"Risk of intracranial haemorrhage in patients with atrial fibrillation treated with novel oral anticoagulants: testing the equivalence margins between dabigratran, rivaroxaban and apixaban"by Andrea Messori, Dario Maratea, Valeria Fadda, Sabrina Trippoli; HTA Unit, ESTAV Toscana Centro, Regional Health Service, Firenze (Italy)

------

This material provides further technical details on the methods employed to generate the results of our analysis.

The incidences of ICH were handled as follows: as regards the 4 direct comparisons between a NOAC (dabigatran 110mg twice daily, dabigatran 150mg twice daily, apixaban at standard dose, and rivaroxaban at standard dose) and warfarin, we converted the values of number needed to treat (NNT) reported by Chatterjee and co-workers (see their Table 2) into the corresponding values of percent risk difference (RD) by using the formula RD=100/NNT. Given that we considered 4 treatment options for the three NOACs in addition to warfarin, the network meta-analysis consisted of 10 comparisons; 4 were direct (based on real clinical trials of single NOACs with warfarin as comparator ) while the remaining 6 were indirect. The values of RD (with 95%CI) for the direct comparisons were already available (though implicitly) in the results of Chatterjee et al. [1]; the same values for the 6 indirect comparisons were estimated by network meta-analysis (method of Bucher [2] as implemented in the Indirect Treatment Comparison software, Canadian Agency for Drugs and Technologies in Health, Ottawa, Canada). All RD values were tehn incorporated in a standard equivalence graph [3-6].

In handling our equivalence test, we did not declare any predefined margin of equivalence, but we preferred to carry out a post-hoc analysis in which we determined which equivalence margins were compatible with the conclusion of equivalence.

Finally, this supplementary document includes a different version of our figure that explicitly shows the numerical values of RDs estimated from our analysis (Figure S1).

Supplementary references

[1]Chatterjee S, Sardar P, Biondi-Zoccai G, Kumbhani DJ. New oral anticoagulants and the risk of intracranial hemorrhage: traditional and bayesian meta-analysis and mixed treatment comparison of randomized trials of new oral anticoagulants in atrial fibrillation. JAMA Neurol. 2013 Oct 28. doi: 10.1001/jamaneurol.2013.4021. [Epub ahead of print]

[2]Bucher HC, Guyatt GH, Griffith LE, et al. The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials. J Clin Epidemiol 1997;50:683-91.

[3]Witte S, Victor N. Some problems with the investigation of noninferiority in meta-analysis. Methods Inf Med. 2004;43(5):470-4.

[4]Ahn S, Park SH, Lee KH. How to demonstrate similarity by using noninferiority and equivalence statistical testing in radiology research. Radiology. 2013 May;267(2):328-38.

[5]Messori A, Fadda V, Maratea D, Trippoli S, Marinai C.Nephrotoxicity of different formulations of amphotericin B: summarizing evidence by network meta-analysis. Clin Infect Dis. 2013 Oct 17. [Epub ahead of print] PubMed PMID:24140972.

[6]Messori A, Fadda V, Maratea D, Trippoli S, Marinai C.High-titre inhibitors in previously untreated patients with severe haemophilia A: an updated estimate of pooled incidence rates in patients treated with plasma-derived concentrates or recombinant Factor VIII. Drugs and Cell Therapy in Hematology (2013) in press(preprint available at ).

Figure S1. The legend to this Figure is the same as that reported for Figure 1 in the published article.