Aims and objectives: A phenotypic drug susceptibility test (pDST) is the standard method for determining Mycobacterium tuberculosis (MTB) drug susceptibility. However, TB treatment can be complicated by errors in pDST. We evaluated the differences in pDST error rates based on the types of samples.

Methods: We retrospectively reviewed medical records of patients diagnosed with tuberculosis in a single tertiary hospital for three years. Patients with at least one cultured MTB and pDST were included. NTM mixed culture, insufficient growth of control medium, and strain contamination were defined as pDST errors.    

Results: During the study period, 503 samples were collected from 447 patients. There were 282 men (61.4%) and an average age of 60.9 years. Among the samples collected, there were 354 samples of sputum (77.9%), 79 samples of washing (14.7%), 11 samples of lung tissues (2.4%), 8 samples of pleural fluid (1.6%), and 5 samples from other sources (1.2%). There were pDST errors in 44 samples (8.7%) of 36 patients. Mixed cultures with NTM were found in 18 cases, insufficient control medium growth was found in 17, and strain contamination was found in 9 cases. pDST error rates were higher in sputum samples than in non-sputum samples (9.7% vs. 5.4%; P = 0.186). In 36 patients with errors, 12 had performed additional pDSTs more than once. Five of them found no additional errors. Despite successful treatment for tuberculosis, two patients relapsed, and all were multidrug-resistant.

Conclusion: The error rate of pDST varied according to the cultured TB specimens. Whenever errors occur, it is necessary to check pDST repeatedly.