Abstract
Purpose :
Selecting the right endpoints, sample size and study duration for clinical trials in orphan diseases is challenging. Trials are most efficient when sufficient power is reached with the shortest study duration and smallest sample size. Incorrectly chosen endpoints may lead to failure of trials. As endpoints should reflect clinical benefit, visual acuity (VA) is currently one of the few accepted endpoints in clinical trials in ophthalmology. With upcoming new trials for Stargardt disease (STGD1), we conducted a natural history study on the natural course of VA in a large cohort of STGD1 patients.
Methods :
We derived VA data from the medical records of 317 STGD1 patients. In total, we collected data from 2295 measurements. Daily decline of monocular VA was calculated using linear regression analysis. Subgroups were made based on ABCA4 genotype severity. Inter-eye correlations were calculated for baseline and follow-up (2 and 4 year) using Lin’s concordance correlation coefficient (LCCC). Sample size calculations were performed based on the calculated daily decline of VA.
Results :
VA decline was fastest in patients with VA 0.22 LogMAR (20/32 Snellen) (2.71*10-3±0.47*10-3 LogMAR/day), 0.30 LogMAR (20/40 Snellen) (3.95*10-3±0.50*10-3 LogMAR/day) and 0.40 LogMAR (20/50 Snellen) (2.99*10-3±0.42*10-3 LogMAR/day). Maximum rate of VA decline did not differ significantly between ABCA4 genotype subgroups (p=0.062). At baseline, the inter-eye correlation was substantial for patients with two severe ABCA4 mutations (LCCC=0.966), but was poor (<0.90) for other genotypes, as well for follow-up. When choosing VA as a primary endpoint in a 3-year clinical trial and assuming a therapeutic effect size of 80% in inhibiting disease progression, the smallest sample size with 80% power could be reached when including patients with VA 0.22 LogMAR (11 patients per study arm).
Conclusions :
Patients with a VA between 20/32 and 20/50 Snellen have the most rapid decline of VA and may therefore be the most suitable for clinical trials with small population size. Since we found VA poorly correlated between a patient’s eyes, a study design in which a patient’s second eye serves as control seems suboptimal. In addition to analysis of VA, natural history data on other potential endpoints such as atrophy growth and microperimetry should be analyzed before the best study design for clinical trials can be determined.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.