SAT0466 The "table-1 P value" issue; baseline group comparison is inappropriately omitted in non-randomized studies. (23rd January 2014)

• Record Type: Journal Article
• Authors: Tascilar, K.; Atac, E.; Esen, F.; Yazici, H.
• Is Part Of: Annals of the rheumatic diseases. Volume 71(2012)Supplement 3
• Page Start: 630
• Page End: 630
• Publication Date: 2014-01-23
• Subjects: Rheumatism -- Periodicals; 616.723005
• DOI: 10.1136/annrheumdis-2012-eular.3412
• Languages: English
• ISSNs: 0003-4967

Abstract

Abstract : Background: Guidelines on reporting randomized controlled trials advise against performing baseline comparison in randomized groups (1). Such comparison is considered futile as any difference will be by chance by definition (2). In non-randomized studies, on the contrary, the baseline characteristics of the study groups should be compared because the differences may have direct implications on the results. Objectives: The aim of this study was to determine the frequency of inappropriate commission or omission of baseline comparisons in randomized and non-randomized studies, respectively. Methods: A PubMed search was conducted to identify clinical studies among humans, designed to test the effect of a treatment or an observation in more than one group in 3 major rheumatology journals; Annals of the Rheumatic Diseases, Arthritis and Rheumatism and Rheumatology (Oxford) from June 2010 to June 2011. Studies were grouped as either randomized or non-randomized. We evaluated each study as to how the baseline characteristics were presented. Also the frequencies of inappropriate commission or omission of statistical comparisons were tabulated. Results: 205 studies were found, 107 were included. There were 60 randomized trials and 47 non-randomized studies. Among the randomized studies, 6/60 (10%) reported inappropriate baseline statistical comparisons. On the other hand among the non-randomized studies, 27/47 (57%) omitted baseline statistical comparison. In 10 of these 27 studies, there was no table presenting the characteristics of the groups subject to comparison. 17 studies on the other hand presented baseline groups in a table 1 but did no statistical comparisons although necessary. Conclusions: The CONSORT statement seems to be better adhered to when reporting randomized controlled trials. On the other hand there is common and inappropriate omission of the necessary statistical comparisons at baseline in non-randomized studies. In this line there's an obvious need to improve adherence to the TREND statement (3). References: J Pharmacol Pharmacother. 2010;1:100-7. Lett Ed Rheumatol 1:e110002. doi:10.2399/ler.11.0002. Am J Public Health. 2004; 94: 361–366. Disclosure of Interest: None Declared