OBJECTIVES To compare the relative efficacy of ustekinumab to alternative therapies in anti-TNF treatment-naïve adult patients with active PsA, while adjusting for the variability in placebo response rates across trials. METHODS A Bayesian network meta-analysis (NMA) was performed to compare ustekinumab with adalimumab, golimumab, etanercept, certolizumab pegol and infliximab. Four outcomes (PASI75, PASI90, PSARC, ACR20) were analysed after 12-16 and 24 weeks. The NMA was conducted using a meta-regression model, with the trial-specific estimated baseline risk included as a covariate (Dias et al., 2013). A sceptical prior was used for this covariate. Both fixed (FE) and random effects models were considered. RESULTS Nine placebo-controlled trials were identified for inclusion in the NMA based on systematic literature review, including 2 ustekinumab trials. The placebo response rates varied significantly across trials, with ustekinumab trials having generally higher values. The median value of the meta-regression coefficient ranged between -0.02 and -1.69 (P[coef<0] between 50% and 95%) over the different scenarios and models, suggesting an interaction effect between baseline risk and treatment effects. At week 24 using the FE, the probability for ustekinumab 45mg and 90mg to be more effective than the comparators based on the PASI75 ranged from 87% (etanercept) to 46% (golimumab 100mg) and from 88% (etanercept) to 50% (golimumab 100mg), respectively. The variability around the point estimates was however large. CONCLUSIONS This analysis indicates baseline risk in PsA-trials to be a treatment effect modifier. Any NMA not correcting for baseline risk might generate biased results. After adjusting for differences in baseline risk between the trials, ustekinumab indicates comparable efficacy to alternative therapies, however with high uncertainty around the point estimates.