Challenging the gold standard: critical limitations in clinical detection of drug- resistant tuberculosis

Heteroresistant infections - defined as infections in which minority drug-resistant (DR) populations are present - are a challenge in infectious disease control. In Mycobacterium tuberculosis, heteroresistance poses challenges in diagnosis and has been linked with poor treatment outcomes. We compared the analytic sensitivity of molecular methods, such as GeneXpert and whole genome sequencing (WGS) in detecting heteroresistance when compared to the gold standard phenotypic assay: the agar proportion method (APM). Using defined mono-resisitant BCG strains we determined the limit of detection (LOD) of rifampin-R (RIF-R) detection was 1% using APM, 60% using Xpert MTB/RIF and 10% using Xpert MTB/RIF Ultra. To evaluate clinical WGS pipelines, a blinded panel of BCG mixtures was sent to 3 clinical labs. These were composed of either a) RIF-R plus isoniazid-R (INH-R) BCG or b) fluoroquinolone-R (FQ-R) plus clofazimine-R/bedaquiline-R (CLZ/BDQ-R) BCG. No labs called resistance at 1%; all labs called RIF-R at 10% or greater and two out of three labs reported FQ-R at 10%. Two labs were able to detect the majority population (either INH-R or CLZ/BDQ-R) at 50%. Importantly, where labs did not report resistance in the majority population, the mutations were present in the raw data but excluded from the final analysis. In conclusion, the gold standard APM more reliably detects minority resistant populations than molecular tests. Further research is required to determine whether the higher LOD of molecular tests is associated with deleterious patient outcomes and the potential effects on transmission of resistance at the population level.