Abstract
PURPOSE
Proper statistical analysis is an essential element in the evaluation of clinical trial outcomes. We had the informal observation that double-organ bias was a neglected issue during the statistical analyses of clinical trials on eyelid ptosis. The aim of this study was to formally document the prevalence of this bias in these studies.
METHODS
Clinical trials on eyelid ptosis, published in the last 20 years, were searched in PubMed with the terms; "((blepharoptosis) OR upper eyelid ptosis) OR eyelid ptosis" and with the filters "Randomized Controlled Trial (RCT), Clinical Trial, Humans." Two independent observers evaluated the articles for eligibility, field of the journal, field of the authors, presence of randomization, type of the study (surgical, medical), and statistical approach to double organ bias.
RESULTS
The PubMed search yielded 101 articles and 23 of them met the above-mentioned criteria. In 3 articles, primary outcome measure was not related with the eyes. Among the remaining 20 articles, 14 (70%) had double-organ bias in the statistics. The bias was slightly less common in randomized trials (60% vs. 80%), but the difference was not statistically significant. Its prevalence was similar between ophthalmology journals and the rest (p = 0.64). Interestingly, the bias tended to be more in Q1 journals (87.5%) compared to Q2 and Q3 journals (58.3%) and median impact factor was higher in biased articles (1.82 vs. 1.29), but the differences weren't statistically significant (p = 0.32, p = 0.24). There was no difference between the last 2 decades (66.6% vs. 75%, p = 0.64).
CONCLUSIONS
The prevalence of the double-organ bias was very high in published clinical trials on eyelid ptosis (70%) and even among RCTs (60%). The prevalence of the bias didn't prevent publication in higher impact factor journals and didn't change between journals from different disciplines or over time.
Collapse