The World Health Organization (WHO) recommends three pillars for tuberculosis (TB) control and eradication: integrated patient-centered care and prevention; bold policies and support systems, especially for vulnerable populations; and intensification in research and innovation1 with emphasis on resistance detection and new diagnostic assays.2 Multidrug-resistant TB (MDR-TB), (resistance to isoniazid [INH] and rifampicin [RIF]), impacts transmission, outcome, and health costs. MDR-TB occurs, in about, 20.5% of retreated patients and in 3.5% of new TB cases.2 Besides RIF and INH resistance, cases with extensively drug-resistant TB (XDR-TB) have also resistance to fluoroquinolones (FQ) and second-line injectable drugs (pre-XDR-TB). About 10.2% of MDR-TB cases in São Paulo state presented XDR-TB or pre-XDR-TB.5

Universal access to drug susceptibility testing (DST) represents the standard of care for TB. Phenotypic DSTs (pDST) are based in mycobacterial cultures and require significant infrastructure with slow turnaround time. Considering this scenario, the focus should be on rapid and accurate MDR-TB detection with molecular assays. Paradigms have been broken in TB diagnosis, mainly in the last decade.3,4 Currently, the molecular rapid assay Xpert MTB/RIF (Xpert, Cepheid, Sunnyvale, California) is available in developing countries, including in Brazilian Public Health Care System (SUS) since 2016.6–8 This real-time polymerase chain reaction (PCR) based assay detects the Mycobacterium tuberculosis (Mtb) and resistance to RIF, the backbone of TB regimens. Several genotypic DST have been evaluated to improve the adequacy of M/XDR-TB regimens.

Molecular hybridization assays based on strips, line probe assays (LPAs) are useful and accurate in different settings and infectious diseases. The rationale of these assays is that specific probes for naive and mutant alleles of resistance genes are conjugated with PCR products.9 Commercial LPA is available for detection of MDR-TB (Genotype MTBDRplus assay [Hain Lifescience, CmbH, Germany]). This product has shown 92.2% sensitivity and 99.3% specificity,10 but not for INH. More recently, a hierarchical cascade for Mtb resistance diagnosis has been proposed using LPA, aiming to provide quick decisions on TB treatment.11

According to the WHO, LPAs assays for Mtb resistance detection are recommended to reduce the time of MDR-TB diagnosis. LPAs are applicable for patients with signs and symptoms of TB with AFB positive in sputum directly from clinical specimens.12 For second-line drugs, LPA has shown sensitivity around 97% and 100% specificity even in smear sputum samples.9 This version detects resistance mutations related to FQ, second-line injectable drugs aminoglycosides (amikacin, kanamycin), and glycopeptides (capreomycin).9 FQ resistance is investigated in regions of 83 and 96 codons of gyrA and 538 and 540 of gyrB, representing more than 70% of resistance mutations.13,14 Regarding aminoglycosides, target regions are the codons of rrs gene (1401, 1402, 1484) and some positions of eis gene.15

Herein we evaluate the frequency of pre-XDR-TB and XDR-TB among MDR-TB patients, in addition to assess the pattern of resistance mutations for FQ and the clinical outcome in a regional reference center for MDR-TB.

During the study period 33 patients with MDR-TB were eligible for this study. Of these, five (15.1%) patients were detected with pre-XDR or XDR-TB. Two patients already had XDR phenotypic resistance profiles and three others, in addition to rifampicin and isoniazid, had additional resistance to ofloxacin, pre-XDR-TB.

Patients with pre-XDR-TB or XDR-TB had pulmonary forms of the disease. All patients had a history of previous TB treatment; moreover, two of them had seven courses of previous treatments. On the other hand, the MDR group had an average of 1.8 previous TB treatments. The number of previous treatments was statistically different between the two groups. Despite the onset of the disease and identification of extensive resistance, two patients had negative smear microscopy. Only one patient with pre-XDR-TB or XDR-TB resistance profile was HIV-infected (Table 1).

Overall death rate was in 12%; higher lethality was registered in patients with pre-XDR and XDR-TB (60.0%) compared to MDR-TB (3.6%). Furthermore, 75.0% of MDR patients have completed treatment against 40.0% of the other group.

All 33 isolates from MDR-TB patients were tested for first-line drugs with pDST, except for pyrazinamide. For the second-line drugs, the evaluation was available for 32 patients; ofloxacin resistance was more prevalent, followed by capreomycin (Table 2). Twenty-five isolates of Mtb underwent LPA: four isolates showed FQ resistance genes mutations pattern and one M. tuberculosis isolate had low level resistance to kanamycin (eis gene). For this investigation, three mutations associated with the gyrA gene were found and only one for gyrB gene: gyrA D94G in two patients, gyrA S91P and gyrB N538D (Table 3).

Three cases of disagreement between LPAsl and pDST were as follows: one with FQ resistance gene mutation pattern, but susceptible in the phenotypic assay, and other two patients with a phenotypic pre-XDR result without any genotypic resistance pattern on LPA (Table 4).

WHO has reported that 17% of patients with MDR resistance profile also had additional resistance to FQ.18 In our study, despite the limited number of patients evaluated, a similar prevalence was observed. Gallo et al. found higher rates (26%) of pre-XDR-TB among MDR-TB patients in São Paulo state treated from 2006 to 2013.8

Pre-XDR-TB or XDR-TB cases had many previous treatments as demonstrated by others.19 In this group, with multiple exposures to antimycobacterial regimens, the LPAsl represented a rapid alternative to define the appropriate therapy, together with pDST as proposed by Pan American Health Organization.11 The rapid recognition of FQ resistance demands the incorporation of new drugs like linezolid, bedaquiline, and delamanid into the regimen.

Good agreement of 88% was demonstrated between LPAsl and pDST in our study. Others have reported favorable diagnostic performance of LPAsl. In a multicentric cohort, involving 353 MDR-TB patients, the sensitivity and specificity of LPAsl for FQ compared to phenotypic DST was 80.5% and 100%, respectively.20 In another report, LPA (MDR and sl) and pDST results agreed in 228 of 243 cases (94%).21

Discordant results were observed in three cases: in two false negatives and in one false positive, considering the pDST as the reference standard. Ideally, potential causes of the discrepancies should be analyzed by whole genome sequencing and minimum inhibitory concentration (MIC) determinations. Of note, some authors have considered that pDST likely underestimates the true rate of resistance for key drugs due to high critical concentrations. Multicentric studies should evaluate adequate clinical and epidemiological MIC breakpoints for FQ considering clinical outcomes. In our study, the rate of favorable clinical follow-up was higher among patients with FQ susceptibility compared to FQ resistance. Regarding cross-resistance between different FQ, some in vitro data have shown that Mtb strains have different susceptibility patterns when comparing old and new generation, as well as different MICs.22

In addition, our findings showed genotyping patterns associated with high resistance. Some authors have demonstrated that the MIC for all FQ in isolates that carried gyrA-D94G or gyrB-N538D had significantly higher resistance levels than those isolates with different genotyping.23 A systematic review has demonstrated that gyrA gene mutations are more prevalent in different countries.13

Our study has some limitations: a small number of pre-XDR and XDR-TB was included. Therefore, MICs were not measured, and samples had not undergone genetic sequencing to explore new mutations, which could explain the absence of resistance genotype in LPAsl in correspondence to phenotypic evaluation.

In conclusion, this study observed a few cases of pre-XDR and XDR-TB among an MDR-TB cohort. In some strains, the disagreement was observed between phenotypic methods and LPAsl. Clinical outcome was more favorable for patients with susceptibility to FQ. Evaluation of larger Brazilian cohorts is warranted.

The authors declare no conflicts of interest.