Drug-resistant tuberculosis treatment

GDG considerations. The GDG acknowledged that, during the analysis, the intervention and 
comparator groups were made as comparable as possible. However, the GDG considered possible 
unmeasured confounding due to a lack of systematic collection of information on comorbidities and 
radiological findings through the EDRWeb system, as well as methodological challenges, such as a 
potential selection bias. However, apart from the selection criteria listed, the risk of major selection 
bias was considered low, given that this intervention represented a complete and comprehensive 
switch in the countrywide programmatic approach. 
The GDG further discussed modifications to the shorter all-oral bedaquiline-containing regimen, 
which was adjusted in 2018 by removing ethionamide
30
 and including linezolid for countrywide 
enrolment (50).
31
However, in the patient records for 2017 provided for this guideline development 
process, only 0.5% of patients received linezolid, and no treatment outcomes were available for 
these patients. Given the incomplete data and inability to analyse these data separately owing to the 
small numbers, the GDG decided to exclude all patients who received linezolid. A further sensitivity 
analysis within subgroups defined by the use of specific combinations of drugs was attempted, to 
explore whether the addition of medicines such as linezolid to bedaquiline-containing regimens 
would improve treatment outcomes. No data on this combination were available in the data source 
used for the primary analysis; therefore, longer regimens containing both bedaquiline and linezolid 
were compared to longer regimens in which other companion drugs plus bedaquiline only were 
used. Results of this analysis suggested that regimens containing both bedaquiline and linezolid were 
associated with significantly lower rates of death (aOR: 1.6; 95% CI: 1.1–2.3) as well as a significantly 
better composite outcome of success versus all unfavourable outcomes (aOR: 1.5; 95% CI: 1.1–2.0). 
29 
Recommendations released by WHO in December 2018 emphasized that fully oral longer regimens should be prioritized and become 
the preferred option for most patients, and that injectable agents were no longer among the priority medicines to consider when 
designing longer MDR-TB regimens (11).
30 
The decision to modify this regimen followed results from a National TB Drug Resistance Survey conducted in South Africa in 2012–2014, 
and published in 2018. The survey found that 44.7% of M. tuberculosis isolates were resistant to ethionamide (95% CI: 25.9–63.6%) (49). 
31 
Linezolid was to be included routinely within the regimen up front, to protect bedaquiline in the early stages of treatment, particularly 
in MDR/RR-TB cases where resistance to fluoroquinolones was yet to be
detected.

---

WHO consolidated 
guidelines 
on
tuberculosis: 
drug-resistant tuberculosis treatment
15
Owing to the evidence being insufficient and indirect, the GDG was unable to use these data and 
consider the value of further modification of the all-oral shorter regimen at this
stage.
During the evidence assessment process, members of the GDG further assessed the overall certainty 
in the quality of evidence, the balance between benefits and harms of the shorter all-oral bedaquiline-
containing regimen in addition to treatment outcomes, values and preferences, as well as considerations 
on equity, acceptability and feasibility (51, 52). One of the areas that required further discussion was 
related to potential for confounding and generalizability. Even though confounding was reduced 
through double-adjustment in propensity score matching,
32
 members of the GDG remained concerned 
about the risk of unmeasured or residual confounding and potential selection bias by indication. In 
addition, the GDG acknowledged that although well-collected and unbiased programmatic data hold 
promise and may better reflect real-world practices, these data typically lead to a low-quality grading 
of evidence and have important shortcomings compared with more robust randomized controlled 
trial (RCT) generated data. It is also important to consider the extent to which these findings could 
be applied in other settings; factors that may limit the generalizability of the study findings to other 
settings include high prevalence of HIV, use of ART, specific M. tuberculosis strains and drug-resistance 
patterns, and the quality of health care services, including adherence strategies in South Africa. 
Overall, the GDG agreed that the certainty in the evidence on the efficacy of all-oral shorter regimens 
was “very low” due to concerns about unmeasured or residual confounding and potential risk of bias. 
The GDG considered all outcomes of interest, without any prioritization; the outcome for success versus