Abstract
Purpose: :
To compare definitions used to determine outcome in studies of amblyopia treatment
Methods: :
Using data from completed amblyopia clinical trials conducted by the Pediatric Eye Disease Investigator Group to develop a statistical model, we simulated visual acuity data for clinical trials of treatment for moderate amblyopia over a range of ages and treatment effects and compared type I and II error rates for 8 different definitions of treatment success that have been used or proposed as outcome measures for treatment studies of amblyopia.
Results: :
Of the continuous outcome measures considered (lines improved, proportion of deficit corrected (PDC), and intraocular difference (IOD)), lines improved had good statistical properties across the range of scenarios studied. Both PDC and IOD had high chance of type I error when the treatment could adversely affect sound eye acuity. Of the dichotomous outcomes considered (better than 20/25, better than 20/32, improved 2 or more lines, equal vision in both eyes, and better than 20/32 or improved 3 or more lines), improved 2 or more lines generally had the best statistical power. Like PDC and IOD, the equal visual acuity outcome had high type I error when the treatment could adversely affect sound eye visual acuity. In general, given a fixed sample size, continuous outcomes had greater power to detect treatment differences than the dichotomous outcomes.
Conclusions: :
Choice of an outcome should be based on biological plausibility, clinical relevance, and safety, as well as statistical properties that maximize the chances of detecting a treatment difference and minimize chances of coming to an erroneous conclusion. Either continuous or dichotomous outcome measures may be reasonable choices for a primary outcome measure for clinical trials of amblyopia treatment. The outcomes of lines improved and improvement of 2 or more lines of visual acuity had good statistical properties across a wide range of conditions likely to be encountered in studies of amblyopia treatment, although other choices of outcome may perform well under specific circumstances. IOD and equal visual acuity had poor statistical properties in most circumstances and should not generally be used as primary outcome measures.
Keywords: amblyopia • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology • visual acuity