Comparison of the Variational Bayes Linear Regression Visual Field Test Algorithms and the Swedish Interactive...

To compare 10-2 visual field (VF) test results obtained with the Swedish Interactive Threshold Algorithm (SITA) Standard and those obtained with a family of variational Bayes linear regression (VBLR) VF algorithms-VBLR-VF (standard program), VBLR-VF Fast, and VBLR-VF Fast+-in patients with glaucoma.

Multicenter prospective observational cohort study.

One eye from each of 133 patients with glaucoma.

Among the 133 patients, 43 underwent VBLR-VF and the SITA-Standard, 45 underwent VBLR-VF Fast and the SITA-Standard, and 45 underwent VBLR-VF Fast+ and the SITA-Standard, all of which were performed with the 10-2 test program on the same day. Each test pair was repeated within 6 months in reverse order.

The mean deviation (MD), pattern standard deviation (PSD), pointwise VF sensitivity, test-retest repeatability (root mean square error [RMSE]), and test duration were compared between the SITA-Standard and each VBLR-VF algorithm.

There was no significant difference in MD between the SITA-Standard and VBLR-VF. In contrast, the MD values obtained with VBLR-VF Fast and Fast+ were significantly greater than those obtained with the SITA-Standard by 0.61 to 0.81 dB (P < 0.01). Pattern standard deviation with the VBLR-VF method was 0.62 dB greater than that with the SITA-Standard in the second test (P < 0.01), whereas there was no significant difference in the first test. No significant differences in PSD were detected between VBLR-VF Fast or Fast+ and the SITA-Standard. The pointwise VF sensitivities with VBLR-VF were 0.8 to 0.9 dB lower than those with the SITA-Standard (P < 0.01), whereas 0.2 dB higher values were observed with VBLR-VF Fast and Fast+ (P < 0.05). The RMSE did not significantly differ between any of the VBLR-VF algorithms and the SITA-Standard. Test durations with VBLR-VF, VBLR-VF Fast, and VBLR-VF Fast+ at the first visit were significantly shorter than those with the SITA-Standard by 12.2%, 37.8%, and 46.3%, respectively, and were further reduced at the second visit by 10.5%, 26.0%, and 26.3%, respectively (all P < 0.01).

All the VBLR-VF algorithms substantially reduced the test duration while maintaining repeatability. Significant or nonsignificant minor differences in 10-2 VF sensitivity were observed between the SITA-Standard and the VBLR-VF algorithms.

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.