Polymyxins are important therapeutic options for treating infections, mainly those caused by carbapenem-resistant Klebsiella pneumoniae. Specific chemical characteristics of polymyxins make it difficult to perform antimicrobial susceptibility testing, especially within the clinical laboratory. Here we aimed to evaluate the performance of three phenotypic methods: Rapid NP Polymyxin Test, ColiSpot test and the SuperPolymyxin medium. To accomplish this, 170 non-duplicate clinical K. pneumoniae isolates were analysed (123 colistin-resistant and 47 susceptible). The sensitivity and specificity obtained for Rapid Polymyxin NP Test, Colispot and SuperPolymyxin medium were, respectively, 90% and 94%, 74% and 100%, and 82% and 85%. Very major errors occurred more frequently in low-level colistin-resistant isolates (MICs 4 and 8 μg/mL). Rapid Polymyxin NP proved to be a method capable of identifying colistin-resistant strains in acceptable categorical agreement. However, major errors and very major errors of this method were considered unacceptable for colistinresistance screening. Although the Colispot test is promising and easy to perform and interpret, the results did not reproduce well in the isolates tested. The colistin-containing selective medium (SuperPolymyxin) showed limitations, including quantification of mucoid colonies and poor stability. Nevertheless, Colispot and SuperPolymyxin medium methods did not present acceptable sensitivity, specificity and categorical agreement. It is essential to use analytical tools that faithfully reproduce bacterial resistance in vitro, especially in last-line drugs, such as polymyxins, when misinterpretation of a test can result in therapeutic ineffectiveness.