Comparative gastrointestinal safety of bisphosphonates in primary osteoporosis: a network meta-analysis-reply to Pazianas and Abrahamsen.

Abstract

Dear Editor, We appreciate the interest of Drs. Pazianas and Abrahamsen [1] in our recent article [2]. Their letter raises some important issues related to the application and interpretation of network meta-analysis (NMA). First, it is important to appreciate the distinction between traditional pairwise meta-analysis and NMA. A pairwise meta-analysis typically includes direct evidence and synthesizes results using frequentist statistics. In contrast, NMA utilizes direct and indirect evidence using Bayesian-based simulations. This difference in statistical methodology produces results that are interpreted in unique ways [3–5]. The results in Table 2 that are highlighted by Drs. Pazianas and Abrahamsen are not proportions of patients (incidence) experiencing the outcomes of interest; rather, Table 2 presents the percentage of times that simulations conducted within the NMA showed a medication to be superior with regard to the outcome of interest. This is referred to as the probability of superiority [2, 5]. For example, the analysis of esophageal adverse events (AEs) found that alendronate had a 54 % probability of superiority; that is, in 54 % of simulations, alendronate had the most patients with esophageal AEs, not that 54 % of patients had esophageal AEs. Second, the letter raises concerns about the validity of including nausea as a gastrointestinal (GI) AE with zoledronic acid since there is no known mechanism to explain such reactions for an intravenous (IV) bisphosphonate. However, nausea is a commonly reported GI AE with zoledronic acid and oral formulations [6, 7]. Moreover, not only is nausea commonly reported separately in trials, but international standards classify nausea under the gastrointestinal subheading [8]. We were not surprised to find that zoledronic acid had the highest probability of superiority with nausea due to the means of administration of this agent. We agree that nausea may often be part of the acute phase reaction with IV administration, yet also acknowledge that nausea can occur outside of the acute phase reaction. We are puzzled by the suggestion to consider acute phase reaction specifically in our study since acute phase reactions do not occur with oral formulations and would not allow for indirect comparisons. We agree that the results of our study do not warrant strong conclusions of differences in adherence, but suggest that they bring into question the previously held assumption that less M. Tadrous (*) :M. M. Mamdani :M. D. Krahn : S. M. Cadarette Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto, ON M5S 3M2, Canada e-mail: mina.tadrous@utoronto.ca