Ophthalmic RCTs need better design, statistical analysis

This study in the April issue of Ophthalmology assessed the design of ophthalmic randomized controlled trials (RCTs). The authors found that there is substantial heterogeneity in the quality of published studies in terms of the proper use of study design, sample size calculation, randomization method and statistical tools.

They note that ophthalmic studies can present a statistical problem in RCTs because individual subjects potentially have two interrelated data points, that is, two eyes, to contribute to the study, and many published articles do not account for this. They recommend that investigators for future ophthalmic RCTs consult a statistician and follow guidelines, such as the CONSORT statement, in order to improve the quality of ophthalmic research.

These conclusions were based on the results of a systematic review of 69 ophthalmic RCTs published in 2009 in the American Journal of Ophthalmology, Archives of Ophthalmology, the British Journal of Ophthalmology and Ophthalmology.

The authors found that the most common study design was a one-eye design (48 percent). Within this group, only half described the method of selecting the study eye, among which five chose the study eye by random selection. The remaining trials included paired-eye design (13 percent), subject design (19 percent) and two-eye design (19 percent).

Among the two-eye design studies, four allocated both eyes of subjects to the same group, four allocated the eyes to different groups and four did not restrict the allocation. None of these studies adjusted for the clustering effect in sample size calculation. Only five studies used statistical methods adjusting for nonindependence.

The authors conclude that researchers should be aware of and account for the fact that different study designs require the use of different methods with specific statistical tools.