OBJECTIVES:  To assess the suitability of a network meta-analysis (NMA) of targeted synthetic and biologic DMARDs in patients with active RA despite biologic DMARD (bDMARD) treatment. METHODS:  A systematic literature review was performed to identify relevant randomized controlled trials (RCTs). Efficacy outcomes of interest were ACR20 response and change in DAS28(CRP) at 24±4 weeks. Suitability of a valid NMA was assessed according to the following criteria: 1) presence of 1 connected network where each trial has ≥1 intervention in common with another trial; 2) the distribution of study characteristics, patient characteristics, and contextual factors that are effect-modifiers are balanced across trials; and 3) the ability to overcome bias due to differences in effect-modifiers with statistical analysis. Identification of potential effect-modifiers was based on reported subgroup results in the literature and analysis of patient level data from a sirukumab phase 3 RCT (SIRROUND-T). RESULTS:  Eight RCTs (2005–2016) were identified. Meaningful between-trial variation was observed, including differences in inclusion criteria, prior bDMARD use, and rescue therapy. Six trials included patients who failed anti-TNFs only, whereas 2 included patients who failed any bDMARD. Percentages of patients failing ≥2 prior bDMARDs ranged from 21%–61%. Where permitted, rescue therapy was administered after 12–18 weeks. Pre-study cDMARD stabilization periods varied (4–8 weeks). Unexplained variability in the placebo response (10%–34% ACR20 response) further highlighted the differences in patient characteristics between trials. A meta-regression analysis attempting to control for between-trial differences in effect-modifiers or high placebo response was not feasible due to the small number of trials. CONCLUSIONS:  Available trials in the post-bDMARD population are characterized by meaningful variation in likely effect-modifiers. In the absence of patient-level data to adjust for patient-related between-trial differences, a NMA based on reported study-level data is likely to be biased.