Drug-resistant tuberculosis treatment

Evidence base and analyses. Following close communication with key stakeholders and NTPs, the 
South African Department of Health provided WHO with access to programmatic data on injectable-
free regimens that have been phased in since 2017, when a majority of eligible patients were enrolled 
on a shorter regimen, with bedaquiline replacing the injectable (personal communication, Dr Norbert 
Ndjeka, South African Department of Health, November 2019). In August 2019, WHO issued a public 
call for individual patient data on the use of all-oral shorter regimens of 9–12 months duration (13), 
but this call yielded no additional evidence on the implementation of such regimens. Consequently, 
the evidence review was based primarily on programmatic data from South Africa, recorded in the 
Electronic Drug-Resistant Tuberculosis Register (EDRWeb). Secondary comparative analyses were 
carried out using IPD, to balance the assumptions and adequacy of the data, and adding to the 
generalizability of findings – in particular, the applicability to a global population. As mentioned earlier, 
the IPD is a global dataset of the records of individual patients who have been treated for MDR/
RR-TB; as of November 2019, it contained 13 273 records from 55 studies or centres in 38 countries 
overall. The evidence review focused on the performance of a standardized shorter regimen in which 
the injectable agent was replaced by bedaquiline, in combination with levofloxacin (or moxifloxacin), 
clofazimine, and high-dose isoniazid, ethambutol, pyrazinamide and ethionamide (or prothionamide). 
Patients on this regimen did not receive any injectable agents, nor were they administered cycloserine, 
terizidone, p-aminosalicylic acid, delamanid or linezolid. According to the clinical guidance issued by 
the Department of Health of South Africa, at the time of regimen roll-out (2016–2017), patients were 
not enrolled on the all-oral shorter regimen if they had extensive disease and severe extrapulmonary 
TB, with fluoroquinolone resistance and previous exposure to second-line treatment for more than 
1 month, or had LPA-based DST showing mutations in both inhA and katG
genes.
As part of the primary analysis, the all-oral shorter regimen detailed above was compared with 
the following treatment regimens: standardized shorter regimen with the inclusion of an injectable 
agent; longer regimens in which at least one new TB drug was used, particularly bedaquiline; and 
additional comparison to longer regimens without use of new drugs (based on WHO guidelines 
issued in 2016). No data on all-oral longer regimens recommended by WHO in 2018 were available 
for analysis and comparison. Based on an assessment of the certainty of the evidence, carried out 
using predefined criteria and documented in the GRADEpro software, the certainty of the evidence 
was rated as very
low.
A total of 10 152 records of patients with MDR/RR-TB initiating TB treatment anytime between January 
and June 2017
27
were considered, of which the following were included for primary
28
analyses: 891 
patients who received a shorter all-oral bedaquiline-containing regimen (intervention), 987 patients 
treated with a shorter regimen that included an injectable agent, 1437 patients treated with longer 
(2016) regimens, and 474 treated with longer regimens that included at least bedaquiline. 
27 
Data extraction from EDRWeb was limited to the first semester of 2017, to allow for patients to have reached at least a 2-year post-
treatment follow-up.
28 
Primary analysis was conducted using EDRWeb data from South African patients who started treatment in 2017 and for whom there 
was information on end-of-treatment outcome, and data from the Civil Registration and Vital Statistics database to evaluate or cross-
check mortality
data.

---

Recommendations 
14
The primary analysis comparing South African programmatic data indicated that the use of a shorter 
all-oral bedaquiline-containing regimen in patients with MDR/RR-TB was associated with higher 
treatment success rates (73% all-oral versus 60% standardized shorter regimen success rates, aOR 
for success versus failure/recurrence: 2.1, 95% confidence interval [CI]: 1.1–4.0; aOR success versus 
death: 1.6, 95% CI: 1.2–2.1; aOR success versus failure/recurrence/death: 1.7, 95% CI: 1.3–2.2; and 
aOR success versus all unfavourable outcomes: 1.9, 95% CI: 1.6–2.4); and lower loss to follow-up 
than a standardized shorter regimen in which an injectable agent was used (aOR loss to follow-up 
versus all other outcomes: 0.5, 95% CI: 0.4–0.7). A similar effect for subgroups of patients with acid-
fast bacilli (AFB) smear-positive sputum and HIV-positive and negative patients was observed with 
the use of the shorter all-oral bedaquiline-containing regimen. 
The analysis also suggested that when the shorter all-oral bedaquiline-containing regimen was 
compared to an injectable-free longer
29
regimen containing bedaquiline, there seemed to be no 
marked differences in the outcomes observed. However, relatively modest beneficial effects were 
noted in the direction of the intervention; in particular, success versus failure/recurrence (aOR: 3.9; 
95% CI: 1.7–9.1), success versus all unfavourable outcomes (aOR: 1.6; 95% CI: 1.2–2.2) and loss to 
follow-up (aOR: 0.5; 95% CI: 0.4–0.8), all favouring the use of the all-oral shorter regimen. Further 
subgroup analysis seemed to indicate that there were consistent differences in treatment outcomes, as 
observed in primary analyses among subgroups, in particular among AFB smear-positive patients and 
in HIV-positive individuals on ART; however, differences in treatment outcomes in all-oral shorter and 
longer regimens were no longer significant when looking at outcomes for HIV-negative individuals, 
with the exception of loss to follow-up, which favoured the intervention. The additional comparison 
also illustrated the effect of a shorter all-oral bedaquiline-containing regimen in comparison to longer 
regimens without any new drugs. The all-oral shorter regimen performed significantly better across 
all outcomes and all subgroups in this comparison.