Abstract
Purpose:
In glaucoma studies, mean IOP is often the primary outcome used as evidence of efficacy during therapeutic intervention. The objective of this study is to explore the incremental clinical utility from diurnal IOP measurements versus reduced IOP sampling strategies. Specifically, this study aimed to determine to what extent using a single random IOP measurement (SRM) instead of diurnal (3) measurements (DM), affects estimates of group mean IOP and proportion of patients above or below a target IOP.
Methods:
IOP data were used from subjects enrolled in the COMPASS FDA clinical study of the CyPass suprachoroidal Micro-Stent. Sets of data, representing varying “study sizes,” were generated by sampling with replacement from a database of 470 records with complete (3 measures) diurnal IOP data. Resampling was performed 2000 times for each study size (range: 25-1000). The average bias for the point estimate of the sample mean using a SRM compared with using DM was examined. The confidence interval of the point estimate of the IOP mean from each method was also compared. In addition to resampling, the base sample was used to determine differences in outcomes when a SRM vs DM based on various cut points (e.g., IOP <21). All programming was performed in SAS version 9.2.
Results:
For all sample sizes tested, there was no consistent directional difference in mean IOP using SRM or DM. The confidence interval around the mean IOP estimate was reduced with increasing sample size for both endpoints. For study sizes of 50, 100, 300 and 500 patients, 95% of the time the SRM was within 0.50, 0.36, 0.21 and 0.16 mmHg of the DM. The DM estimates fewer patients reaching an IOP target than SRM, for targets ≤ 18-25 mmHg and ≥ 28 mmHg and greater.
Conclusions:
Using a SRM can accurately estimate a group’s mean IOP from multiple measurements. The CI of the group mean IOP estimate is reduced, more precise, with larger sample sizes. In analyses of proportions of patients reaching an IOP target (IOP ≤ or ≥ a target level), using SRM will overestimate the proportion of patients that reach or exceed a target. Depending on outcomes of interest, it may be practical in future studies to use a SRM to approximate the DM.
Keywords: 568 intraocular pressure •
468 clinical research methodology •
458 circadian rhythms