Using Risk Difference As Opposed to Odds-Ratio in Meta-Analysis

IJCA-15035; No of Pages 2 International Journal of Cardiology xxx (2012) xxx–xxx

Contents lists available at SciVerse ScienceDirect

International Journal of Cardiology

journal homepage: www.elsevier.com/locate/ijcard

Letter to the editor Using risk difference as opposed to odds-ratio in meta-analysis

Andrea Messori ⁎, Dario Maratea, Valeria Fadda, Sabrina Trippoli

HTA Unit, ESTAV Toscana Centro, Regional Health Service, 50100 Firenze, Italy

article info so it is likely to be less useful that risk difference or relative risk for communicating risk”. Article history: We reanalysed the data set of Aradi et al. [1] on cardiovascular Received 22 June 2012 mortality (Figure 2 of their article) by choosing the risk difference Accepted 24 June 2012 rather than the OR. All calculations (random-effect model) were Available online xxxx carried out using the OMA software that automatically implements Keywords: a correction for the presence of zero values as event frequencies [3]. Clopidogrel The results of our analysis (Fig. 1) do not entirely confirm the find- Meta-analysis ings of Aradi et al. [1]. In fact, while the p-value of Aradi et al. was Odds ratio 0.008 [1], our analysis found a p-value of 0.0515. More importantly, Risk difference fi Percutaneous coronary intervention the bene t estimated by our analysis was a risk reduction of 0.47% (95% confidence interval: from a reduction of 0.89% to an increase of 0.003%). It is interesting to see that a statistical level initially far from the conventional threshold of p=0.05 has been converted into a non- In comparing cardiovascular death between patients receiving significant value by using a more conservative (and more appropriate) fi intensi ed antiplatelet therapy after percutaneous coronary interven- statistical analysis. Does this have any implication in interpreting tion and controls given standard-dose clopidogrel, the meta-analysis Aradi's findings? by Aradi et al. [1] handled a very critical data set because only 10 randomised studies were available and more than half of the event References frequencies were zero (13 out of 20). Besides this large number of zero frequencies, another weakness [1] Aradi D, Komócsi A, Price MJ, et al. On behalf of the tailored antiplatelet treatment fi fi of this study is the choice of the odds-ratio (OR) as effect measure. study collaboration. Ef cacy and safety of intensi ed antiplatelet therapy on the basis of platelet reactivity testing in patients after percutaneous coronary intervention: More than 10 years ago, Walter [2] stressed that the OR tends to systematic review and meta-analysis. Int J Cardiol Jun. 15 2012 [Epub ahead of print]. “exaggerate the effect of treatment” as a direct consequence of its [2] Walter SD. Choice of different effect measure for epidemiological data. J Clin – mathematical definition. Another drawback pointed out by Walter Epidemiol 2000;53:931 9. [3] Open Meta-Analyst software, version 4.16.12, Tufts University, May 2012 url [2] is that “the OR is undeniably the most difficult measure to intuit, http://tuftscaes.org/open_meta/

⁎ Corresponding author at: HTA Unit, Area Vasta Centro Toscana, Regional Health System, Via Guimaraes 9‐11, 59100 Prato, Italy. Tel.: +39 3476053933; fax: +39 0574 701319. E-mail address: [email protected] (A. Messori).

Please cite this article as: Messori A., et al, Using risk difference as opposed to odds-ratio in meta-analysis, Int J Cardiol (2012), doi:10.1016/ j.ijcard.2012.06.078 2 A. Messori et al. / International Journal of Cardiology xxx (2012) xxx–xxx

Fig. 1. Comparison of cardiovascular mortality between patients receiving intensified antiplatelet therapy and controls given standard-dose clopidogrel. This meta-analysis was based on risk differences and random-effect model as implemented in the OMA software [3].Blacksquaresdenotestudy-specific risk differences (with 95% confidence interval) while diamonds indicate pooled risk difference (with 95% confidence interval); I2 is a measure of heterogeneity (accompanied by its level of statistical significance); the dotted vertical line highlights the meta-analytical value of risk difference. Abbreviations: Ev=number of events; Trt=number of treated patients. The complete references of the original studies have been reported by Aradi et al. [1].

Please cite this article as: Messori A., et al, Using risk difference as opposed to odds-ratio in meta-analysis, Int J Cardiol (2012), doi:10.1016/ j.ijcard.2012.06.078