PURPOSE: Most of the statistical tests used in significance testing are based on the assumption that each data entry is independent from other entries, however, we observe that in many articles researchers include data from 2 eyes as independent variables while performing these statistical tests. The aim of this study was to formally survey the prevalence of the above-mentioned "double-organ bias" in randomized controlled trials (RCT) of glaucoma. MATERIALS AND METHODS: We did a PubMed search with the terms "glaucoma" and limitations "Humans" and "Randomized Controlled Trials" in 15 highest-impact-factor ophthalmology journals between November 2002 and November 2012. We only included RCTs published as an original article, where the aim was treating glaucoma. Two independent observers (M.K. and A.S.E.) read through each article and classified the articles according to treatment modality (medical, laser, or surgical) and presence of double-organ bias. RESULTS: The PubMed search yielded 270 articles. A total of 130 articles qualified for the survey. Eighty-five of the RCTs were medical studies, 11 were laser studies, and 34 studies evaluated the outcome of a surgical procedure. In 17 of the 130 articles (13.1%), double-organ bias was found. Prevalence of the double-organ bias was not significantly different between medical (12.9%), laser (14.7%), and surgical (9.1%) studies. CONCLUSION: Double-organ bias was observed around 13.1% of the published RCTs, leading to inaccurate statistical testing.
PURPOSE: Most of the statistical tests used in significance testing are based on the assumption that each data entry is independent from other entries, however, we observe that in many articles researchers include data from 2 eyes as independent variables while performing these statistical tests. The aim of this study was to formally survey the prevalence of the above-mentioned "double-organ bias" in randomized controlled trials (RCT) of glaucoma. MATERIALS AND METHODS: We did a PubMed search with the terms "glaucoma" and limitations "Humans" and "Randomized Controlled Trials" in 15 highest-impact-factor ophthalmology journals between November 2002 and November 2012. We only included RCTs published as an original article, where the aim was treating glaucoma. Two independent observers (M.K. and A.S.E.) read through each article and classified the articles according to treatment modality (medical, laser, or surgical) and presence of double-organ bias. RESULTS: The PubMed search yielded 270 articles. A total of 130 articles qualified for the survey. Eighty-five of the RCTs were medical studies, 11 were laser studies, and 34 studies evaluated the outcome of a surgical procedure. In 17 of the 130 articles (13.1%), double-organ bias was found. Prevalence of the double-organ bias was not significantly different between medical (12.9%), laser (14.7%), and surgical (9.1%) studies. CONCLUSION: Double-organ bias was observed around 13.1% of the published RCTs, leading to inaccurate statistical testing.
Authors: António Campos; Elisa J Campos; Anália do Carmo; Francisco Caramelo; João Martins; João P Sousa; António Francisco Ambrósio; Rufino Silva Journal: Eye Vis (Lond) Date: 2018-10-11