To evaluate the presence of spin in SRs of interventions from the top-5 general medical journals, as well as to investigate characteristics of SRs that were associated with spin in the abstract.

This overview of reviews included a random sample of full-published SRs with meta-analysis. Selected journals were: 1) Lancet; 2) Journal of American Medical Association (JAMA); 3) British Medical Journal (BMJ); 4) Annals of Internal Medicine (AIM); and 5) JAMA Internal Medicine. The electronic search on each selected journal website and on PubMed covered the period from January 2011 to November 2017. Methodological quality of the included SRs was assessed through the AMSTAR-2 checklist (A MeaSurement Tool to Assess systematic Reviews). Data selection, extraction and quality assessment were performed in duplicate. Descriptive analysis was performed. We used a seven-item SPIN checklist to assess spin in the abstracts and full texts. Two linear regression models were built to investigate the association between the presence of spin in the abstract and full text with some characteristics of the SRs.

196 SRs were included (total of 11,881,507 participants), from which 32 were from JAMA Internal Medicine, 39 from BMJ, 40 from Lancet, 42 from JAMA and 43 from AIM. The assessment of SRs methodological quality resulted in 152 SRs with critically low overall confidence, 37 with low, 6 with moderate and only 1 high overall confidence. The journals presented from 1 to 4 items of spin (out of 7), ranging from 84% to 100% in abstracts and 51% to 81% in full texts. Higher levels of spin in the abstracts were associated with critically low methodological quality reviews (ß 0.22; 95% CI 0.14, 0.63). The main limitation of this study was the exclusion of the New England Journal of Medicine.

More than half of full texts and most of abstracts presented spin, with the latter associated with reviews methodological quality. Ultimately, the flaws could only be directed to the authors, journal editors and reviewers.