Abstract Background To assess the ability of procalcitonin (PCT) to distinguish between bacterial and nonbacterial causes of patients with severe acute exacerbation of COPD (AECOPD) admitted to the ICU, we conducted a retrospective analysis of two prospective studies including 375 patients with severe AECOPD with suspected lower respiratory tract infections. PCT levels were sequentially assessed at the time of inclusion, 6 h after and at day 1, using a sensitive immunoassay. The patients were classified according to the presence of a documented bacterial infection (including bacterial and viral coinfection) (BAC + group), or the absence of a documented bacterial infection (i.e., a documented viral infection alone or absence of a documented pathogen) (BAC- group). The accuracy of PCT levels in predicting bacterial infection (BAC + group) vs no bacterial infection (BAC- group) at different time points was evaluated by receiver operating characteristic (ROC) analysis. Results Regarding the entire cohort (n = 375), at any time, the PCT levels significantly differed between groups (Kruskal–Wallis test, p