Hospital-acquired infections by multidrug-resistant Gram-negative bacteria (MDRGN) have become a global public health problem. Colistin is considered one of the last therapeutic options due to its limited clinical effectiveness and high rate of adverse effects. Unfortunately, its use has increase in recent years given the increase of MDRGN-associated infections and narrow therapeutic alternatives. With the aim of improving effectiveness using pharmacokinetic/pharmacodynamic knowledge, higher doses of colistin have been used in recent years but with few reported outcomes. Methods: A retrospective cohort analysis was performed in patients with MDRGN infections treated with high-dose colistin with the aim of evaluating clinical improvement as primary outcome and the incidence of acute kidney injury (AKI) and other adverse events as secondary outcomes. Results: Fifty-five courses of colistin treatment were identified in 50 patients, with 45% applied on intensive care unit. Clinical improvement was achieved in 35 (63.6%) treatments, while extrapulmonary infections were significantly associated with a higher clinical failure rate (OR 10; 95%CI: 1.18-84.5). By multivariate analysis, only the failure to control the infection source influenced significantly for mortality of patients (aOR= 19.6; IC95 3.0-126, p= 0.002). AKI was observed in 30 treatments (54.5%) and was only significantly associated with the use of a loading doses (OR= 6.0, 95%CI: 1.61-22.3). Eosinophilia was frequent (35.7%) and, besides, two respiratory depression events were observed. Conclusion: High doses of colistin could be associated with a favorable clinical improvement in patients with MDRGN infections but has a limited effectiveness in extra-pulmonary infections, especially when a source-control procedure is not performed. AKI is frequently observed and limits its use, while eosinophilia and respiratory depression should be considered also as part of safety monitoring. Prescription of this drug should be judiciously analyzed weighing benefits-to-risk ratio.