You are on page 1of 4

Comparing significance and magnitude of glaucomatous visual field defects using the SITA and Full Threshold strategies

Boel Bengtsson and Anders Heijl

Department of Ophthalmology, Malmo University Hospital, Malmo , Sweden

ABSTRACT. Purpose: To evaluate and compare visual field test results as presented by the Statpac interpretation tools in tests obtained with the new short SITA Standard and the even shorter SITA Fast strategies to the traditional Humphrey Full Threshold strategy. Subjects and Methods: One eye of each of 44 glaucoma patients was examined four times with each of the Humphrey SITA Standard, SITA Fast, and Full Threshold strategies. Another 21 eyes of 21 normal subjects had one eye examined once with each of the three strategies. Results: Average light sensitivity was highest with the shortest SITA Fast strategy and lowest with the longest Full Threshold strategy. Magnitude of field loss as defined by the Statpac Mean Deviation (MD) did not differ between the three strategies. In the glaucoma patients, both SITA strategies showed larger number of significantly depressed points in Statpac probability maps than the Full Threshold strategy. In the normal subjects SITA Standard showed more significantly depressed points, close to the statistically expected number, at the lowest probability level (p5%) than both Full Threshold and SITA Fast. At higher probability levels (p1%) SITA Standard and Full Threshold showed similar numbers of significantly depressed points. Conclusion: Both SITA Standard and SITA Fast identified at least as much significant glaucomatous field loss as the Full Threshold using the Statpac interpretation tools.
Key words: visual eld strategies threshold perimetry probability maps clinical testing glaucoma. Acta Ophthalmol. Scand. 1999: 77: 143146
Copyright c Acta Ophthalmol Scand 1999. ISSN 1395-3907

e have developed two new perimetric threshold strategies aimed at reducing test time without decreasing quality of test results as compared to traditional threshold tests. SITA Standard (Bengtsson et al. 1997a) was designed to be comparable to the Humphrey Full Threshold (Heijl 1985; Haley 1986) in terms of accuracy, and SITA Fast (Bengtsson & Heijl 1998a) to be similar to the Humphrey Fastpac strategy. Clinical testing has shown that the SITA strategies have met our goals. Test times

with SITA were reduced by nearly 50% compared to the strategies they were intended to replace, while reproducibility was similar or slightly better with SITA (Bengtsson et al. 1998; Bengtsson & Heijl 1998a; 1998b; Cunliffe et al. 1998; Tsuji et al. 1998). The greyscale printouts of SITA elds often appear lighter grey than those of Full Threshold and Fastpac, and glaucomatous visual eld defects may appear less pronounced in such representations of SITA tests as compared to the traditional

threshold programs. This is most obvious in results from the shortest SITA Fast strategy. In earlier studies we have seen that SITA elds in general have slightly higher light sensitivities than Full Threshold. The largest differences were seen in the upper tail of the threshold distribution, i.e. in the number of points with threshold values 20 dB (Bengtsson & Heijl 1998a; 1998b). Early glaucomatous visual eld loss is often characterised by areas of reduced light sensitivity. It is often difcult to interpret the importance of reduced light sensitivity when using simple numerical threshold maps or greyscale representations of such values alone. Probability maps are very useful and commonly used for early detection of visual eld defects (Heijl et al. 1989) because they quantify and account for the way in which normal variability depends upon test point location (Heijl et al. 1987a). Thus, to be signicant, the numerical reduction of sensitivity must exceed the variability seen among normal subjects at a particular test point. Normal variability is smaller in both SITA strategies than in the traditional Full Threshold test (Bengtsson & Heijl 1999). Therefore SITA probability maps are able to single out pathologically based reductions in light sensitivity which would be within normal variability when using the Full Threshold strategy. The aim of the current paper was to compare size and signicance of visual eld loss detected when using the two new SITA strategies with that found when using the traditional Full Threshold strategy. The interpretation tool used for making this comparison was the Statpac (Heijl et al. 1987b; Bengtsson & Heijl 1999) normative signicance limits appropriate to each strategy.


Table 1. Average individual differential light sensitivity (dB). Full ThresSITA SITA hold Standard Fast Normal subjects Glaucoma patients 30.3 18.0 31.3 19.2 31.7 20.0

Table 2. Glaucoma patients (n44) tested at 4 different visits with each of the Full Threshold, SITA Standard and SITA Fast strategies using the 302 test point pattern. Average percentage of Pattern Deviation points signicant at each visit. Visit . strategy Full Threshold SITA Standard SITA Fast Full Threshold SITA Standard SITA Fast Full Threshold SITA Standard SITA Fast Full Threshold SITA Standard SITA Fast Signicance of sensitivity depression 5% 5% 5% 2% 2% 2% 1% 1% 1% 0.5% 0.5% 0.5%

1 37 41 42 31 35 35 26 30 30 21 26 25

2 35 42 42 29 36 35 24 31 30 20 26 24

3 36 41 40 30 35 34 25 31 30 21 27 23

4 37 41 42 30 35 34 25 30 30 21 26 24

Mean of all visits 36 41 41 30 35 34 25 31 30 21 26 24

Subjects and tests

One group of 44 perimetrically experienced glaucoma patients was tested four times each with the SITA Standard, SITA Fast, and Full Threshold strategies using the 30-2 test point pattern. All patients gave informed consent. One eye of each subject was tested. Among the 44 subjects were 19 men and 25 women with a mean age of 71 years (range: 5981). Test order between strategies was randomised at the rst visit and then repeated at the following visits. Tests obtained at the same visit were separated by at least 15 minutes. Intervals between visits were 7 days in most patients, but never shorter than 2 or longer than 15 days. The glaucoma diagnosis was based on previous visual eld test results and/or typical glaucomatous optic disc ndings, e.g. notches or cupping reaching the disc margin. Eyes with early concomitant cataract were accepted, but visual acuity had to be 0.7 or better. MD values ranged from 21.43 dB to 0.71 dB (mean 9.15 dB) in Full Threshold tests. Patients on miotics (n13) were tested at approximately the same time of the day at each visit. Another group of 21 healthy volun-

teers, all with experience of automated threshold perimetry, was examined once with each of the SITA Standard, SITA Fast, and Full Threshold strategies. Fifteen women and six men participated, mean age was 40 years (range: 2956). One eye of each subject was used. All tests were obtained during the same session, allowing at least 15 minutes rest between examinations. Test order between strategies was randomised.

Methods of analysis
Analyses of test results aimed at comparing these three strategies in terms of encountered differential light sensitivity

values, and magnitude of eld loss as dened by MD-values, and numbers of signicantly depressed test points in Pattern Deviation probability maps. We chose Pattern Deviation probability maps, as these maps are more robust tools for identication of glaucomatous eld loss, less inuenced by miosis and cataract than those based on Total Deviations maps (Statpac users guide, Bengtsson et al. 1997b). All test points in the 302 pattern were included in the analyses except for the two points located in the blind spot area. Individual average threshold values and MD-values were compared using ANOVA. In the glaucoma group, where repeated measurements were obtained, within patient and within strategy

Fig. 1. Number of signicantly depressed points at p0.5% vs. average threshold value. A: SITA Standard Full Threshold. B: SITA Fast Full Threshold. Larger differences in number of signicantly depressed points were seen between strategies in eyes with similar average threshold values, while similar number of depressed points were seen in eyes with larger differences in threshold values.


variability was taken into account in the ANOVA model. Numbers of signicantly depressed points were compared between strategies using contingency tables and chi-square tests for independence. Differences between SITA Standard and Full Threshold in the number of signicantly depressed points at the 0.5% level in Pattern Deviation probability maps were compared to differences in threshold values for each subject using linear regression. The same comparison was performed between SITA Fast and Full Threshold tests. The differences were calculated as averages of the four tests obtained with each strategy in each glaucoma patient.

Light sensitivities differed signicantly between strategies in both the glaucoma and the normal group (p0.0001). Differential light sensitivities were higher, i.e. dimmer stimuli were seen, when measured with the shortest strategy, SITA Fast, while the longest Full Threshold test gave lowest sensitivities (Table 1). Magnitude of glaucomatous eld loss, as described by MD values, was similar with the three different strategies (differences were rejected at p0.12). Average SITA Standard MD value was 8.90 dB, ranging from 9.01 dB to 8.77 dB over the four visits. SITA Fast had an average MD of 8.55 dB, range between visits: 8.71 dB to 8.41 dB. Full Thresholds tests had 8.87 dB, range between visits: 9.13 dB to 8.61 dB. MD-values were also similar in the normal eyes (p0.38 when testing for differences); SITA Standard had an average MD-value of 0.99 dB, SITA Fast had 1.01 dB, and Full Threshold showed 0.67 dB. In glaucoma subjects numbers of signicantly depressed points were on the average higher in both SITA Standard and SITA Fast tests when compared to Full Threshold (p0.0001), at all signicance levels. No signicant differences were seen between SITA Fast and SITA Standard. Numbers of signicantly depressed points at each visit were similar (Table 2). In normal subjects, the average percentage of points showing signicant reduction of light sensitivity at the p5% level in Pattern Deviation probability maps were fewer than expected (p0.001) both with the shortest, SITA Fast, and the longest, Full Threshold test, showing 3.3% and

Fig. 2. Test results obtained in one of the glaucoma patients included in current study. Differences between strategies are seen in greyscale representation of threshold values, but defects appear much more similar in Pattern Deviation probability maps.

2.1% such points, respectively. SITA Standard showed 4.2% signicantly depressed points, which was closer to the expected value of 5%. Differences between strategies at this probability-level were also signicant (p0.001). Number of points showing signicant depressions at the p1% level, were 0.6% with SITA Standard, 0.1% in SITA Fast, and 0.5% when using the Full Threshold test. These differences between strategies at this probability level were not signicant (p0.07).

Those glaucoma patients who showed larger number of signicantly depressed points in SITA tests than in Full Threshold tests had similar mean light sensitivity, while the number of signicant points were similar with SITA and Full Threshold in patients with higher light sensitivity on the average with the SITA strategies. The dependence on mean sensitivity was stronger between SITA Standard and Full Threshold (p0.0001), Fig. 1A, than between SITA Fast and


Full Threshold (p0.158, p0.0018 if excluding one outlier), Fig. 1B.

In the current study both mean differential light sensitivities and number of signicantly depressed test point locations differed signicantly between test strategies, but in opposite directions. The magnitude of eld loss expressed in dB was similar. Highest light sensitivity was measured when using that strategy which produced the shortest tests, and light sensitivity was lowest when using the slowest strategy. The signicance of measured eld defects varied between test algorithms in a different way. In the glaucoma patients, more signicant visual eld defects were identied with each of the SITA strategies than with the Full Threshold strategy. To be signicant a depression has to exceed the normal random variability. Normal random variability is smaller at almost all test point locations with both the SITA Standard and the SITA Fast as compared to the Full Threshold strategy (Bengtsson & Heijl 1999). In normal subjects one would have expected similar numbers of signicantly depressed points with all strategies, and that those numbers also should be close to the expected ones. Both the SITA Fast and the traditional Full Threshold, however, had signicantly lower number of points depressed at the mildest probability level (p5%), while SITA Standard was close to the expected. This result is difcult to explain. One could, of course, argue that the normal subjects represented an experienced elite normal material performing better than average population based normal material. But even then the results should be similar at all probability levels with all strategies, also at the p5% SITA Standard level. Such systematic differences between centres are regularly found in multi-centre studies. The SITA Statpac multi-centre data collection is no exception (Bengtsson & Heijl 1999). If applying normal limits for Full Threshold based on the same normal subjects as the SITA limits on our 21 normal subjects, we still have fewer signicantly depressed points than expected with the Full Threshold strategy, 3.3% at p5% level. Thus, the difference between SITA Standard and Full Threshold in number of signicantly depressed points cannot be explained by normal limits based on different ma-

terials. No important differences in number of signicantly depressed points at a higher signicance level, p1%, were seen between strategies. In the glaucoma group magnitudes of eld loss as described by MD values were as expected, similar with the three strategies. MD values are the mean difference of encountered thresholds values and age-corrected normal values. The MD values correct for general differences in height of the visual eld and should be similar for all strategies. Thus the SITA strategies are associated with higher light sensitivities than Full Threshold tests. This could give an impression that the SITA strategies are less sensitive to detect glaucomatous visual eld loss. This is probably due to reduced visual fatigue when using the shorter strategies, and thereby in SITA larger inuence of the a priori visual eld model on the threshold estimate. However, in the current study they detected slightly but signicantly more glaucoma eld defects as the traditional more time-consuming Full Threshold test when using Statpac probability maps, Fig. 2. This is explained by narrower normal limits, which in turn could be explained by reduced visual fatigue in shorter SITA tests and/or higher accuracy in the SITA test algorithms as compared to the traditional longer strategy. The sensitivity and specicity of different test algorithms are affected in probability maps more than in greyscale printouts. Comparisons between results obtained with different tests are therefore easier with probability maps than with numerical greyscale printouts and the task of interpreting different visual elds would be facilitated if probability maps were available for all tests.

This study was supported by grants from Malmo University Hospital and by the Ja rnhardt foundation. This paper was presented in part at the International Perimetric Society Meeting 1998 in Gardone Riviera, Italy.

with manifest and suspect glaucoma. Acta Ophthalmol Scand 76: 268272. Bengtsson B & Heijl A (1999): Inter-subject variability and normal limits of the SITA Standard, SITA Fast, and the Humphrey Full Threshold computerized perimetric strategies, SITA STATPAC. Acta Ophthalmol Scand 77: 125129. Bengtsson B, Heijl A & Olsson J (1998): Evaluation of a new threshold strategy, SITA, in normal subjects. Acta Ophthalmol Scand 76: 165169. Bengtsson B, Lindgren A, Heijl A & Lindgren G (1997b): Perimetric probability maps to separate change caused by glaucoma from that caused by cataract. Acta Ophthalmol Scand 75: 184188. Bengtsson B, Olsson J, Heijl A & Rootze n H (1997a): A new generation of algorithms for computerized threshold perimetry, SITA. Acta Ophthalmol Scand 75: 368375. Cunliffe IA, Pacey IE, Wild JM, Sedgewick G & ONeill (1998): The validity and reproducibility of the SITA Standard and SITA Fast threshold algorithms in glaucoma. Abstract, Invest Ophthalmol Vis Sci vol 39, ARVO suppl, s: 26. Haley MJ (1986): The Field Analyzer primer. Allergan Humphrey, San Leandro, CA. Heijl A (1985): The Humphrey Field Analyzer, construction and concepts. In Heijl & Greve (eds.) Documenta Ophthalmol Proc series 42: 7784. Proceedings of the 6th International Perimetric Society Meeting 1984. Junk Publishers, Dordrecht. Heijl A, Lindgren G & Olsson J (1987a): Normal variability of static threshold values across the central visual eld. Arch Ophthalmol 105: 15441549. Heijl A, Lindgren G & Olsson J (1987b): A package for statistical analysis of computerized visual elds. In Greve & Heijl (ed.) Documenta Ophthalmol Proc series 49: 153168. Proceedings of the 7th International Perimetic Society Meeting 1986, Junk Publishers, Dordrecht. Heijl A, Lindgren G, Olsson J & sman P (1989): Visual eld interpretation with empirical probability maps. Arch Ophthalmol 107: 204208. Statpac users guide (1989). Allergan Humphrey, San Leandro, California. Tsuji A, Inazumi K, Yamamoto T & Kitazawa Y (1998): Evaluation of the Swedish Interactive Threshold Algorithm, a new thresholding algorithm, of the Humphrey eld analyzer in normal subjects. Nippon Ganka Gakki Zasshi 102: 359364. Received on September 29th, 1998. Accepted on November 22nd, 1998. Corresponding author: Boel Bengtsson Department of Ophthalmology Malmo University Hospital S-20502 Malmo Sweden Fax: 46 40336212 E-mail: boel.bengtsson/

Bengtsson B & Heijl A (1998a): SITA Fast, a new rapid perimetric threshold test. Description of methods and evaluation in patients with manifest and suspect glaucoma. Acta Ophthalmol Scand 76: 431437. Bengtsson B & Heijl A (1998b): Evaluation of a new threshold strategy, SITA, in patients