Cardiovascular safety of the class of JAK inhibitors or tocilizumab compared with TNF inhibitors in patients with rheumatoid arthritis: Systematic review and a traditional and Bayesian network meta-analysis of randomized clinical trials

Alisson Pugliesi,Daniela Gomes Chicre Oliveira,Vani Abreu de Souza Filho,Júlia de Oliveira Machado,Aline Gonçalves Pereira,Júlia de Castro Silveira Bichuette,Zoraida Sachetto,Luiz Sérgio F de Carvalho,Manoel Barros Bertolo

DOI: https://doi.org/10.1016/j.semarthrit.2024.152563

Abstract:Background: We aimed to compare the risk of major adverse cardiovascular events (MACE) and all-cause of death (ACD) in patients with rheumatoid arthritis (RA) treated with JAK inhibitors (JAKi) or tocilizumab (TCZ) versus tumor necrosis factor (TNF) inhibitors (TNFi). Methods: We performed a systematic review of six medical databases until May, 2024 for phase 2-4 randomized controlled trials (RCTs) evaluating patients with RA treated with TCZ or JAKi (intervention arm) compared with controls (TNFi or placebo). The study data were independently assessed by 3 investigators. The risk of bias was assessed using the Cochrane Collaboration tool. We performed a network meta-analysis with random effects to evaluate the risk of MACE (primary outcome) and ACD (secondary outcome) compared to TNFi. We also calculated the posterior probability of increasing the primary and secondary outcomes by 15% or more (PP15%) following Bayes' theorem. Results: This meta-analysis included 18 RCTs with 21,432 patients and 57,040 patient-years. JAKi were linked to a non-statistically significant increase in the risk of MACE and ACD as compared to TNFi (ORs of 1.232 [95%CI 0.86-1.76]; p = 0.56 and ORs = 1.3903[95%CI 0.94-2.07]; p = 0.10, respectively). By Bayesian analysis, a high clinical probability of more frequent MACE (PP15% of 61%) and ACD (PP15% of 84%) was found in the JAKi group than in the TNFi group. No statistical difference was found between TCZ and TNFi in relation to MACE (1.029 [95%CI 0.75 -1.40]; p = 0.86) and ACD (1.072 [95%CI 0.78-1.48]; p = 0.67]) in the traditional meta-analysis. In the Bayesian approach, the probability of a difference in clinical relevance was low (PP15% for MACE of 11% and PP15% for ACD of 25%). Discussion: The main limitation of this study is the small number of events with JAKi other than tofacitinib, reflecting the importance of ORAL SURVEILLANCE. Despite this, these data reinforce the recommendations of regulatory agencies and rheumatology societies on the use of JAKi in the context of RA, but above all call for more direct comparison studies involving primary safety outcomes with JAKi, since the outcome data available are still small and heterogeneous. In both meta-analyses, no difference was found between TNFi and TCZ.