OBJECTIVES: To evaluate differences across randomized controlled trials (RCTs) concerning interventions for moderate to severe chronic obstructive pulmonary disease (COPD) in terms of baseline exacerbation rates and the association with treatment effects by means of a network meta-analysis (NMA). This NMA is performed based on RCTs evaluating the long-acting bronchodilators indacaterol 75/150/300µg OD, salmeterol 50µg BID, formoterol 12µg BID, tiotropium bromide 18µg/5µg OD, and glycopyrronium bromide 50µg OD. METHODS: The rate of moderate or severe exacerbations was extracted from RCTs identified with a systematic literature review. A Bayesian NMA was used to synthesize the treatment effects of the different trials. The association between treatment effects and baseline exacerbation rate with placebo was assessed with a meta-regression model assuming a constant treatment-by-baseline risk interaction term. RESULTS: Twenty-four RCTs were included that differed mainly in terms of smoking status, COPD severity, use of inhaled corticosteroids, exacerbation definition, and exacerbation history. Across the RCTs the rate of exacerbations per patient year for patients in the placebo arm ranged from 0.40 to 1.91. Baseline risk was negatively associated with the rate ratios reflecting treatment effects across the RCTs. The coefficient for baseline risk was -0.35 (95% credible intervals: -0.49, -0.18). CONCLUSIONS: Based on a NMA of RCTs regarding the efficacy of long-acting bronchodilators in terms of the rate of exacerbations per patient year, baseline risk of exacerbations acts as a significant treatment effect modifier and should be accounted for in the model.