Abstract Objective To compare the risk of diverticulitis and gastrointestinal perforation (GIP) in RA treated with tocilizumab (TCZ) compared with rituximab (RTX) and abatacept (ABA). Methods We conducted a population-based study using three observational French registries on TCZ, RTX and ABA in RA. Using a propensity score approach, we compared the risk of diverticulitis or GIP in these patients. Results With inverse probability weighting, there was an increased risk of diverticulitis in TCZ-treated patients compared with RTX- or ABA-treated patients [hazard ratio (HR)=3.1 (95% CI: 1.5, 6.3), P =0.002]. Moreover, patients treated with TCZ had also an increased risk of GIP due to diverticulitis compared with those treated with RTX or ABA [HR=3.8 (1.1–13.6), P =0.04], resulting in an overall increased risk of GIP [HR=2.9 (1.1–7.8), P =0.03], while no significant increased risk of GIP due to any other aetiology was found in TCZ treated patients. Diverticulitis and GIP occurred earlier with TCZ than other drugs after the last perfusion (P =0.01), with atypical clinical presentation (slow transit in 30%, P =0.04) and lower acute-phase reactants at the time of the event (P =0.005). Conclusion TCZ for RA was associated with increased odds of diverticulitis as well as GIP due to diverticulitis as compared with RTX and ABA. Our study confirms the increased odds of GIP in patients receiving TCZ, which might be explained by an increased risk of diverticulitis with misleading clinical presentation.