Abstract
Purpose :
To guide the appropriate design and statistical analysis for future randomized clinical trials (RCTs) for dry eye disease (DED), we performed a systemic review of primary result papers from the RCTs for DED.
Methods :
We searched the primary result papers published between 2013-2023 in 9 journals (Ophthalmology, Cornea,Ocular Surface, JAMA Ophthalmol, Am J Ophthalmol, Br J Ophthalmol, NEJM, Eye, Eye & Contact Lens). Two authors (SZ, AZ) independently extracted data including number of study eyes (e.g., one-eye design or two-eye design), treatment and follow-up, control group, number of arms, severity of DED, masking, run-in period, sample size and power, primary and secondary outcomes, correction for multiple comparisons and inter-eye correlation in DED signs, and missing data. The discrepancies between two authors were adjudicated by the senior author (GSY).
Results :
Among 45 RCTs for DED, 56% used one-eye design, 40% used two-eye design, and 4% used paired design (Table 1). These trials primarily tested drugs (82%) that were administered locally (82%) for a median treatment length of 84 days (range: 14-360 days). Placebo (36%), vehicle (29%) or active control (29%) were used as controls. The Sample size justification was provided in 30 (67%) trials with median sample size 218 participants (range: 16-840) and 19 (42%) trials using a statistical power 90% or higher. The top primary outcome of DED symptoms and signs were OSDI (38%) and corneal staining score (42%) respectively, while the top secondary outcomes of DED symptoms and signs were eye dryness score (38%) and corneal staining score (49%) respectively (Table 2). Statistical correction for multiple comparisons was utilized in 5 (50%) trials with multiple arms, 6 (45%) trials with multiple primary outcomes and 10 (27%) trials with multiple secondary outcomes. Missing data in primary outcome occurred in median of 5% participants (range: 0-52%), and 41% trials applied statistical methods for missing data. Among 20 trials that used two-eye design, 35% trials accounted for inter-eye correlation in analyzing DED sign data.
Conclusions :
Dry eye treatment trials varied in design and statistical analysis. Future trials can be improved by properly choosing the study eye(s), sample size, primary and secondary outcomes and correcting for multiplicity and inter-eye correlation.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.