Comparing the Rate of Retinal Nerve Fiber Layer and Visual Field Loss as Outcomes in Glaucoma Trials.

PURPOSE

This study aimed to compare the statistical power of structural and visual field (VF) outcomes for randomized clinical trials (RCTs) in glaucoma.

DESIGN

Analysis of retrospectively collected data.

PARTICIPANTS

Eighty-two patients with glaucoma were recruited to a test-retest study, during which up to 10 24-2 Swedish Interactive Thresholding Algorithm (SITA) Standard VF and circumpapillary retinal nerve fiber layer (cpRNFL) Spectralis OCT scans were collected in separate sessions over 3 months.

METHODS

Eyes with at least 3 sessions with a reliable VF (false-positives < 15%) and cpRNFL scan (quality index ≥ 25 dB) were selected (127 eyes, 68 patients) to model the test-retest variability and the structural floor effect. These estimates were combined with a published realistic structure-function progression model from the United Kingdom Glaucoma Treatment Study to simulate longitudinal RCTs (30% neuroprotective effect). Simulations only included data from eyes with early to moderate VF loss (mean deviation [MD], ≥-10 dB, 107 eyes, 65 patients). Simulations were repeated 5000 times to estimate sample size requirements to detect a significant difference (P < 0.05) in the rate of change of MD and average cpRNFL thickness, estimated with a linear mixed-effects model. We also tested the power of a significant outcome with either metric (P < 0.025). A supplementary analysis was performed including eyes with early VF loss only (MD ≥ -6 dB).

MAIN OUTCOME MEASURES

Sample size at 80% power for the linear rate of MD, cpRNFL, and their combination.

RESULTS

At 80% power, the required sample size (patients [95% confidence interval]) was 38% smaller for the MD rate (292 [300-283]) than the cpRNFL rate (470 [481-459]). The sample size for the combined outcome was only marginally smaller than the MD alone (275 [283-268]). The supplementary analysis on eyes with early VF loss showed similar results.

CONCLUSIONS

Using realistic modeling of structure-function progression and test-retest data, MD progression showed higher statistical power cpRFNL as an outcome measure for clinical trials. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.