| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1From the Vision Assessment Unit, Department of Ophthalmology, Royal Liverpool Children’s Hospital, Liverpool, United Kingdom; the 2Unit of Ophthalmology, University of Liverpool, Liverpool, United Kingdom; and the 3Infant Vision Laboratory, The Smith-Kettlewell Eye Research Institute, San Francisco, California.
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
METHODS. The Lea Symbols Chart is a constant-crowding, equal-logMAR increment chart similar in design to the Early Treatment Diabetes Retinopathy Study [ETDRS] chart. LogMAR visual acuity was tested twice in each eye of 32 amblyopic and 11 normal children. Each test commenced with screening in which one of the three central symbols was chosen for identification starting with the 1.0- or 0.9-logMAR line, progressing to every second line until incorrect identification occurred. Symbol-by-symbol presentation then commenced at the logMAR line containing the last correctly identified symbol. The threshold was recorded as the last logMAR line where four of four or four of five correct responses occurred (i.e., line-by-line scoring). Retesting by the same examiner was identical and occurred within the same session.
RESULTS. There was no significant difference in repeatability among normal, fellow, or amblyopic eyes. The difference between test and retest thresholds lay between ±0.10 logMAR in 93% of eyes. The 95% limits of agreement for the difference was ±0.18 logMAR. Repeatability in eyes tested first did not differ from that in those tested second in either the normal or amblyopic groups.
CONCLUSIONS. In the age-group tested, the line-by-line method of threshold scoring compares favorably with previous reports of both line-by-line and interpolated threshold scoring. There was no clinically meaningful difference in repeatability between the normal and amblyopic children tested.
Visual acuity is an important clinical measure of visual function. The ETDRS-type chart and scoring protocol3 standardizes the administration and scoring of visual acuity and has arguably become the gold-standard instrument in both research and clinical settings. The method of threshold scoring used in the ETDRS protocol is that of letter-by-letter, or interpolation. In this method, each letter, or optotype, presented is assigned an equal logMAR score. Charts displaying five optotypes per line and graduated in 0.10-logMAR increments therefore assign a 0.02-logMAR score per optotype.
An alternative method of determining acuity thresholds is the line-by-line method. This method of assessment interprets each line as an individual psychophysical task that the patient must pass to progress. The line-by-line scoring system is accepted by authoritative bodies4 5 and is also in widespread use in clinical and pediatric research settings,6 7 8 9 10 11 including The Amblyopia Treatment Study.1
The stated benefit of interpolation (letter-by-letter) is that it is theoretically more sensitive to the measurement of change owing to the subdivision of the 0.10 logMAR incremental graduations of each chart line into an incremental scale five times finer (each optotype becomes an increment).12 13 14 The implication is that the line-by-line scoring rules may be insufficiently sensitive to detect and monitor treatment effects in for example amblyopia. However, a recent modeling study15 has shown relatively small differences between the two termination rules.
Empiric limits of the line-by-line and interpolation rules have been established only in adults and may differ in pediatric populations. Furthermore, despite a theoretical advantage in favor of the interpolation method, Carkeet15 makes a practical recommendation of using a line-by-line termination rule.
In addition to determining the test–retest reliability of line-by-line scoring in children, we also wanted to control for the effects of uncorrected refractive error. Several studies16 17 18 have indicated that uncorrected refractive error contributes to test–retest variability. A recent study involving normal adults18 highlighted wide variations in published repeatability data. Based on their work and that of others,6 12 13 14 16 17 18 19 20 21 22 Rosser et al.18 have suggested (hyperopic) optical defocus as one of the main sources of repeatability error. Other sources of error, such as the presence or absence of disease,14 16 21 whether the measurements were conducted in a single session,6 16 and the subject’s age19 20 have been considered and found not to contribute significantly to repeatability error. Of note regarding the issue of subject age, however, Beck et al.19 tested mostly adults with a mean age of 50 ± 22 years (SD) and Lovie-Kitchin and Brown20 tested adults older than 20 years.
Many studies1 2 8 9 10 11 23 24 25 26 27 deal with repeatability of recognition acuity in the <12-year-old group. In those in which crowding control was uniformly used1 2 9 10 23 24 25 26 and amblyopes and normal subjects were compared,1 2 23 24 26 only Holmes et al.1 and Moke et al.2 used line-by-line scoring. Furthermore, regardless of the method of scoring, only Moke et al.2 controlled for optical defocus over most (87.5%) of their study population. The test used by Moke et al. is a computer-based version of the HOTV test, which is not commercially available and costs approximately US $750 (raw materials and customization costs). In addition, the majority (78%) of their amblyopic group had acuities better than or equal to 0.40 logMAR (20/50).
The designs of previous studies in this area (i.e., repeatability of visual acuity in children) have ignored amblyopes, not uniformly controlled for optical defocus, or used acuity testing devices that are not commercially available. In this study, we addressed all the limitations of the prior studies. We chose crowded Lea symbols because, compared with HOTV, it has been shown to have superior testability in two10 28 of three10 28 29 large-scale pediatric studies. In addition, it is commercially available, inexpensive, and portable; is in widespread use; and most important, has been standardized.30 We chose the line-by-line scoring method because it also is in widespread use.1 6 7 8 9 10 11
In summary therefore, the primary purpose of this study was to inform clinical procedure by investigating test–retest variability of line-by-line scoring of the Lea chart in normal and amblyopic children having corrected uniformly for optical defocus in a standard protocol.
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Children prescribed spectacles were assessed every 4 weeks, to determine whether they were as yet accustomed to them. In amblyopes, this was deemed to have occurred on the first occasion after the spectacle prescription was issued when the fellow eye’s acuity became equal to or better than its previously recorded threshold in the unaided state (mean duration, 7.8 ± 3.4 weeks [SD]). In normal children, this criterion was deemed to have been met when the acuity in both eyes achieved the threshold recorded in the unaided state or better, which required 8.4 weeks on average (with some noncompliance, n = 2). At this time, the child was reexamined, and the diagnosis of amblyopia or normal control was made.
Amblyopia was diagnosed when the interocular acuity difference (IAD), measured with a 3-m Crowded Lea Symbol chart (catalog no. 2503; Precision Vision, Bloomington, IL), was 
0.20 logMAR. Normal subjects were recruited from referrals for suspected amblyopia that was disproved after identical examination, and also through local newspaper announcements. All normal participants were required to have been born fewer than 15 days prematurely, to have had a history of normal birth weight and, to be free of systemic disease. Furthermore, undilated ophthalmic examination by a fully qualified ophthalmologist (SC, AC) confirmed no ocular disease in addition to bifoveal fixation (4-D base-out test, 3/4 criterion) and, in those able to perform it, normal stereothreshold (Frisby near stereotest: two-up, one-down staircase). Potentially normal recruits found to have greater than 1 D of sphere (hyperopia or myopia) and/or greater than 0.5 D of cylinder (hyperopia or myopia) on manifest retinoscopy (n = 2) were subjected to cycloplegic refraction, corrected with full spectacle prescription, and confirmed to surpass normality criteria before recruitment. Only data collected after participants became accustomed to their spectacles were included in the analysis.
The entire study group consisted of 32 amblyopes of mean age 6.44 years (range, 3.63–11.90 ± 1.94; anisometropic, 15; strabismic, 6; anisostrabismic, 11) and 11 normal subjects of mean age 6.69 years (range, 4.84 –8.58 ± 1.12). Slightly more difficulty was encountered in recruiting younger normal subjects, and although there was a difference in age range between the two groups, it did not reach statistical significance (P = 0.69). Mean logMAR acuity of the amblyopic eyes was 0.50 ± 0.31 (SD), of the fellow eyes was 0.13 ± 0.15, of the normal right eye was 0.10 ± 0.09 and, of the normal left eye was 0.06 ± 0.13. The mean spherical equivalent refraction of the amblyopic eye was +4.66 ± 3.44 D (SD) and of the fellow eye was +2.83 ± 2.42 D. Only two children in the normal group needed spectacles (mean spherical equivalent right, +4.94 D, and left, +4.63 D) but each achieved normal acuity (mean logMAR of 0.08 in each eye) and stereoacuity (20 seconds of arc) for their age.
Local IRB approval was received. The protocol was in compliance with the tenets of the Declaration of Helsinki, and written, informed consent was obtained.
Study Design
Only participants providing complete data for the test and retest procedure in both eyes contributed data to the study. Data were collected by one author (SC) at each clinic session during the course of a controlled longitudinal study of amblyopia treatment. Most of the participants had more than one clinic session in which test and retest thresholds were collected in each eye, and therefore, in these instances, one record was randomly (computer random number generator) chosen in each patient for analysis. The visual acuity test (first determined threshold for a given eye) and retest (second determined threshold for that eye) procedures were identical and adhered rigidly to the protocol that will be described in the Procedure section. This procedure was based on the ETDRS-Fast method.31 For the younger age group of our participants a modification was necessary to shorten the testing procedure. Monocular acuity thresholds were determined for each eye during the test phase. Each eye was then retested (not necessarily in the same sequence) within a 20-minute period. The threshold determinations for each eye were interleaved with orthoptic testing and breaks as necessary to maintain the interest and cooperation of the child.
At each visit, the acuity was attempted first as a matter of course and without knowledge of previous acuity thresholds. For previously noted reasons32 and reasons related to pediatric cooperation levels, there was limited ability to mask the tester from the test thresholds when conducting the retest thresholds. The tester was fully aware, however, of possible bias in this respect and adhered rigidly to the acuity testing procedure at all times (see the Discussion section for further detail).
Procedure
Best corrected distance acuity was determined with the same ETDRS-type, high-contrast, standardized, retroilluminated chart in the same room throughout. The mean luminance was 266 cd/m2, in compliance with recommendations for the standardization of acuity measurement.4 33 The chart is logarithmically calibrated and, at the specified testing distance, ranges in logMAR from 1.0 to –0.4. The interoptotype distances relate to the size of the symbols, of which there are four: circle, house, heart, and square. The symbols are chosen to be universally recognizable among children, and each differs in only a few details from a circle, so that, beyond the threshold of recognition, all shapes blur to resemble a circle.34 The procedure for test and retest threshold determination was identical and involved recording a threshold determined by line-by-line rather than interpolated35 scoring. As mentioned earlier, line-by-line scoring interprets each line as an individual psychophysical task that the patient must pass to progress. The four-alternative choice design of the Lea chart in this study dictated that the subject surpass 62.5% (halfway between chance and 100%) correct responses for each size optotype (logMAR line) before progressing. Because three of five represents 60% correct, in this study a four of five (80% correct) criterion or better was universal.
The first stage of testing involved a screening stage in which the child was requested to identify one of the central three symbols chosen randomly by the tester, starting from the 1.0 or the 0.9 logMAR line (if the 0.9 line was incorrectly identified, the 1.0 symbol was used). No second chances were given at any time, and the child was expected to guess if uncertain. If the child was unable to identify correctly the first symbol presented from the 1.0-logMAR line, the testing distance was reduced in 1-m increments to a minimum of 1 m from the chart and then to 0.5 m from the chart. Screening progressed in –0.2-logMAR increments (every second line), avoiding the presentation of circles, until an incorrect response was recorded. At this point, screening stopped, and the line containing the previously correct response was tested from left to right or vice versa. A threshold was recorded as the logMAR of the line last correctly (4/4 or 4/5 correct) identified (i.e., the line just above the one missed).
Several measures were used to achieve a reliable threshold in cases in which false starts (child unwilling to cooperate with identifying symbols e.g., due to playfulness or fussiness) were encountered. For example, the design of this chart (catalog no. 2503; Precision Vision) provides space for two additional sets of symbols on either side of the central set (below the 0.50-logMAR level), allowing the tester to obviate any attempts by the child to memorize the symbols. If false starts occurred before the 0.50-logMAR level, the child’s view of the chart was shielded, and the child was misinformed that the entire chart was being changed, a break would be given, and testing would resume when cooperation was again appropriate.
Statistical Analysis
All units of acuity were recorded and analyzed in logMAR notation. The 95% limits of agreement (LOA) as described by Bland and Altman32 were evaluated for the entire group and for each subgroup. Student’s t-test was used to evaluate the differences (difference = test threshold – retest threshold) between eyes and also in groups of participants, to evaluate whether differences were affected by the sequence of testing (i.e., results of first eye chosen to be tested compared with that of the second eye chosen to be tested).
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
The line-by-line method categorizes repeatability in 0.10 logMAR increments starting from 0.00 (no variation). Considering the lack of significant difference in repeatability related either to the sequence of testing or the diagnostic category, the repeatability for the entire group (normal and amblyope) was calculated. A difference of ±0.10 logMAR between test and retest thresholds occurred in 93% of cases, and a difference of ±0.20 logMAR occurred in 7% of cases. No significant difference (P = 0.13) was found between the magnitude of threshold improvement (26/48) versus disimprovement (22/48) on retesting.
In the entire study group, the mean difference between the test threshold and that of the retest was an improvement in threshold of 0.01 logMAR (Table 1) , and the largest difference that occurred was an improvement of 0.05 logMAR (2.5 symbols), neither of which is clinically meaningful.
|
|
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
0.4 logMAR. In the present study and that of Moke et al.,2 both normal subjects and amblyopes were assessed, with no difference found in repeatability between these two groups. In addition, we and Moke et al.2 (as well as Holmes et al.,1 in which optical defocus was not reported as uniformly controlled1 ), found that, in 93% of cases, test and retest thresholds lay within ±0.10 logMAR. This study and the other two (Moke et al.2 and Holmes et al.1 ) used line-by-line scoring.
|
In practical terms, when using line-by-line scoring, however, a 95% LOA of ±0.18 would effectively amount to a threshold for detecting a real difference of 2 lines, and in our sample 100% of test and retest scores lay within ±0.2 logMAR—generally the accepted threshold for the detection of amblyopia and better than the interocular difference threshold specified in a recent large amblyopia study.36
Despite their use of the interpolated scoring method, the reported 95% LOA in Kheterpal et al.23 are actually greater than those reported by us (Table 2) . This most likely reflects the smaller study group and possibly also the effect of optical defocus.16 17 18 Whether the careful control of optical defocus affected our results and those of Moke et al.2 is unclear, since Holmes et al.1 reported similar results in a group of children in which the refractive status of the normal subjects was not specified. Unfortunately, recalculation of our data using the interpolated scoring system is not possible, because the study criteria recorded only line-by-line thresholds from the outset.
Our analysis failed to identify any statistically significant difference in the magnitude of threshold differences between test and retest of the normal and the amblyopic groups. This result is in keeping with those in previous reports in both adult and pediatric groups, which showed either no statistically significant difference in repeatability1 2 14 21 between normal and abnormal groups or, if a difference was found (±0.02 logMAR greater test–retest variability in cataract patients than normal),16 it was not clinically meaningful.
The adequacy of masking in the single observer, single-session data collection design of this study is a possible source of bias. In young pediatric age groups, cooperation is notoriously fickle, and in this study, intervals between acuity thresholds were occupied by conducting other orthoptic assessments rather than allowing the child to leave the examination room for non–clinic-related activity. It was decided also to collect test and retest data within the same session, because there is a high nonattendance rate (24.76% in the research clinic, 19.66% in the nonresearch clinic), thereby reducing the total number of visits required. Despite this, no significant difference was detected between the magnitude of improvement versus disimprovement (P = 0.13) for the entire group of eyes in which a change in threshold occurred (n = 48). A further argument against the possibility of bias in this report is related to a crucial finding in the work (i.e., the magnitude of repeatability is similar across three classes of eyes: normal, fellow, amblyopic). To have achieved this result, one would have had to inflate or deflate or adjust the bias systematically, depending on the group being tested at the time. We did not make such adjustments.
The Lea symbol chart has been shown to have testability superior to that of the HOTV test,10 28 but the question of its reliability, particularly in amblyopic children and also under conditions of appropriate optical correction has been investigated in this study. We have now shown that the Lea symbol chart, scored using a line-by-line method, produces test–retest reliability that is comparable to the HOTV test scored by the same method,1 2 and reliability that is comparable to other reports23 25 in which interpolated scoring was used. Thus, a degree of accuracy comparable to that of the HOTV design can be obtained with a test that has been shown to have superior testability in children. Best corrected acuity scores were measured throughout this study, and therefore the effect of optical defocus on repeatability was not measured, but there is no reason to suspect that it would be any less than that reported by Rosser et al.18 in normal adults. There appeared to be no clinically meaningful difference in repeatability between the normal and amblyopic children tested in this study.
|
Footnotes |
|---|
Submitted for publication May 17, 2005; revised July 25, September 13, and October 25, 2005; December 15, 2005.
Disclosure: S.I. Chen, None; A. Chandna, None; , A.M. Norcia, None; M. Pettet, None; D. Stone, None
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked "advertisement" in accordance with 18 U.S.C. 
1734 solely to indicate this fact.
Corresponding author: Sean I. Chen, Vision Assessment Unit, Department of Ophthalmology, Royal Liverpool Children’s Hospital, Eaton Road, Liverpool L12 2AP, UK; s.chen{at}blueyonder.co.uk.
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
