VFQ-25 With and Without Rasch Scoring Assessed Using Simulated Data in Trials Comparing Laser With Anti-VEGF for...

To evaluate treatment group differences in Rasch-calibrated and non-Rasch-calibrated health-related quality-of-life domain scores and visual acuity (VA) across randomized clinical trials of laser versus anti-vascular endothelial growth factor (VEGF) in proliferative diabetic retinopathy.

We simulated 10,000 randomized controlled trials (RCTs), using random sample-and-replace bootstrapping methods from open-source data obtained from individual patient VA and National Eye Institute Visual Function Questionnaire-25 (VFQ-25) responses in the Diabetic Retinopathy Clinical Research (DRCR) Retina Network Protocol S, comparing ranibizumab with panretinal photocoagulation for proliferative diabetic retinopathy. We calculated Pearson correlation coefficients between treatment group differences in mean VA and mean domain scores and ranked them. We also calculated mean effect sizes for each domain and composite across RCTs.

Correlations between treatment group differences (mean VA vs. mean domain score) in rank order were as follows: Rasch combined, 0.408; Rasch socioemotional, 0.394; original composite, 0.388; near activities, 0.385; Rasch visual function, 0.383; and mental health, 0.380. Correlations were weaker for the remaining 10 domains. The largest effect sizes were found with the original composite, mental health, dependency, and peripheral vision, slightly larger than Rasch combined and Rasch socioemotional scales.

Treatment group differences in Rasch calibrated VFQ-25 domains (visual function, socioemotional, combined) had the highest correlation with differences in VA, although the original multidimensional non-Rasch composite and near activities were similar.

Use of Rasch calibrated scores is recommended for desirable psychometrics, but it may be acceptable to use the raw scored original composite instrument.