster of points with three points showing larger SDs. With the size lenses in place, the dispersion of both the means and the SDs increases markedly.
Table 1 shows the analysis of variance that includes the individual trials. This analysis was done primarily to determine the within- subject variance used to calculate the repeatability coefficients for the individual adjustments. There are significant effects of experimental condition and trial along with a significant interaction effect between grade and condition. The effect of condition is due to the size lens, and the effect of trial is due to alternating the child's task between enlarging and reducing the variable target.
The effect of trial is small as shown in Fig. 2. There is a trend from a small negative value to a small positive value over the four adjustments. The significant difference shown in the analysis of variance table is the result of trials 1 and 4 being significantly different (Tukey's unequal Honest Significance Difference test, p = 0.006). On average, children underestimated the necessary enlargement to match the nonvariable semicircle at trial 1 and underestimated the necessary reduction at trial 4. These changes are all smaller than the smallest discernable aniseikonia measure of 0.5%.
Table 1 also shows a significant interaction between grade and condition. This is because grade 1 subjects have aniseikonia measures that tend more toward 0 than the other two grades. This interaction disappears when the analysis of variance is done on the average of the adjustments (see below and Table 2).
The repeatability coefficients for the adjustments under the four measurement conditions are shown in Fig. 3 for all subjects and by each grade. Each coefficient can be used to calculate the value below which the difference between any two measures will lie for 95% of the time. Note that the coefficient is expressed in the same percentage units as the adjustments themselves. Thus, a repeatability coefficient of 1.0 would mean that 95% of the differences between successive adjustments would be ≤1.0%. Larger coefficients reflect more variability in the successive differences, not a shift in the mean of the differences to higher values; the grand mean of successive differences is, in fact, only −0.12%.
The repeatability coefficient is lowest for the control condition in which the subjects adjusted the size of the half-circles without any filters in place. The coefficients increase by ∼0.5% when the filters are added but no size lens is used. With the size lens in front of the right eye, the coefficients increase furthermore by ∼1.5%. Variability increases even more when the size lens is shifted to the left eye. It is clear that the repeatability coefficients decrease as a function of grade for each measurement condition. Note that all the repeatability coefficients are greater than the smallest discernible aniseikonia measurement of 0.5%.
The plots in Fig. 4 and the data in Table 3 show the results relevant to the accuracy and reliability of the measures of aniseikonia. The measured aniseikonia in this figure is the mean of the four adjustment trials discussed above. Consider the issue of accuracy first. If the aniseikonia measurements were perfect, the means would fall on the 1:1 lines. In all cases, the measured aniseikonia is less than the induced aniseikonia. The slopes of the regression lines are all <1.0 and, as is evident in the confidence limits for the regression lines, significantly different from 1.0.
The whiskers in Fig. 4 are the 95% confidence intervals for the means under each experimental condition. First, note that there is a considerable difference for all three grades between the width of the confidence intervals and ranges when there is no size lens in place and the two conditions when the size lens was used. Second, note that the width of confidence limits is greater for grade 1 than for grades 3 and 5. Part of that is because of the smaller sample size for grade 1. However, the ranges are not as consistent. In fact, Table 3 shows that the largest range is for grade 5 when the size lens was in front of the left eye.
Although these features of accuracy and reliability and the repeatability coefficients discussed earlier seem consistent with the notion that older students are better at making the adjustments, the analysis of variance using the means of the four adjustments show a significant effect for condition only—again, not unexpected. Unlike the previous analysis of variance, which included the adjustment trials, there is no significant interaction effect between grade and condition.
Children were included regardless of their vision measures. Because refractive data were not available, our measures of aniseikonia might be influenced by uncorrected refractive error. Also, our results apply to conditions where adjustments are made by children who are viewing a red target through a green filter and may be different under different chromatic conditions.
Based on the results of the first condition, without dissociation, we conclude that children in grades 1, 3, and 5 are capable of understanding the task and making the adjustments necessary to measure aniseikonia with the Aniseikonia Inspector software. In this control condition, the subjects were simply required to adjust the size of one semicircle to match the size of a fixed semicircle under associated binocular conditions.
Based on the results of the second condition in which the subjects were required to make the same adjustments while viewing the two semicircles with red/green dissociating filters, we conclude that children as young as 6 years are capable of making the adjustments necessary to obtain a measure of aniseikonia with sufficient accuracy and reliability to be clinically useful. As shown in Table 3, the mean and the 95% confidence intervals for the “no size lens” condition fall within one software-driven 0.5% step in all grades tested. This means that the chance of a false positive reading of 0.5% or more for a first grader with no aniseikonia is <5%. The results are even better for the third- and fifth-grade students. Because this test condition in children without aniseikonia most resembles testing required in children suspected of having aniseikonia, these results suggest that the Aniseikonia Inspector (version 1 with large targets in the dark) may accurately determine aniseikonia in situ. Aniseikonia <0.5% may not be detected; however, as noted earlier, most subjects are not symptomatic until aniseikonia reaches 2.0% and binocular inhibition may not begin until 5.0%.
In contrast, the introduction of the 3.5% size lens in front of either eye decreases the accuracy and the reliability as shown in both Table 3 and Fig. 4. The means for all three groups of children indicate that the full magnitude of the induced aniseikonia is not being measured. Averaged across all subjects, the measured aniseikonia is ∼1% less than the induced aniseikonia. These underestimates result in the slopes of the regression lines being significantly different from 1.0 as shown in Fig. 3.
Similarly, the widths of the confidence limits are much larger when the size lens is in place than when it is not. The ratio of the widths of the confidence limits with the size lens in place to those with only the filters in place is 1.6 for all subjects. It is clear from Table 3 that the first-grade students have confidence limits that are two to three times as large as those of the other two grades.
The source of this increase in variability with a size lens in place can be traced to the repeatability of the four individual adjustments that are averaged to get the aniseikonia measures. This is evident in Fig. 1, which shows that, although they are not systematically related, both the means and the SDs increase with the size lens in front of either eye.
The degree to which this happens is shown very clearly in Fig. 3. The repeatability coefficients indicate the values below which 95% of the differences between any two adjustments should fall. All of these differences, including those for the control conditions, are >0.5%. It is not surprising that these values are as large as they are because they represent the individual adjustments that are averaged to get the final measure of aniseikonia. What is important about the repeatability coefficients is that they show a substantial increase when the size lens is in front of either eye. Furthermore, despite randomized order, the test condition with size lens combined with the red filter over the left eye is consistently less repeatable in all grade levels than any other test condition including the size lens and green filter over the right eye. The effect of chromatic influences and other factors, including neurological adaptation, which may affect reliability measures between eyes when aniseikonia is induced with size lenses should be considered and evaluated in future studies. However, such factors may not be relevant to values obtained from children with natural aniseikonia because of anisometropia.
The question that naturally arises from these data is whether children in the age range tested here who have a natural aniseikonia will have greater difficulty making the adjustments than those without aniseikonia. Fullard et al.32 found that, although there appeared to be a slight underestimation of the amount of induced aniseikonia with size lenses in adults, the differences were not significant, and the slope of the line was not significantly different from 1.0. In addition, it appears that the variability in the measurements was fairly constant over the entire range of induced aniseikonia.32 This is in contrast to the current findings where variability with no induced aniseikonia was much smaller than that observed with the size lens in place. Such variability should be further studied in children with natural, axial anisometropia, who may be at greatest risk for aniseikonia in situ.
Children as young as 6 years are capable of making the adjustments necessary to obtain a measure of aniseikonia under dissociated conditions with sufficient accuracy and reliability to be clinically useful using the Aniseikonia Inspector software (version 1 with large targets in the dark). Confidence level widths suggest a measurement error of 0.5% or one software-driven measurement step. Therefore, the software as tested here is capable of detecting aniseikonia as small as 0.5 to 1% in children as young as 6 years. Further study is necessary to show how children with natural aniseikonia respond to testing.
We thank Drs. Robert Rutstein and Roderick Fullard at UAB and Gerard de Wit of Optical Diagnostics for providing insight during the study.
This study was supported by the Department of Optometry at the University of Alabama at Birmingham.
The authors have no conflicts of interest related to the product evaluated. This paper was presented as a poster at the American Academy of Optometry meeting in Anaheim, CA, on October 22–25, 2008.
Katherine K. Weise
Henry B. Peters Building
1716 University Boulevard
Birmingham, Alabama 35294
1. Repka MX, Beck RW, Holmes JM, Birch EE, Chandler DL, Cotter SA, Hertle RW, Kraker RT, Moke PS, Quinn GE, Scheiman MM; Pediatric Eye Disease Investigator Group. A randomized trial of patching regimens for treatment of moderate amblyopia in children. Arch Ophthalmol 2003;121:603–11.
2. Repka MX, Cotter SA, Beck RW, Kraker RT, Birch EE, Everett DF, Hertle RW, Holmes JM, Quinn GE, Sala NA, Scheiman MM, Stager DR Sr, Wallace DK; Pediatric Eye Disease Investigator Group. A randomized trial of atropine regimens for treatment of moderate amblyopia in children. Ophthalmology 2004;111:2076–85.
3. Scheiman MM, Hertle RW, Kraker RT, Beck RW, Birch EE, Felius J, Holmes JM, Kundart J, Morrison DG, Repka MX, Tamkins SM; Pediatric Eye Disease Investigator Group. Patching vs atropine to treat amblyopia in children aged 7 to 12 years: a randomized trial. Arch Ophthalmol 2008;126:1634–42.
4. Brown SA, Weih LM, Fu CL, Dimitrov P, Taylor HR, McCarty CA. Prevalence of amblyopia and associated refractive errors in an adult population in Victoria, Australia. Ophthalmic Epidemiol 2000;7:249–58.
5. Multi-Ethnic Pediatric Eye Disease Study Group. Prevalence of amblyopia and strabismus in African American and Hispanic children ages 6 to 72 months the multi-ethnic pediatric eye disease study. Ophthalmology 2008;115:1229–36.e1.
6. Leon A, Donahue SP, Morrison DG, Estes RL, Li C. The age-dependent effect of anisometropia magnitude on anisometropic amblyopia severity. J AAPOS 2008;12:150–6.
7. Lubkin V, Kramer P, Meininger D, Shippman S, Bennett G, Visintainer P. Aniseikonia in relation to strabismus, anisometropia and amblyopia. Binocul Vis Strabismus Q 1999;14:203–7.
8. Jiménez JR, Ponce A, del Barco LJ, Diaz JA, Pérez-Ocón F. Impact of induced aniseikonia on stereopsis with random-dot stereogram. Optom Vis Sci 2002;79:121–5.
9. Oguchi Y, Mashima Y. The influence of aniseikonia on the VEP by random-dot stereogram. Acta Ophthalmol (Copenh) 1989;67:127–30.
10. Lovasik JV, Szymkiw M. Effects of aniseikonia, anisometropia, accommodation, retinal illuminance, and pupil size on stereopsis. Invest Ophthalmol Vis Sci 1985;26:741–50.
11. Jiménez JR, Ponce A, Anera RG. Induced aniseikonia diminishes binocular contrast sensitivity and binocular summation. Optom Vis Sci 2004;81:559–62.
12. Sharma P, Prakash P. Effect of aniseikonia on fusion. Indian J Ophthalmol 1991;39:170–3.
13. Nordlow W. Anisometropia, amblyopia, induced aniseikonia and estimated correction with iseikonic lenses in 4 year-olds. Acta Ophthalmol (Copenh) 1970;48:959–70.
14. Scheiman MM, Wick B. Clinical Management of Binocular Vision, 2nd ed. Philadelphia, PA: Lippincott Williams and Wilkins; 2002.
15. Hawkswell A. Routine aniseikonic screening. Br J Physiol Opt 1974;29:126–9.
16. Katsumi O, Tanino T, Hirose T. Effect of aniseikonia on binocular function. Invest Ophthalmol Vis Sci 1986;27:601–4.
17. Ogle KN. Association between aniseikonia and anomalous binocular space perception. Arch Ophthalmol 1943;30:54–64.
18. Bannon R. Clinical Manual on Aniseikonia. Buffalo, NY: American Optical Corporation; 1954.
19. Kramer PW, Lubkin V, Pavlica M, Covin R. Symptomatic aniseikonia in unilateral and bilateral pseudophakia. A projection space eikonometer study. Binocul Vis Strabismus Q 1999;14:183–90.
20. Hirai T, Sato M, Piao CH, Miyake S, Terasaki H, Ohyagi W, Miyake Y. Dynamic aniseikonia measurement: prismatic effect appears on the Hess chart. Binocul Vis Strabismus Q 2004;19:234–45.
21. Lubkin V, Shippman S, Bennett G, Meininger D, Kramer P, Poppinga P. Aniseikonia quantification: error rate of rule of thumb estimation. Binocul Vis Strabismus Q 1999;14:191–6.
22. Liu AN, Yan SM. Study and application of “binocular aniseikonia test chart” [in Chinese]. Zhonghua Yan Ke Za Zhi 1994;30:348–50.
23. Lovasik JV, Robertson KM. Electrophysiological detection of aniseikonia. J Am Optom Assoc 1984;55:499–504.
24. Katsumi O, Hirose T, Tanino T, Uemura Y. Pattern reversal VER as a tool for evaluating unbalanced visual inputs between the two eyes. Jpn J Ophthalmol 1988;32:86–97.
25. Venger GE. Examination of aniseikonia in unilateral aphakia with the use of the apparatus “Kheiroscope” [in Russian]. Oftalmol Zh 1974;29:149–50.
26. Yoshida M, Sato M, Awaya S. Evaluation of the clinical usefulness of the New Aniseikonia Tests [in Japanese]. Nippon Ganka Gakkai Zasshi 1997;101:718–22.
27. McCormack G, Peli E, Stone P. Differences in tests of aniseikonia. Invest Ophthalmol Vis Sci 1992;33:2063–7.
28. Antona B, Barra F, Barrio A, Gonzalez E, Sanchez I. The validity and repeatability of the New Aniseikonia Test. Optom Vis Sci 2006;83:903–9.
29. Awaya S, Romano PE. Aniseikonia tests: valid = “true” but “true” is not equal to p < or = .05 or p anything. The NAT is useful. Binocul Vis Strabismus Q 2003;18:69–71.
30. de Wit GC. Clinical usefulness of the Aniseikonia Inspector: a review. Binocul Vis Strabismus Q 2008;23:207–14.
31. de Wit GC. Evaluation of a new direct-comparison aniseikonia test. Binocul Vis Strabismus Q 2003;18:87–94.
32. Fullard RJ, Rutstein RP, Corliss DA. The evaluation of two new computer-based tests for measurement of Aniseikonia. Optom Vis Sci 2007;84:1093–100.
33. British Standards Institution. Precision of Test Methods: Guide for the Determination of Repeatability and Reproducibility for a Standard Test Method, Part I: BS 5497. London, United Kingdom: British Standards Institution; 1979.
34. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–10.