4.3 Justification and evidence

The recommendation in this section addresses the following question:

PICO question 10 (MDR/RR-TB, 2020): In XDR-TB patients or patients who are treatment intolerant or with non-responsive MDR-TB, does a treatment regimen lasting 6–9 months composed of bedaquiline, pretomanid and linezolid safely improve outcomes when compared with other regimens conforming to WHO guidelines?

This is a new recommendation for a defined patient group; it is to be used under operational research conditions, and thus does not apply to routine programmatic use. Given these limitations, this recommendation complements other WHO recommendations on the use of longer regimens for patients with MDR/RR-TB (see Section 3) because individuals with MDR/RR-TB and additional fluoroquinolone resistance would usually receive a longer regimen comprising medicines from Groups A, B and C, according to their drug susceptibility profile and other parameters. Several conditions are therefore necessary for the implementation of this new recommendation on BPaL, and these are described in Section 4.5.

Treatment of XDR-TB presents multiple challenges to clinicians and NTPs, because of the limited range of medicines available and the life-threatening nature of the disease. Patients with MDR/RR-TB and additional fluoroquinolone resistance have typically experienced poor treatment outcomes since XDR-TB was first described in 2006 (90). Based on data reported by Member States to WHO, for the cohort of XDR-TB patients who started treatment in 2016 (and for whom treatment outcomes were available in 2018), only 39% completed treatment successfully, while 26% died, treatment failed for 18% and an additional 17% were lost to follow-up or were not evaluated (1).

Evidence base and analyses. The pressing need for more effective treatment regimens for patients with extensive drug resistance, including fluoroquinolone resistance and more extensive drug-resistance profiles, has been the driver for a number of studies and initiatives to test more effective and novel treatment regimens, including newer and repurposed medicines. One such study is the Nix-TB study, conducted by TB Alliance. The Nix-TB study was a one-arm, Phase III, open-label observational cohort study that assessed the safety, efficacy, tolerability and pharmacokinetic properties of a 6-month BPaL treatment regimen, extendable to 9 months for those who missed doses, or who remained culture positive or reverted from culture negative to positive between months 4 and 6 of treatment (89). The study was conducted between 2014 and 2019 at three study sites, all in South Africa, with the first patient enrolled in April 2015. Eligible patients were aged 14 years and above, weighed at least 35 kg, had a documented HIV result and had bacteriologically confirmed sputum culture positive XDR-TB or bacteriologically confirmed MDR/RR-TB, but were treatment intolerant or non-responsive to previous MDR/RR-TB treatment. A number of other inclusion criteria were applied (Box 2).


Patients were followed up for a period of up to 24 months after completion of treatment. The primary outcome measure was the incidence of bacteriological failure, relapse or clinical failure (including TB-related deaths) through follow-up until 6 months after the end of treatment. Secondary outcome measures comprised:

• incidence of bacteriological failure or relapse or clinical failure through follow-up until 24 months after the end of treatment (as a confirmatory analysis);
• time to sputum culture conversion to negative status through the treatment period;
• the proportion of subjects with sputum culture conversion to negative status at 4, 6, 8, 12, 16 and 26 or 39 weeks;
• linezolid dosing (actual) and efficacy;
• change from baseline in TB symptoms;
• change from baseline in patient reported health status; and
• change from baseline in weight.

The Nix-TB study regimen comprised pretomanid administered at 200 mg once daily, bedaquiline administered at 400 mg once daily for the first 2 weeks of treatment (days 1–14) and then 200 mg three times a week thereafter, and linezolid commenced at 1200 mg per day (additional information on linezolid dosing is included in Section 4.5) (89). Close microbiological, clinical and adverse event monitoring were features of the Nix-TB study (89).

The evidence to inform PICO question 10 was derived from the Nix-TB study, and included information on 108 patients. The total study population was 109 patients, but one patient withdrew informed consent to participate in the study; this person was included in safety analyses but not in the analyses for effectiveness. These data were compared to a subset of data from the IPD, which overall included 13 273 individual patient records from 55 different studies or centres in 38 countries. For the primary analyses, the comparator group included patients from the IPD on longer treatment regimens (with a mean duration of treatment of 21.0–25.5 months), who received both bedaquiline and linezolid as part of the regimen (no patients received pretomanid in the IPD). This comparison group included data of 456 patients reported from Belarus, India, France, Russian Federation and South Africa, and from one study in a group of countries.⁴⁹ The intervention and comparison groups were matched exactly for XDR, MDR and fluoroquinolone resistance, and for HIV status, with propensity score matching for the variables of age, sex, baseline culture result, extent of disease (determined by baseline AFB smear or chest X-ray findings of cavitation or bilateral disease if the AFB smear result was missing) and country income level (using the World Bank Atlas method (91)). Treatment outcomes used in these analyses comprised the investigator-defined outcomes for the intervention group (i.e. for the Nix-TB study) and those largely based on WHO definitions⁵⁰ for the comparator group (i.e. for the patients included in the IPD). To allow an equal opportunity for treatment outcomes to occur from the start of treatment when comparing the two groups, all outcomes were included from the start of treatment to 24 months after the start of treatment. Thus, in the intervention group, these outcomes occurred after completion of treatment and in the comparator group the outcomes were end-oftreatment outcomes (because patients in the IPD received a longer regimen and were not followed up after completion of treatment). Three other comparator groups from the IPD included patients on longer treatment regimens who received one of the following: a regimen that included bedaquiline, a regimen that included linezolid, or a regimen that included neither bedaquiline nor linezolid. The GDG’s initial intention was to assess the intervention regimen against all three comparison groups; however, during their deliberations, the panel agreed that the judgements should be based on the comparison group who received bedaquiline and linezolid as part of their regimen, because these patients most closely resemble patients who would receive currently recommended longer regimens composed with medicines from Groups A, B and C. However, a direct comparison of BPaL with alloral longer regimens constructed according to the most recent WHO recommendations issued in March 2019 was not possible, because these regimens may have only been in use since mid-2019, so treatment outcomes for these patients are not yet available. Based on an assessment of the certainty of the evidence, carried out using predefined criteria and documented in GRADEpro, the certainty of the evidence was rated as very low.

Additional data reviewed by the GDG relevant to PICO question 10 were a cost–effectiveness analysis, a study on the acceptability and likelihood of implementation of the BPaL regimen, modelled pharmacokinetic data based on the development of a pharmacokinetic toxicodynamic model, and a summary review of preclinical and early clinical data on pretomanid. The cost–effectiveness analysis, acceptability study and modelled pharmacokinetic studies were conducted along with the Nix-TB study and were sponsored by TB Alliance.

The aims of the cost–effectiveness study were twofold:

• to estimate the cost–effectiveness of a new regimen (BPaL) compared with the standard of care at a given price; and
• to estimate the maximum drug price at which the BPaL regimen could be considered cost effective or cost neutral in each setting.

There were two patient populations: XDR-TB patients; and XDR-TB patients combined with MDR-TB patients who failed treatment or who were treatment intolerant. The comparison treatment regimen was the standard of care in Georgia, the Philippines and South Africa, and the main outcome of interest was the incremental cost per disability adjusted life year (DALY) averted. A health service perspective was taken.

The objective of the acceptability study was to assess the acceptability and likelihood of implementation of the BPaL regimen as anticipated by key stakeholders based on a number of criteria including perceived benefits and challenges of implementation of BPaL and also longer individualized treatment regimens and other practical requirements of implementation. The study was a mixed-methods, multicountry, cross-sectional study conducted in 2018–2019 among stakeholders in Indonesia, Kyrgyzstan and Nigeria. A total of 188 participants offered their views; they included caregivers, programmatic stakeholders (including national and international programmatic stakeholders and patient advocacy groups), and public and private laboratory stakeholders.

Additional evidence presented included modelled pharmacokinetic data based on the development of a pharmacokinetic toxicodynamic model designed to quantify the relationship between pharmacokinetic features of linezolid and toxicity as part of a 6-month BPaL regimen. Modelled data from 88 patients who received linezolid as part of the Nix-TB study were presented. This patient population included information on seven patients who had died. A total of 21 individuals from the Nix-TB study were excluded from these analyses – 16 because they had incomplete dosing histories (i.e. they were receiving ongoing treatment at the time of analysis) and five because they had an unverifiable dosing history.

Additionally, the GDG was presented with an independent summary review of the preclinical and early clinical data on pretomanid. This review included background information, preclinical data and early phase clinical data detailing safety and efficacy, including data submitted to the United States Food and Drug Administration (US FDA) as part of the original new drug application or as supplemental information.

GDG considerations. The GDG considered the desirable effect of treatment success, which was higher in the intervention group than in the comparator, for all four treatment outcomes assessed. Overall, when comparing treatment success versus failure or recurrence, the treatment success rate was 97.0% in the Nix-TB study and 91.7% in the comparator group (resulting in 6 more outcomes of treatment success per 100 patients). For the comparison of treatment success versus death, treatment success was 93.2% in the Nix-TB study and 91.9% in the comparator group (resulting in 1 more outcome of treatment success per 100 patients). For the comparisons of treatment success versus failure, recurrence or death, and treatment success versus all unfavourable outcomes combined (i.e. failure, relapse or death, and loss to follow-up) the proportions of patients with treatment success in the intervention and comparator groups were 90.5% versus 84.8% (6 more outcomes of treatment success per 100 patients) and 88.9% versus 82.2% (2 more outcomes of treatment success per 100 patients), respectively. Based on these figures, the primary analysis yielded adjusted odds ratios of 3.3 for treatment success (versus the combined outcome of failure and recurrence; 95% Cl: 0.8– 13.7), 1.0 for success versus death (95% Cl: 0.1–8.2), 1.8 for success versus failure, relapse or death (95% Cl: 0.7–4.4), and 1.2 for success versus all unfavourable outcomes (95% Cl: 0.5–3.1), with a mean duration of follow-up of 24 months (range 21.0–25.5 months), when BPaL was compared to longer regimens containing bedaquiline and linezolid. The GDG considered lower rates of loss to follow-up as a desirable effect; the proportion of patients who were lost to follow-up was lower in the intervention (BPaL) group (1.8%) than in the comparison group (3.1%); however, this difference was not considered by the panel to be large. The panel also considered that a shortened duration of treatment and less drug exposure were desirable effects of the intervention, and that both of these were components of the overall burden of a given MDR/RR-TB treatment regimen, which may not be wholly reflected in rates of loss to follow-up alone. Subgroup analyses could not be undertaken because of the small sample size.

The BPaL regimen was also associated with a high rate of adverse events, relative to the study drugs, which was a concern for GDG members. Of the 109 patients in the Nix-TB study, 28 (25.7%) experienced at least one serious adverse event. This included one (0.9%) death related to acute haemorrhagic pancreatitis, 29 (26.6%) other serious adverse events including hospitalizations and lifethreatening events, and two (1.8%) adverse events that resulted in persistent or significant disability or incapacity. A total of 53 (49%) patients experienced at least one Grade 3–4 adverse event considered to be related to the study drugs; these comprised 25 with peripheral neuropathy (11 resolved), 16 with increased hepatic transaminases (13 resolved), nine with haematologic adverse events (all resolved), eight with increased pancreatic enzymes (7 resolved) and two with optic neuritis (both resolved). This led to all three drugs being discontinued in one patient, and linezolid (initial dose 1200 mg/day) being discontinued in another 35 (32%) patients. Only 20 (18%) patients completed a full course of linezolid at 1200 mg/day. The GDG discussed the difficulty of comparing these adverse event rates with other studies, owing to major and important differences in ascertaining, assessing and reporting adverse events. However, in the IPD studies (where 90% of patients received a linezolid dose of ≤600 mg/day), the pooled rate of permanent discontinuation of linezolid was 17.9%, and in the endTB observational study (where all patients received a linezolid dose of ≤600 mg/day), the rate of linezolid discontinuation was 13.1%. In both of these studies, more than 80% of patients received a starting dose of linezolid of 600 mg per day. In preliminary analyses of the endTB observational study, nine patients out of 1094 (0.8%) died of an adverse event that was possibly or probably drug related, including two individuals with sudden cardiac death; these two patients were receiving bedaquiline, clofazimine, capreomycin and p-aminosalicylic acid, and both had hypokalaemia.

Information from the independent review on the preclinical and early phase clinical data highlighted that pretomanid possesses activity against replicating and non-replicating bacilli that is both concentration and dose dependent. Safety signals were comprehensively described, many of which were observed at exposures that are higher than would be used in humans; however, safety signals of note included liver toxicities (hypertrophy of hepatocytes, transaminase elevation and increased liver weight, observed at higher doses in rodents and lower doses in monkeys) and reproductive toxicities in males observed in animal (murine and simian) models, which appear to be both time and dose dependent. The observations in monkeys may have been due to the general decline of health in these animals; however, the same signals were observed in rodent models, with some evidence that these effects may be irreversible. In mouse models, these effects were observed at exposures that would be used in humans, and in females, reproductive toxicities were also observed.

Additional information on adverse events presented to the GDG included the results of a pharmacokinetic toxicodynamic model (Savic R, unpublished data, University of California, San Francisco, November 2019). Based on these data, it was concluded that the pharmacokinetics of linezolid are nonlinear in XDR-TB patients, and that individual linezolid concentration–time profiles are the best predictor of toxicity. Higher toxicity rates were observed at higher total daily doses, with comparable toxicity rates for BID (twice a day) and QD (once a day) dosing schedules. The results of the modelled data highlighted that anaemia can be managed by closely monitoring changes in haemoglobin over the first 4 weeks of treatment (in particular, changes in haemoglobin >10% decrease from baseline should trigger a reduction in the dose of linezolid; haemoglobin levels recover well after dose reductions). Thrombocytopenia was not a major concern. The study investigators recommended that peripheral neuropathy should be closely monitored; based on the modelled data, when peripheral neuropathy did occur, for most patients, it was reversible within 3 months. The GDG was concerned that the two studies had different findings with regard to reversibility of peripheral neuropathy (i.e. one study found that it was largely reversible, the other study had different findings).

The panel considered at length the desirable and undesirable effects, and the balance of these effects, noting the very low certainty of the evidence and acknowledging the efforts to match patients in the intervention and comparator groups (through exact and propensity score-based matching). Overall, the GDG agreed that the BPaL regimen showed high rates of treatment success when used in XDR-TB patients in South Africa, and they noted the additional positive treatment outcomes resulting from the use of the regimen when compared with patients receiving longer regimens with bedaquiline and linezolid. However, there were important residual concerns about the likelihood and severity of adverse events, possible reproductive toxicity signals in the preclinical data, limitations in the study design and the overall very low certainty of the evidence. The GDG was concerned that the Nix-TB study was a one-arm study with no in-study comparison group, including 109 participants overall recruited from one country setting (South Africa), with the inclusion of a group of patients who were sputum culture negative at baseline, and many patients receiving second-line treatment of variable duration before being enrolled on BPaL. The GDG agreed that this might limit the generalizability of the study findings to all populations and to all regions.

The GDG was concerned about issue of serious adverse events, particularly those related to linezolid, and pretomanid-associated potential safety signals related to male infertility observed in animal models (murine and simian). They highlighted the potential difficulties in monitoring infertility in a programmatic setting. Additional human sperm studies recommended by the US FDA will be carried out by TB Alliance; however, these data were not available for the GDG to consider at the time of the meeting. The GDG determined that infertility is a serious issue because it affects both patients and their families. The GDG also acknowledged that, at the time the Nix-TB study started, treatment options for patients with MDR-TB and additional fluoroquinolone resistance were limited in South Africa, with an associated high case fatality rate, which meant that patients may have placed a different value on potential male infertility at that time than they might now. The judgement about the balance of the desirable and undesirable effects was therefore assessed as not favouring the intervention or the comparison.

With regard to how much patients might value the outcomes, no research evidence was available for the GDG to consider. Although no research evidence was identified, the GDG felt that there was possibly important uncertainty or variability in how much people would value the main outcomes. Fertility was raised as an outcome for which there is less information at the present time. The GDG thought that this issue would increase the complexity of the judgement about how much people would value the outcomes. They discussed the fact that there are other safety outcomes (e.g. other adverse events, including peripheral neuropathy) that may vary, and that people may value these outcomes differently. The GDG considered indirect evidence in the form of a separate qualitative study conducted among 16 drug-resistant TB patients from high burden countries (which informed the judgements on PICO question 2, described in Section 2). The aim of the study was to determine the most acceptable treatment regimen for drug-resistant TB from the patient perspective. Based on the results of this study, the preferred MDR/RR-TB treatment regimen from the perspective of the patients who were interviewed was a short, injectable-free regimen with few to no physical or mental health side-effects, and a low pill burden.

The cost–effectiveness study determined that the use of BPaL for the treatment of XDR-TB is likely to be cost saving at the proposed price (US$ 364 per treatment course for pretomanid), in the settings in which it was studied. The study found that cost savings are a function of the cost of care and the magnitude of XDR-TB burden, and are about US$ 4490 (not including ART costs) in South Africa, US$ 4060 in Georgia and US$ 3860 in the Philippines. In high HIV-TB prevalence settings, such as South Africa, related future costs such as those from the HIV programme (ART costs) reduce the magnitude of expected cost savings to US$ 1400 per patient. Overall, when BPaL is introduced to a larger population (including MDR/RR-TB treatment failure and people with MDR-TB who are treatment intolerant), the GDG observed an increase in the incremental benefits, both in terms of deaths and DALYs averted, and incremental costs. The study investigators concluded that the impact of BPaL on costs and DALYs averted will depend on the overall performance of the NTP (e.g. considering rates of loss to follow-up or mortality). The GDG noted that food costs were not included in the cost– effectiveness analysis, and would need to be considered when BPaL is implemented. During the Nix-TB study, all study medications were administered with food (attributable to the administration of bedaquiline, which will also now feature as a core medicine in longer treatment regimens for MDR/ RR-TB). The GDG noted that the cost–effectiveness analysis was not specific to the other evidence that was presented; most importantly, the undesirable effects were not considered according to the GDG’s judgements. This made the cost–effectiveness analyses informative, but not directly based on the evidence that the GDG was assessing.

Although no research evidence was identified on equity, the GDG felt that the impact on health equity would be a probable increase, given the option of a shorter regimen that could be available to all patients, globally. The GDF identified a price for pretomanid and stated that it will make the BPaL regimen available to NTPs, depending on the recommendation from the panel. Children and pregnant women were ineligible for inclusion in the Nix-TB study, and therefore will not be an eligible population for BPaL at the present time.

Acceptability was higher for BPaL than for the longer MDR-TB regimens in six of the seven acceptability categories assessed (ranging from patient friendliness to treatment safety monitoring), with the exception of treatment safety monitoring, where the difference was negligible. The main drivers of the acceptability of BPaL were the shorter duration, absence of injectables, lower pill burden, anticipated patient preferences and lower financial burden for patients, anticipated higher treatment success, lower costs for the health system, minimal additional requirements for diagnostic processes, and a lower anticipated per patient burden to TB laboratories for bacteriological treatment monitoring. The GDG thought that the intervention was probably acceptable to key stakeholders. There were residual concerns about the acceptability of the intervention, given the potential reproductive toxicities (which were not specifically discussed with study participants because they were not known to the study investigators at the time) and the comparator that was used in this study (which was not the same as the comparator assessed by the GDG). Therefore, the GDG felt that the acceptability study was informative, but was not based directly on the evidence that was assessed.

In line with the overall high acceptability of the BPaL regimen, the likelihood of implementing the BPaL regimen as a standard of care for the treatment of patients with XDR-TB and MDR-TB treatment failure or persons with MDR-TB who are treatment intolerant scored at 88%; only 1% of the 166 participants scored the implementation as “unlikely”, and 11% had a neutral opinion. The likelihood of implementation was similar for the BPaL regimen as a standard of care for MDR-TB patients with fluoroquinolone resistance, with a score of 84%, regardless of additional resistance to second-line injectables. This indicated that those interviewed generally felt that it may be feasible to implement the regimen.

The GDG noted that more research studies are needed in the future, which may allow a more closely aligned comparison group, with less potential for residual confounding and more certainty in the evidence overall. The GDG also emphasized that it will be necessary for WHO to review, revise or update this guidance, as additional substantive data on the efficacy and safety of the BPaL regimen become available. Therefore, as the evidence base on the use of BPaL increases, researchers and NTPs are encouraged to make these data available for international policy-making.

Unpublished data. After the GDG meeting, the TB Alliance shared with WHO an additional report (unpublished and confidential at the time of the release of the current guidelines) that included comparison of data from the Nix-TB study with a cohort of patients with XDR-TB, treated at one of the study sites. The comparison group in these analyses comprised 102 XDR-TB patients treated in a programmatic setting in South Africa, between November 2013 and April 2016. All patients in the comparison group had laboratory-confirmed XDR-TB, with TB isolates resistant to rifampicin, isoniazid, ofloxacin and amikacin. About half of the patients in the comparison group had HIV infection (51.0%). These comparative analyses were carried out by a data analyst under contract to the TB Alliance, who used a methodology that differed from that used for the analyses that informed the WHO recommendation on the use of BPaL presented in these guidelines, and no IPD were provided. Thus, the comparative analyses are not directly comparable to the analyses presented in Annex 3. Although this report could not be considered by the GDG during the November 2019 meeting (given that is was not made available until after the meeting), it was reviewed by the WHO Guideline Steering Group and the GDG before finalizing the guidelines document. The comparison in this report reinforces the positive treatment success rates that have been observed with the use of the BPaL regimen. In these analyses, the treatment success rate for patients receiving BPaL was 89.9% (98/109) versus 66.7% (56/84) for patients receiving another XDR-TB regimen that also contained bedaquiline and linezolid, yielding an adjusted risk ratio of 1.31 (95% Cl: 1.11–1.55, P=0.0012). The detailed data that supported this comparison, together with the results of operational research in other settings, may be useful for informing policy recommendations on the use of BPaL in the future. 

⁴⁹ China, Philippines, Republic of Korea, Russian Federation, Thailand.

⁵⁰ These treatment outcomes conform to the definitions outlined in the paper by Laserson K.F. et al. (2005) (93) or in the WHO publication: Definitions and reporting framework for tuberculosis – 2013 revision (94).

Book navigation