| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 From the Departments of Ophthalmology and 2 Neurology, University of Iowa, College of Medicine, Veterans Administration Hospital, Iowa City, Iowa; and the 3 Department of Psychology, Eastern Illinois University, Charleston, Illinois.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
METHODS. Twenty-four patients with nonglaucomatous optic neuropathies and 18 patients with a relative homonymous or bitemporal hemianopia were tested with both conventional perimetry (Humphrey 24-2 program) and "back to back" SITA standard tests (SITA 1, SITA 2) to approximate the test time of the FT test conditions. Also, 28 normal subjects between the ages of 20 and 80 were tested with this protocol. The visual field quadrants with the most damage were used to evaluate any fatigue effect (i.e., possible lack of fatigue effect with SITA standard due to the shorter test time) and to compare probability plot data between FT, SITA 1, and SITA 2. Pointwise total and pattern deviation probability plot defects were weighted by degree of significance and summed.
RESULTS. Test times for normal subjects were 45 seconds longer for FT than for the combined test time of SITA 1 + SITA 2. Patients’ test times were 40 seconds longer for hemianopias and 90 seconds longer for optic neuropathies with FT than the combined times for two SITA tests. There were higher sensitivities found with SITA 1 compared with Full Threshold (1.06 dB, P < 0.001) and SITA 2 with Full Threshold (0.73 dB, P < 0.001) in the most damaged quadrant for the optic neuropathy patients; for the hemianopia patients the difference in values were between SITA 1 and Full Threshold (0.96 dB, P = 0.07) and between SITA 2 and Full Threshold (0.11 dB, P = 0.87). The second SITA standard test had lower sensitivity than the first SITA standard test by 0.82 dB in hemianopias and by 0.71 dB in optic neuropathy patients. Analysis of the total and pattern deviation probability plot data showed slightly more defects (number and magnitude) with SITA 1 compared to FT for both groups, but the differences were not statistically significant.
CONCLUSIONS. Sensitivities were higher in patients with hemianopias or optic neuropathies using SITA standard compared with FT by approximately 1 dB. The probability plot comparison suggests SITA standard is at least as good as FT for detection of visual loss in individual examinations. However, efficacy of SITA standard for serial examinations has not yet been evaluated.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
SITA is a method of estimating thresholds and reliability indices that has been optimized for a reduction in test time.1 2 3 4 5 6 7 To accomplish this reduction in test time, information from surrounding test locations is used to compute staircase starting values. This is done by visual field modeling (normal and glaucoma models), using frequency of seeing curves, and by application of a Bayesian posterior probability function. When a predetermined level of uncertainty is reached (error related factor) the 4/2 staircase procedure is interrupted. Test time is further reduced both by test pacing that changes stimulus presentation rate in response to the patient’s reaction time and by a new method of calculating catch trials, mostly from data within the test. SITA is available for the Humphrey Field Analyzer in two versions. SITA standard is analogous to the Full Threshold strategy and uses a double crossing of threshold; SITA fast is analogous to Fastpac and adopts a single crossing of threshold using a 3-dB step. SITA permits a halving of the examination time while maintaining within-test variability at about the same level as the Full Threshold test done with the classic staircase procedure.2
Validation of this new methodology for detection of visual loss has been performed in normals and glaucoma patients.1 3 8 It has been observed that sensitivities are 1 to 2 dB higher than with conventional testing.8 9 10 11 The reason for this is unclear. Although visual fatigue has been suggested, evidence to the contrary exists.8 Shirato and co-workers point out that the test point starting value affects the threshold estimate when using a 4/2 staircase,12 especially in subjects with moderate visual loss. They suggest differences in starting values may contribute to the higher SITA sensitivities. Also, SITA interrupts the staircase procedure when a predetermined level of uncertainty is surpassed. Because starting values are often higher than the true thresholds (the seed point starts at 25 dB), this staircase interruption strategy may account for some of the higher sensitivity found with SITA. Also, confidence limits of normals are of smaller magnitude (require less sensitivity loss to be abnormal) with SITA standard (SITA standard) than with FT. Therefore, although sensitivities are higher, pointwise probability plot analysis appears to be comparable to standard 4/2-staircase testing (Statpac 2).13
SITA was developed for visual field testing in glaucoma patients and has not been validated in other visual disorders, specifically neuro-ophthalmologic ones. Our goal was twofold. First, we aimed to determine whether this test strategy was valid in patients with nonglaucomatous optic neuropathies and hemianopias. Second, we wished to investigate further whether the higher sensitivities found with SITA standard were attributable to less visual fatigue.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Normal subjects were included if they had no history of eye disease except refractive error (no more optical correction than five diopters of sphere or three diopters of cylinder); no history of diabetes mellitus or systemic arterial hypertension; a normal ophthalmologic examination including 20/25 or better Snellen acuity; and normal automated perimetry results (Humphrey Visual Field Analyzer, program 24-2). If a potential normal subject had three or more adjacent abnormal points in a clinically suspicious area at the P < 0.05 level, or two adjacent points abnormal with at least one at the P < 0.01 level, or a mean deviation that fell outside the 95% confidence limit, they were excluded. The included subjects had either undergone a complete eye examination within 12 months before this study or were examined by an ophthalmologist on the day of testing to ensure normal ocular health. The visual field inclusion criteria were based on the conventional perimetry examination done for this study.
The patients were selected by a chart review or were asked to participate as part of a scheduled clinic visit. They all fulfilled the following inclusion criteria: best-corrected acuity of 20/40 or better; objective evidence of an optic neuropathy or hemianopia; ages 20 to 70 years; and acceptable reliability indices (based on manufacturer’s recommendations) on the Humphrey Full Threshold visual field examinations. Patients were excluded if they had any other disease causing visual field loss, including systemic arterial hypertension or diabetes mellitus severe enough to cause retinopathy, media opacity severe enough to interfere with visualization of the optic disc, a history of mental illness, or cancer with metastases.
Visual Testing
All subjects first underwent standard Humphrey 24-2 Full
Threshold testing. We followed the manufacturer’s recommendations and
used a corrective lens when necessary. Care was taken to prevent lens
rim artifact. The 24-2 Full Threshold test was administered to all
normal subjects and to patients who had a stable visual field within 1
year of this study. Normals and optic neuropathy patients had testing
in one eye. Halfway through the Full Threshold test, the normals and
the patients were given a 2-minute rest break. Normals had their right
eye tested, and patients with optic neuropathies had either an involved
eye chosen randomly or one with mild–moderate rather than severe loss.
Only one eye was tested because often the second eye had normal
function. Hemianopia patients had both eyes tested (2 Full Threshold
tests and 4 SITA tests).
SITA standard testing was done after the 24-2 testing as follows. All subjects had two SITA standard examinations in continuity for the eye being tested. A 2-minute rest break was given between tests while the examiner reset the computer (2 minutes is the approximate time needed for resetting for a repeat test). The hemianopia patients had both eyes tested, so we a gave 15-minute rest break after the testing sequence was completed in the first eye before the second eye was tested.13 The order was always Full Threshold, SITA 1, and SITA 2. All patients were experienced in either automated perimetry (34 patients) or Goldmann perimetry (8 patients).
Statistical Analysis
We compared test times for FT, SITA 1, and SITA 2. A repeated
measures analysis of variance (ANOVA) was used to compare sensitivities
among the tests. A 6 x 3 (age, test) split-plot ANOVA was used in
the normals to test for effects of age; a 4 x 3 (zone, test)
repeated measures ANOVA was used to test for eccentric zone effects.
The zones were four areas approximately 6° apart with values from the
blind spot (15, 3 and 15, -3) excluded. Zone 1 was the innermost and
Zone 4 the outermost area.
For statistical analysis of patients, we chose the thresholds from the most damaged quadrant for analyses; we defined this as the quadrant with the lowest sensitivity mean score. This was done because with the hemianopia patients, at least two of the other three quadrants could be normal. For the total and pattern deviation probability plot analyses, results were weighted point-by-point and summed. This was done for all test locations for the optic neuropathy patients and for the involved hemifield in the hemianopia patients. We assigned weighting values to the probability plot results as suggested by Asman and colleagues: a P value of 5% was assigned a value of 2, 2% x5, 1% x10, and 0.5% x20.14 We then summed these weighted values for each patient for a probability score.
Differences among groups were tested for significance with ANOVA. Differences between groups of all test results were interpreted as significant if the probability of their occurrence was less than 0.05.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
|
|
However, when individual subjects were analyzed for a fatigue effect, whether it was a within or between-algorithm effect, the following was apparent. In some patients, the mean deviation of SITA 1 was worse (i.e., more negative) than the MD of SITA 2, indicating that the opposite of the expected fatigue effect was present. This occurred in 8 of 28 normals (27.6%). If the MD of SITA 1 was more negative than that of FT, then again, there would also be no evidence for a fatigue effect in that patient. We found a lack of fatigue effect using these criteria (SITA 2 sensitivity higher than SITA 1 or Full Threshold sensitivity higher than SITA 1) in 9 of 28 normals (32.1%). Thirteen of 28 normals (46.4%) met either of these conditions. For the optic neuropathy patients, lack of fatigue effect was found in 29.2%, 33.3%, and 45.8% of patients meeting one, the other, or either criteria. For the right eyes of the hemianopia patients, these results were: 18%, 50%, and 69% as evidence against a fatigue effect. Therefore, although there is evidence for a between-algorithm fatigue effect, the effect is overshadowed by another process in many of the patients; this process is likely retest variability. In the hemianopia patients, 11 of 18 patients had areas of absolute loss in their hemianopic fields. This took the form of full quadrant loss in 5, partial quadrant(s) loss in 5, and hemifield loss in 1. These areas of absolute loss likely contributed to inability to demonstrate any evidence of a fatigue effect.
We also analyzed the SITA 1 versus SITA 2 data on a point-by-point
basis (SITA 1 test location result minus SITA 2 result). The data for
normals by age are shown in Figures 2
and 3a
. The split-plot ANOVA (6 x 3: age x test) yielded
significant main effects for age [F(5,23) = 6.297,
P < 0.001; power (
= 0.05) = 0.96] and
test [F(2,46) = 57.178, P < 0.001; power
( = 0.05) = 1.00]; the interaction was also significant
[F(10,46) = 2.373, P = 0.023; power ( =
0.05) = 0.61]. The expected small decrease in sensitivity with
age was observed, but there appears to be little effect of age on the
differences. Only for the 60-year-olds were SITA 1 (31.0) and SITA 2
(30.4) significantly different. For the 30-, 40-, 50-, and
60-year-olds, Full Threshold was lower was lower than both SITA 1 and
SITA 2; for the 20-year-olds, Full Threshold was only lower than S1,
and only lower than S2 for the 70-year-olds.
|
|
|
|
= 0.05) = 1.00] and test [F(2,56) = 47.702,
P < 0.001; power ( = 0.05) = 1.00]; the
interaction was also significant [F(6168) = 3.401,
P = 0.003; power ( = 0.05) = 0.812]. The
eccentric zone data are found in Table 2
. Thresholds between the two SITA tests did not differ significantly at
the three innermost zones (SITA 1 had higher sensitivities by 0.12 in
Zone 1, central, 0.23 in Zone 2, and 0.19 in Zone 3), but thresholds on
the Full Threshold test were significantly lower than for both SITA
tests. However, at the outermost eccentricity, Zone 4, thresholds on
the second SITA test were significantly lower by 0.38 dB than
thresholds on the first SITA test. Therefore, the differences in
sensitivity between the two tests were greatest in the periphery.
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Patients with hemianopias present a different problem. Here, defects nearly always respect the vertical meridian rather than fitting nerve fiber bundle-like patterns. We were much more concerned that SITA standard might have some problems in estimating thresholds in patients with hemianopic defects. Again, we found no evidence that neurologic defects were missed by SITA standard.
We believe SITA standard worked well in detecting neurologic defects because the basic SITA model that estimates threshold (the normal and glaucoma model in conjunction with frequency of seeing curve data) is likely similar to a model based on neurologic defects. Frequency of seeing curve data does not appear to be disease specific, rather it varies with threshold.15 Also, the use of correlations between test locations, expected to be present in a glaucomatous visual field, does not appear to be a major factor in the algorithm because we did not encounter any instances of loss of definition of hemianopic defects due to glaucoma-related correlations across the vertical midline.
It is well established that SITA standard testing results in slightly higher sensitivity estimations than Full Threshold testing. This increase in sensitivity with SITA standard has been found to be independent of age16 and does not appear to be dependent on test point position.8 Its relationship to test point threshold has been studied with conflicting results.11 Table 3 summarizes these differences in sensitivity reported in normals and glaucoma patients. SITA standard testing resulted in higher sensitivities in all comparisons. In normals, the range is from 0.8 to 1.6 dB with a mean of 1.2 dB. In glaucoma patients, the range is from 1.0 to 2.5 dB with a mean of 1.3 dB. Our 0.44 dB (0.24 dB for normals, 0.57 dB for optic neuropathy, and 0.05 dB for hemianopia patients) increase with SITA 1 compared to Full Threshold is considerably less than these other studies. The difference might relate in part to the procedure in our study to do Full Threshold testing first. A small fatigue effect is possible, but the patients had at least a 15-minute break between the two tests. Additionally, the small difference in hemianopia patients may be attributable to the frequent presence of absolute sensitivity loss.8
|
The goal of the SITA algorithms is to reduce test time without sacrificing sensitivity for detection of defects or without raising retest variability. The studies to date reporting test-time reduction are summarized in Table 3 . Wild et al.11 found a 53% shorter test time for SITA standard than for Full Threshold testing. Examination time duration increased with visual field damage. Our results are similar to these other studies with a 54% reduction of test time with SITA standard in normals; a 46% reduction in test time with the SITA standard method for the optic neuropathy patients; and, a 40% reduction in test time for hemianopia patients.
Bengtsson and Heijl investigated intersubject variability and the normal limits of SITA standard in 330 eyes of 330 normal subjects.9 They found a 31% decrease in intersubject variance with SITA standard compared with Full Threshold testing. This translated into a tightening of the confidence limits of normality from 9 to 29%; that is, the average sensitivity depressions needed to reach the 5% and smaller limits were modestly reduced. This finding is in concert with our observation that there is not much difference when comparing probability plot data between SITA standard and Full Threshold results in patients with neuro-ophthalmologic disorders.
Wild and colleagues16 showed the pointwise between algorithm, between subject variability was lower with SITA standard. The mean of the ratio of the SD for SITA standard compared with Full Threshold for the 24-2 test locations was 0.93, demonstrating lower variability for SITA standard. Shirato and coworkers8 found the test–retest variability was slightly lower with SITA standard than Full Threshold (2.9% vs. 3.4%). Our results, displayed graphically in Figure 5 , also show similar variability of Full Threshold and SITA results. Wild and co-workers suggest that an interaction of the error related factor, the model for the normal island of vision and post-processing computations artificially smooth the visual field results, thereby lowering the between-subject variability.16
Bengtsson and Heijl compared probability plot results between Full Threshold and SITA strategies.13 They tested one eye of 44 glaucoma patients four times with each strategy. Another 21 eyes of 21 normal subjects were examined once with each strategy. The magnitude of field loss as defined by the Statpac Mean Deviation did not differ between the strategies. However, in the glaucoma patients, SITA showed a slightly larger number of significantly depressed points in the probability maps compared to the Full Threshold strategy. They concluded that SITA standard (and also SITA fast) identified at least as many significant glaucomatous field defects as the Full Threshold strategy. Wild and co-workers compared the variability of the SITA algorithm’s test results with those of Fastpac and Full Threshold testing in patients with primary open angle glaucoma.11 They studied one eye from each of 29 patients. Unlike our study, the total deviation probability plot analysis showed a statistically greater number of defects with SITA standard than with Full Threshold testing. Our findings, though, agree with these studies and extend the use of total and pattern deviation plot results to patients with nonglaucomatous optic neuropathies and hemianopias.
In aggregate, it is clear that SITA standard accomplishes its goal of a substantial reduction of test time in normals, glaucoma, and neuro-ophthalmology patients. The reduction in time is on the order of 50%. The test time is longer in patients than normals and the relative time savings is less, probably due to more error in estimating starting values in damaged visual fields. In the development of SITA standard, the error-related factor was chosen so that SITA standard would have a similar variability to Full Threshold testing. It appears that SITA standard has about the same retest variability compared to Full Threshold testing3 16 and that this algorithm does not solve this critical problem of variability increasing exponentially with decreasing threshold.17 18 19 20 Our results agree with those findings.
It is also clear that with SITA standard compared to Full Threshold testing, mean sensitivities are approximately 1 to 2 dB higher in normals, glaucoma subjects, and patients with other optic neuropathies and hemianopias. However, the mechanism of this difference is complex. Although some of this difference is due to visual fatigue, it is also likely that the SITA method, which interrupts the 4/2 staircase procedure when the error-related factor is reached, also contributes. However, from a clinical standpoint, the absolute differences are small and appear to be accounted for when comparisons are made to normative data from the two tests using the probability plot results.
The issue of fatigue is further complicated by the learning effect that works in the opposite direction with regard to sensitivity. The learning effect is the improvement in sensitivity found with repeated testing. In normal subjects, it is in the range of 1 to 2 dB21 There are three different patterns of learning. In some subjects there is little, if any, learning effect. In others, the learning effect is primarily between the first and second visual fields. In some, there is a gradual learning effect over many visual fields. A substantial effect is found in a minority and is primarily between the first and second examinations but can continue for many with most of the effect in the first five examinations,21 and the effect may disappear after 5 months.23 In our study, this learning effect likely had some contribution tending to offset fatigue effects.
This study has several potential drawbacks. We were unable to reset the computer to do the second SITA test in less than about 2 minutes. This duration of a rest break likely reduced the effect of visual fatigue. Also, the high variability in patients with visual field damage made calculations of amount of sensitivity difference between the second SITA test and the Full Threshold results problematic. Lastly, the confounding influence of a learning effect likely interferes with our analysis of the effects of visual fatigue.
SITA represents a major advance in clinical perimetry. The substantial reduction in test time coupled with no apparent degradation of detection of visual loss is welcome. However, it remains unknown whether SITA will be as good or better than Full Threshold testing for detecting visual field progression. Studies addressing this issue are needed before full adoption of SITA takes place.
|
Footnotes |
|---|
Submitted for publication March 24, 2000; revised September 28, 2000; accepted October 20, 2000.
Commercial relationships policy: N.
Corresponding author: Michael Wall, Department of Neurology, University of Iowa College of Medicine, 200 Hawkins Drive #2007 RCP, Iowa City, IA 52242-1053. michael-wall{at}uiowa.edu
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. Boden, J. Pascual, F. A. Medeiros, M. Aihara, R. N. Weinreb, and P. A. Sample Relationship of SITA and Full-Threshold Standard Perimetry to Frequency-Doubling Technology Perimetry in Glaucoma Invest. Ophthalmol. Vis. Sci., July 1, 2005; 46(7): 2433 - 2439. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Turpin, A. M. McKendrick, C. A. Johnson, and A. J. Vingrys Properties of Perimetric Threshold Estimates from Full Threshold, ZEST, and SITA-like Strategies, as Determined by Computer Simulation Invest. Ophthalmol. Vis. Sci., November 1, 2003; 44(11): 4787 - 4795. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
