Pointwise variability of a tablet perimetry application.

Abstract

Purpose : The Melbourne Rapid Fields (MRF) application provides a perimetry test on an iPad tablet. Despite using different test methods to the Humphrey Field Analyzer (HFA) it returns thresholds with strong correlations to HFA 24-2 SITA-standard . What is not clear is:1. how it compares to 24-2 SITA-fast and, 2. whether retest variability between tests is similar.

Methods : Sixty patients (stable glaucoma/glaucoma suspects) with acuity better than 6/12 (20/40) were recruited into a 6-month longitudinal trial with four, 2-monthly visits (0, 2, 4, 6 months). At each clincal visit patients were tested on the eye having the least HFA MD with the MRF and either HFA SITA-fast (n=21, India) or SITA-standard (n=39, Cambridge). Test associations were evaluated using Lin’s Coefficient of Accuracy (CoA) and Bland-Altman methods. Fifty-eight patients attended on at least 3 of the 4 occasions and the outcomes from these tests has been analysed in terms of an average pointwise threshold and standard deviation.

Results : CoA for Mean Deviation (MD) at the 6-month retest was 0.83 [0.75 – 0.91] for MRF vs SITA-standard and 0.82 [0.74 – 0.90] for MRF vs SITA-fast. Test time for MRF was similar to SITA-fast (MRF 4.6±0.1 vs. SITA-fast 4.3±0.2 mins) and both were faster than SITA-standard (6.2 ± 0.1 mins, p<0.001). Repeatability of MRF MD was excellent, with CoA for baseline and 6-month visit being 0.95 [0.93 – 0.97] compared with SITA-fast 0.89 [0.86 – 0.92] and SITA-standard 0.90 [0.85 – 0.94]. The MRF showed significantly (p<0.05) smaller Bland-Altman 95% Limits of Agreement for MD with repeatability being 2 dB for MRF, 2.8 dB for SITA-standard and 3.8 dB for SITA-fast. Our pointwise analysis finds that variability increases with decreasing threshold for all tests and that some 61% of locations do not change at retest when measured with the MRF compared with 16% for SITA-standard and 3% for SITA-fast. However, the MRF returns greater point-wise variability than does either SITA test when change is found.

Conclusions : MRF gives strong correlations with the MD returned by HFA SITA-standard and SITA-fast. As the 3-step Bayes logic of the MRF implements larger ‘steps’ it returns more stable outcomes than does either SITA test. However, when thresholds change, they can show greater variability for the MRF. Retesting (2-4 times) will reduce the deleterious effect that this increased variability can have on finding change.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.