Arthritis & Rheumatism, Volume 60,
October 2009 Abstract Supplement
The 2009 ACR/ARHP Annual Scientific Meeting
Philadelphia October 16-21, 2009.
Assessment of the Adequacy of Conclusions in Superiority RCTs: The Example of 3 Rheumatic Diseases- Rheumatoid Arthritis, Spondyloarthropathies and Osteoarthritis
Mathieu1, Sylvain, Boutron2, Isabelle, Giraudeau3, Bruno, Soubrier1, Martin, Ravaud2, P.
Rheumatology. Gabriel Montpied Teaching Hospital, Clermont-Ferrand, France
University of Paris VII, Bichat Hospital, Paris, France, Paris, France
INSERM, CIC 202, France; CHRU de Tours, France, Tours, France
To assess the proportion and to define the causes of misleading conclusions in published randomized controlled trials (RCTs) assessing rheumatoid arthritis (RA), osteoarthritis (OA) or spondyloarthropathies (SPA). To determine whether conclusions were based on the primary outcome (PO).
We searched Medline, Embase and the Cochrane Collaboration for RCTs assessing RA, OA or SPA published between January 2006 and April 2008. Characteristics of each article were extracted (journal information, funding source, financial ties of authors, Jadad scale, results for POs and authors' conclusions). The abstract conclusion was judged to be misleading according to the following criteria: PO not reported, conclusion based only on secondary outcome or sub-group results, discrepancy between results and conclusion, negative results suggested as equivalent and lack of balance benefit/risk discussion.
Of the 145 reports selected, 18 (12.4%) were registered with a clear and similar PO in the registry and the published text and no misleading conclusion in the abstract. We analysed the 108 articles with a clear PO. Twenty-seven per cent had a misleading conclusion in the abstract (n=29). Abstract conclusion in disagreement with results (n=10; 34.5%) and lack of PO reporting (n=8; 27.6%) were the most frequent reasons. Negative results suggested as equivalent, conclusion based only on secondary outcome or on sub-group analysis represented respectively 17.2% (n=5); 13.8% and 13.8% (n=4) of cases of misleading conclusions. Finally, lack of benefit/risk balance discussion in cases of serious adverse events was found in 10.3% (n=3) of articles. Of the 108 studies analysed, 63 (53.3%) had positive results and 43 (39.8%) negative results. Studies with negative results had a higher frequency of misleading conclusions in the abstract than those with positive results (22/43 versus 5/63; p<0.001). Articles assessing OA, studies with an industrial affiliation of authors or those without clinically relevant results were more likely to report misleading conclusions in univariate analysis. Other factors did not reach statistical significance. In multivariate analyses, only negative results increased the risk of misleading conclusions (OR=9.27 [1.8945.45]).
This study found a proportion of 27% of misleading conclusions in the abstracts of a sample of RCTs assessing RA, OA and SPA. Negative and clinically irrelevant results were risk factors of misleading conclusion.
To cite this abstract, please use the following information:
Mathieu, Sylvain, Boutron, Isabelle, Giraudeau, Bruno, Soubrier, Martin, Ravaud, P.; Assessment of the Adequacy of Conclusions in Superiority RCTs: The Example of 3 Rheumatic Diseases- Rheumatoid Arthritis, Spondyloarthropathies and Osteoarthritis [abstract]. Arthritis Rheum 2009;60 Suppl 10 :709