Abstract
Purpose: :
In ophthalmology, investigators sometimes use a within-person trial design where eyes within an individual are randomized separately. Because characteristics of eyes within an individual are correlated, within-person paired randomized controlled trials (RCTs) should account for non-independence in trial design and analyses. Our study describes design and analysis characteristics of within-person RCTs in ophthalmology.
Methods: :
Within person RCTs in ophthalmology were identified by searching PubMed, EMBASE, and The Cochrane Library using a specifically designed search strategy. Two individuals independently screened reports for eligibility, including within person RCTs in ophthalmology and excluding non-English reports and reports that were only abstracts. Using an electronic data collection form, two individuals independently extracted design characteristics, including rationale, randomization method, intervention, and statistical tests. Discrepancies were resolved by a third reviewer or discussion between original reviewers. Data were tabulated for descriptive analysis.
Results: :
Of 482 reports identified by the search strategy, 60 were eligible for inclusion. Almost all studies randomized the first eye and assigned the second eye to the alternate treatment (57/60, 95%). Interventions included drugs (24/60; 40%), contact or intraocular lenses (14/60, 23%), laser surgery (12/60, 20%), and other (10/60, 17%). Less than half of investigators reported a rationale for using a within-person design (25/60, 42%) or eligibility criteria related to the similarity of eyes (29/60, 48%). Most authors reported using at least one statistical test that took into account non-independence of eyes: paired t test (35/60, 58%), Wilcoxon signed rank sign (20/60, 33%), McNemar’s (11/60, 18%); but also included tests that did not (ANOVA (10, 17%), Chi square (9/60, 15%), and Fishers exact (3/60, 5%).
Conclusions: :
Investigators sometimes took into account the non-independence of eyes with respect to study rationale, eligibility criteria, and statistical testing. CONSORT guidelines for within-person paired RCTs may be useful in improving within-person RCT design and reporting.
Keywords: clinical (human) or epidemiologic studies: treatment/prevention assessment/controlled clinical trials • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology