Measurement error of visual field tests in glaucoma.

AIM Psychophysical strategies designed for clinical visual field testing produce rapid estimates of threshold with relatively few stimulus presentations and so represent a trade-off between test quality and efficiency. The aim of this study was to determine the measurement error of a staircase algorithm similar to full threshold with standard automated perimetry (SAP) and frequency doubling perimetry (FDP) in glaucoma patients. METHODS Seven patients with early open angle glaucoma (OAG) were prospectively recruited. All were experienced in laboratory based psychophysics. Three matched test locations were examined with SAP (externally driven Humphrey field analyser) and FDP (CRT) in a single arbitrarily selected eye of each subject. Each location was tested twice with a 4-2-2 dB staircase strategy, similar to full threshold, and then with the method of constant stimuli (MOCS). Accuracy (threshold estimation error) was quantified by determination of differences between "true" threshold measurements made by MOCS and single staircase threshold estimates. Precision (repeatability) was quantified by the differences between repeated staircase threshold estimates. RESULTS Precision was relatively high for both tests, although higher for FDP than SAP at depressed sensitivity levels. The staircase strategy significantly underestimated threshold sensitivity for both test types, with the mean difference (95% CI) between staircase and MOCS thresholds being 4.48 dB (2.35 to 7.32) and 1.35 dB (0.56 to 1.73) for SAP and FDP respectively. Agreement levels (weighted kappa) between MOCS and staircase thresholds were found to be 0.48 for SAP and 0.85 for FDP. Although this "bias" appeared constant for FDP across all sensitivity levels, this was not the case for SAP where accuracy decreased at lower sensitivity levels. CONCLUSION Estimations of threshold sensitivity made using staircase strategies common to clinical visual field test instrumentation are associated with varying degrees of measurement error according to visual field test type and sensitivity. In particular, SAP significantly overestimates the "true" level of sensitivity, particularly in damaged areas of the visual field, suggesting that clinical data of this type should be interpreted with caution.