Clinical Effective Dynamic Range and the Measurement Floor of SITA-Faster Visual Field Tests

PURPOSE: To describe the effective dynamic range (defined below) using the Swedish Interactive Thresholding Algorithm (SITA)-Faster perimetry and the effects of reliability. METHODS: Nine thousand six hundred and seventy-two test-retest pairs of the 52 test locations of the 24-2 SITA-Faster visual field tests from 1468 eyes of 748 subjects were analysed. The outcome measures were the number of discernible clinical steps, breakpoints and the measurement floor. Effective dynamic range was first assessed using an established approach of smoothed Loess functions (Method 1) applied to test-retest sensitivity pairs. Method 1 was a geometric approach for estimating the outcome measures. Method 2 used the same Loess functions applied to the data in Bland-Altman form. In Method 3, Gaussian smoothing and multisegmental linear regression were applied to the standard deviation of test-retest pairs as a function of mean sensitivity. Analyses were performed on the total cohort, a false positive rate ≤15% subgroup and a false positive rate 0% subgroup. RESULTS: Method 1 returned four clinical steps with intervals bounded by 37 dB, 27 dB and 20-21 dB (the measurement floor), similar across each reliability condition. Method 2 returned the same number of clinical steps, but with a lower measurement floor (18-19 dB). Method 3 led to one to two additional steps and a floor of 18-21 dB. In general, using a false positive rate of 0% resulted in a relatively lower measurement floor value and more breakpoints in comparison with higher false positive rate tolerances. CONCLUSIONS: Using a conventional reliability criterion of false positive rate ≤15%, SITA-Faster has four to five meaningful clinical steps/intervals and a measurement floor of 18-21 dB, which is slightly higher than estimates from SITA-Standard (likely attributable to the higher sensitivity values and greater variability of SITA-Faster). Below this level, clinicians should consider other perimetric approaches, such as 10-2 and/or testing with a Goldmann size V stimulus.