Mortality of drug-resistant tuberculosis in high-burden countries: comparison of routine drug susceptibility testing with whole-genome sequencing

BackgroundDrug-resistant Mycobacterium tuberculosis (Mtb) strains threaten tuberculosis (TB) control. We compared data on drug resistance obtained at clinics in seven high TB burden countries during routine care with whole-genome sequencing (WGS) carried out centrally.MethodsWe collected pulmonary Mtb isolates and clinical data from adult TB patients in Africa, Latin America, and Asia, stratified by HIV status and drug resistance, from 2013 to 2016. Participating sites performed drug susceptibility testing (DST) locally, using routinely available methods. WGS was done using Illumina HiSeq 2500 at laboratories in the USA and Switzerland. We used TBprofiler to analyse the genomes. We used multivariable logistic regression adjusted for sex, age, HIV-status, history of TB, sputum positivity, and Mtb-lineage to analyse mortality.FindingsWe included 582 TB patients. The median age was 32 years (interquartile range: 27-43 years), 225 (39%) were female, and 247 (42%) were HIV-positive. Based on WGS, 339 (58%) isolates were pan-susceptible, 35 (6%) monoresistant, 146 (25%) multidrug-resistant, and 24 (4%) pre-/ extensively drug-resistant (pre-XDR/XDR-TB). The local DST results were discordant compared to WGS results in 130/582 (22%) of patients. All testing methods identified isoniazid and rifampicin resistance with relatively high agreement (kappa 0.69 for isoniazid and 0.88 rifampicin). Resistance to ethambutol, pyrazinamide, and second-line drugs was rarely tested locally. Of 576 patients with known treatment, 86 (15%) patients received inadequate treatment according to WGS results and the World Health Organization treatment guidelines. The analysis of mortality was based on 530 patients; 63 patients (12%) died and 77 patients (15%) received inadequate treatment. Mortality ranged from 6% in patients with pan-susceptible Mtb (18/310) to 39% in patients with pre-XDR/XDR-TB (9/23). The adjusted odds ratio for mortality was 4.82 (95% CI 2.43-9.44) for under-treatment and 0.52 (95% CI 0.03-2.73) for over-treatment.InterpretationIn seven high-burden TB countries, we observed discrepancies between drug resistance patterns from local DST and WGS, which resulted in inadequate treatment and higher mortality. WGS can provide accurate and detailed drug resistance information, which is required to improve the outcomes of drug-resistant TB in high burden settings. Our results support the WHO’s call for point-of-care tests based on WGS.