Comparative Utility of Genetic Determinants of Drug Resistance and Phenotypic Drug Susceptibility Profiling in Predicting Clinical Outcomes in Patients With Multidrug-Resistant Mycobacterium tuberculosis

Setting: Programmatic management of drug-resistant tuberculosis in Ningbo, China. Objective: To assess whether data-driven genetic determinants of drug resistance patterns could outperform phenotypic drug susceptibility testing in predicting clinical meaningful outcomes among patients with multidrug-resistant tuberculosis (MDR-TB). Design: We conducted a prospective cohort study of 104 MDR-TB patients. All MDR-TB isolates underwent drug susceptibility testing and genotyping for mutations that could cause drug resistance. Study outcomes were time to sputum smear conversion and probability of treatment success, as well as time to culture conversion within 6 months. Data were analyzed using latent class analysis, Kaplan-Meier curves, and Cox regression models. Results: We report that latent class analysis of data identified two latent classes that predicted sputum smear conversion with P = 0.001 and area under receiver-operating characteristic curve of 0.73. The predicted latent class memberships were associated with superior capability in predicting sputum culture conversion at 6 months and overall treatment success compared to phenotypic drug susceptibility profiling using boosted logistic regression models. Conclusion: These results suggest that genetic determinants of drug resistance in combination with phenotypic drug-resistant tests could serve as useful biomarkers in predicting treatment prognosis in MDR-TB.

Automated summary

By analyzing MDR-TB patients with sputum smear testing and genotyping, the study found that latent class analysis based on data-driven genetic determinants could better predict clinical meaningful outcomes, showing promising potential in predicting treatment prognosis for patients.