Abstract
This systematic review aimed to update the current evidence for multidrug-resistant tuberculosis (MDR-TB) treatment.
We searched for studies that reported treatment information and clinical characteristics for at least 25 patients with microbiologically confirmed pulmonary MDR-TB and either end of treatment outcomes, 6-month culture conversion or severe adverse events (SAEs). We assessed the association of these outcomes with patients' characteristics or treatment parameters. We identified 74 studies, including 17 494 participants.
The pooled treatment success was 26% in extensively drug-resistant TB (XDR-TB) patients and 60% in MDR-TB patients. Treatment parameters such as number or duration and individual drugs were not associated with improved 6-month sputum culture conversion or end of treatment outcomes. However, MDR-TB patients that received individualised regimens had higher success than patients who received standardised regimens (64% versus 52%; p<0.0.01). When reports from 20 cohorts were pooled, proportions of SAE ranged from 0.5% attributed to ethambutol to 12.2% attributed to para-aminosalicylic acid. The lack of significant associations of treatment outcomes with specific drugs or regimens may reflect the limitations of pooling the data rather than a true lack of differences in efficacy of regimens or individual drugs.
This analysis highlights the need for stronger evidence for treatment of MDR-TB from better-designed and reported studies.
Abstract
There are many recent published studies, but these provide only very weak evidence on how to improve MDR-TB treatment http://ow.ly/zQa1308SSOi
Introduction
Multidrug-resistant tuberculosis (MDR-TB), defined as TB resistant to at least isoniazid (INH) and rifampin (RIF), and extensively drug-resistant TB (XDR-TB), defined as resistance to INH and RIF plus at least one fluoroquinolone and one second-line injectable drug, have become major public health threats [1]. The World Health Organization (WHO) has estimated that, in 2015, approximately 580 000 people developed MDR-TB, of whom 55 000 had XDR-TB [2].
Treatment for MDR-TB or XDR-TB requires lengthy use of second-line TB drugs, although the regimens used vary widely due to differences in opinions as well as the resources available [3, 4] To date, there are few published phase 3 randomised clinical trials (RCTs) for MDR-TB treatment. Hence, systematic reviews [5–7] and individual patient data (IPD) meta-analyses [8] have provided the majority of evidence for MDR-TB/XDR-TB treatment [9]. However, this meta-analysis only included studies published up until 2008 [8], and since then a large number of studies with new drugs and new regimens for MDR-TB/XDR-TB have been published. We therefore performed this systematic review to update the evidence for MDR-TB treatment to inform the WHO Guideline Development Group.
Methods
Literature search and study selection
The PICO (Patients, Intervention, Comparator and Outcomes) questions were developed by a WHO Guideline Development Group to answer specific questions regarding MDR-TB treatment (see supplementary material for the PICO questions). The main focus of this review was the efficacy and safety of the drugs available for treatment of MDR-TB patients. The following groups of drugs were analysed: first-line drugs (pyrazinamide and ethambutol), injectable drugs (streptomycin, kanamycin, amikacin and capreomycin), fluoroquinolones (ofloxacin, levofloxacin and moxifloxacin), add-on agents (ethionamide/prothionamide, cycloserine, para-aminosalicylic acid (PAS) and high-dose isoniazid) and bedaquiline. Four independent systematic reviews [10–13] have been conducted recently for the group 5 drugs (renamed as “add-on agents” by WHO), so we did not include these drugs specifically in our search. However, if a study reported the use of any group 5 drugs, we abstracted the information.
We searched MEDLINE (through OVID), EMBASE (through OVID) and The Cochrane Library. The search strategy used a combination of Medical Subject Heading terms and free-text words in titles, abstracts and key words. Terms related to MDR- or XDR-TB, drugs and treatment outcomes were included (supplementary material). Because this is an update from previous reviews that included studies published up to December 2008, our search was limited to the period from January 2009 to August 2015.
Titles, abstracts and full texts were screened by two reviewers (M.L. Bastos and Z. Lan), with consensus in each stage. A third reviewer (D. Menzies) was consulted to resolve disagreements. We included studies published in English, French, Chinese, Portuguese and Spanish. All studies that met the following inclusion criteria were selected: 1) MDR-TB confirmed by phenotypic tests (GeneXpert® alone for diagnosis of MDR-TB was not considered adequate unless confirmed by phenotypic tests), 2) pulmonary TB, 3) cohorts or RCTs with a minimum of 25 patients treated, 4) a clear description of the regimen and the drugs received, and 5) at least one of the following reported: end of treatment outcomes; 6-month sputum culture conversion; treatment adverse events. Studies that evaluated short regimens (<18 months) were excluded, as these have been reviewed elsewhere [14]. For studies that reported patients with extra-pulmonary disease, due to the difficulty of microbiological confirmation for the initial diagnosis and the even greater difficulty of confirmation of cure at the end of treatment, we excluded studies in which more than 10% of patients had extra-pulmonary disease, and did not report results stratified by disease site.
Data abstraction
We recorded information about age, sex, HIV (and use of antiretroviral treatment), acid-fast bacillus smear results, chest radiography cavitation, prior TB treatment (with first-line drugs or second-line drugs), drug susceptibility test results, number of patients that received each drug, duration of treatment, and whether the regimen was standardised or individualised. Outcomes abstracted included end of treatment outcomes, 6-month sputum culture conversion, and severe adverse events defined as grade 3–4 events, or defined operationally as events resulting in permanent discontinuation of a drug.
Quality assessment
There are no validated criteria for evaluating quality in MDR-TB studies, so we developed a checklist based on the presence of 12 indicators, grouped into four major categories:
1) Diagnostic information (three items): i) reported methods of confirmation of TB and of MDR-TB; ii) reported results of drug susceptibility testing for ethambutol or pyrazinamide; and iii) reported results of drug susceptibility testing for fluoroquinolones or second-line injectable.
2) Treatment regimen information (five items): i) duration of intensive phase; ii) number of drugs used in intensive phase; iii) duration of continuous phase; iv) number of drugs used in continuous phase; and v) dosage of drugs used.
3) Adverse event information (two items): i) provided a definition of adverse event (e.g. graded, classified by severity); and ii) the drug considered related to the adverse event was identified.
4) End of treatment information (two items): i) if the end of treatment outcomes were defined using Laserson et al. [15] or WHO [1] definitions (2013); and ii) if default rate was ≤8% (this threshold for quality was calculated by subtracting the pooled estimation of fail/relapse and death (17%) of all cohorts included in the review, from the WHO predefined target of a total of 25% non-success for MDR patients) [16].
The score was an unweighted sum of the 12 indicators.
Data synthesis and statistical analysis
For end of treatment outcomes, we compared success (defined as cured or treatment completed) to 1) failure or relapse or 2) failure or relapse or death. Analysis was stratified according to resistance pattern (MDR or XDR) and whether the regimen was individualised or standardised. In additional analyses, we examined the relationship of the pooled success and the prevalence of additional resistance to the second-line injectable or fluoroquinolones (i.e. pre-XDR-TB patients) among the MDR patients. We examined the relationship between each effectiveness end-point (end of treatment outcome and 6-month sputum culture conversion) and the number of patients receiving each specific drug, the average number of drugs used and duration of treatment, as well as the average value, for each cohort, of the major clinical and demographic characteristics of the patients. If HIV or age were missing, values were estimated using information from other included studies from the same country, or if this was not available, from data published by the World Bank [17] or WHO [18]. Variables were categorised according to the distribution observed (median, terciles or quartiles).
The occurrence of severe adverse events was pooled across studies that reported the following: the drug related to the event; events classified as grade 3 or 4 severity; or permanent discontinuation of a drug related to the event.
All statistical analyses were performed using SAS (version 9.2 Institute, Cary, NC, USA). Linear mixed models were used to pool the proportion with events and generalised linear mixed model for pooling adverse events.
Results
Description of studies
As shown in figure 1, 2336 titles were identified and, after eliminating duplicates and non-relevant publications based on a review of titles and abstracts, 250 were selected for full text review. In total, 176 studies were excluded, including 24 classified as “MDR-TB not confirmed”. These 24 studies reported cohorts in which outcomes of patients with MDR-TB were reported together with outcomes of patients with fully susceptible TB, or poly-drug resistant, or non-confirmed MDR-TB (“suspicious of drug resistant TB”). None was excluded due to reliance only on GenXpert results.
PRISMA diagram of study selections in the review (and guide to applicable tables). TB: tuberculosis; MDR: multidrug-resistant; XDR: extensively drug-resistant; AE: adverse event.
Of the 74 studies that met the review inclusion criteria [19–92], seven [27, 34, 36, 51, 60, 66, 68] reported more than one cohort, yielding a total of 84 cohorts with 17 494 patients with MDR- or XDR-TB. Of these, 64 studies reported microbiological outcomes at the end of treatment and/or 6-month sputum culture conversion, and 44 reported adverse events. Of the latter group, only 19 studies reported events that were classified as grade 3 or 4, or required permanent discontinuation of the drug and identified the drug considered related.
Of the 74 studies, seven reported results in patients with XDR-TB only. 14 (13 cohorts) studies included patients with MDR-TB and XDR-TB, and treatment outcomes (one cohort reported 6-month sputum culture conversion and 12 cohorts reported end of treatment) were reported separately for these two groups of patients. Seven studies reported only MDR-TB patients, and 46 studies (57 cohorts) reported MDR-TB patients but did not provide information regarding drug susceptibility testing to fluoroquinolones and second-line injectable. Hence, some patients in these studies may have had XDR-TB. As seen in table 1, of the 17 494 patients, 4623 (66% of those with information) had a history of prior treatment with first-line drugs, 5088 (55% with chest radiography information) had cavitation on chest radiograph and 6057 (69% with acid-fast bacilli (AFB) results) were AFB sputum smear positive. Only 3111 (19% of tested) had HIV co-infection, of whom 1311 (42%) were on anti-retroviral treatment. More detailed study design, demographic, clinical treatment and outcome information are summarised in supplementary tables S1–S7.
Summary of available clinical treatment and outcome information for multidrug-resistant and extensively drug-resistant tuberculosis (MDR- and XDR-TB) patients
Study quality
All studies were assessed for quality based on the reporting of essential diagnostics, treatment and outcome information. As seen in table 2, only half to two-thirds of studies reported each of the three diagnostic information items considered essential and, only one-third of studies reported all three items adequately. Only 20–39% of studies reported each of the five treatment information items, and only two studies (3%) reported all five items adequately. Of the two items considered essential for adverse events, only half of the studies who reported any adverse events provided both. On the other hand, 51 of 54 studies reported end of treatment outcomes, defined as recommended by WHO [1] or Laserson et al. [15]. Only 25% of studies achieved a lost to follow-up rate less than 8%. Supplementary table S2 provides details about these criteria.
Quality of the studies included in the review
Treatment outcomes and correlates
As shown in table 1, the pooled treatment success was 26% (95% CI 23–30%) in XDR-TB patients, compared to 60% in all cohorts of MDR-TB patients (with or without “pre-XDR”). In the studies providing this information, studies with a higher proportion of pre-XDR-TB patients reported similar outcomes to all other studies of MDR-TB patients (supplementary table S8), so these results were pooled together. Because the outcomes for XDR-TB patients were substantially worse than in MDR-TB patients, the XDR-TB patients were analysed separately from MDR-TB patients in the end of treatment outcome and in 6-month sputum culture conversion.
The pooled treatment success for MDR-TB patients that received individualised regimens was significantly higher when compared with patients who received standardised regimens (64% versus 52%; table 1). As shown in table 3, in 61 cohorts of MDR-TB patients, the end of treatment outcomes were not associated with any patients' characteristics or treatment parameters, including duration, number of drugs or individual drugs. When the same analysis was stratified in cohorts of MDR-TB patients who received only standardised or only individualised regimens, similar results were found (supplementary tables S9 and S10). Similar results were found when we analysed failure separately from relapse (data not shown).
Covariates associated with end of treatment (EOT) outcomes in the 61 cohorts of multidrug-resistant tuberculosis patients (extensively drug-resistant tuberculosis (XDR-TB) excluded)1
No treatment or patient characteristics examined were associated with end of treatment outcomes in the 15 cohorts of XDR-TB patients (supplementary table S11).
The pooled estimation of the 6-month sputum culture conversion was 69% in MDR-TB patients and 19% in XDR-TB patients (table 1). As shown in table 4, in 16 cohorts of MDR patients, sputum conversion was not associated with any of the reported patients' characteristics or treatment parameters. We could not assess the relationship between treatment regimens, clinical characteristics and 6-month sputum culture conversion for XDR-TB patients due to unstable results when we pooled the data.
Covariates associated with 6-month sputum culture conversion in the 16 cohorts of multidrug-resistant tuberculosis patients (extensively drug-resistant tuberculosis (XDR-TB) excluded)1
Only 19 studies (20 cohorts) met the criteria for pooling severe adverse event data, and 86% of the patients included in these cohorts had received individualised regimens (supplementary table S7). All 20 cohorts reported the number of patients that had adverse events and not the number of events, so the denominators used in adverse event analysis were the number of patients that received the specific drug.
As shown in table 5, the occurrence of severe adverse events ranged from 0.5% of 1325 patients receiving ethambutol to 12.2% of 1706 patients who received PAS. Fewer than 3% of patients receiving fluoroquinolones or pyrazinamide experienced an severe adverse event, compared to more than 5% of patients receiving second-line injectables or a thiamide (ethionamide or prothionamide). Detailed information on the reporting of adverse events is provided in supplementary tables S6 and S7.
Occurrence of severe adverse events (SAEs), attributed to specific drugs in the treatment of multidrug-resistant or extensively drug-resistant tuberculosis#
Discussion
This review identified 74 studies, with 84 distinct cohorts, published since January 2009, that reported treatment regimens and outcomes in 17 494 MDR-TB and XDR-TB patients. These studies have reported adverse events, 6-month sputum culture conversion, and end of treatment outcomes. The pooled overall success of MDR patients was 60%: similar to previously reported studies [5–8] and well below the WHO target of 75% [16]. Treatment outcomes were substantially worse in patients with XDR-TB and in patients who received standardised regimens for MDR-TB.
However, despite the large number of studies and patients, no other treatment parameter, including number or duration of drugs and individual drugs, were associated with 6-month culture conversion, or end of treatment outcomes. This lack of association in many treatment parameters (use of any individual drugs, treatment length, number of drugs) and clinical characteristics (such as HIV co-infection) was also observed in three previous systematic reviews [5–7]. This may reflect the limitations and difficulties of pooling the data rather than a true lack of differences in efficacy of regimens or individual drugs. This limited pooling to simple characterisation, such as stratifying analyses at the median value for proportion receiving a certain drug. This crude characterisation resulted in misclassification of exposure for many patients treated with individualised regimens and reduced our chances of finding any effects, even if present.
This review highlights the need for more standardised reporting as well as evidence from well-designed randomised trials, or from meta-analysis of pooled individual patient data from multiple observational studies.
This review has a number of strengths, the most important being the identification of a large number of studies, published within the past 7 years, describing three important outcomes of MDR-TB treatment: 6-month sputum culture conversion, severe drug-related adverse events, and end of treatment outcomes. This comprehensive aggregate meta-analysis was used to update the WHO recommendation [93] of MDR-TB treatment. The treatment regimen should include four core second-line drugs: one from group A (fluoroquinolones), one from group B (second-line injectable agents) and two from group C (ethionamide/prothionamide, cycloserine/terizidone, linezolid, clofazimine), plus pyrazinamide. If there is clinical or in vitro evidence of resistance to these drugs, other add-on agents from group D2 (bedaquiline or delamanid) and group D3 (PAS, imipenem–cilastatin/meropenem, amoxicillin–clavulanate) can be added, preferably drugs from group D2 [93].
However, our study had a number of important limitations, the most important being that almost all included studies were observational, and the majority described results with individualised treatment regimens. Selection bias is an important limitation for these studies of individualised regimens, because sicker patients with more extensive disease or drug resistance may have been more likely to receive certain drugs such as later-generation fluoroquinolones. Additionally, the 23 cohorts describing standardised regimens used very different regimens (supplementary table S4), so we could not pool by different types of standardised regimens.
Another major limitation was the incomplete reporting in many of the studies. Only one-third of studies provided adequate information about laboratory methods for diagnosis and drug susceptibility testing. Information about treatment regimens was particularly poorly reported, with missing information about drugs used initially and in the continuation phase, plus the duration of these two important phases.
Despite these important limitations, the findings of this review have several interesting implications. We speculate that the nearly complete reporting of the definition of treatment outcomes reflects the effort made over a decade ago by an international collaboration to develop a consensus definition for the major treatment outcomes in MDR-TB patients [15]. This was in striking contrast to the incomplete reporting of all other information (where no international consensus effort has been attempted).
The finding of inferior results with standardised regimens compared to individualised regimens was also found in two previous reviews [5, 7], but must be interpreted with caution due to potential confounding between centres, in terms of patient populations and resources available for treatment. It is plausible that the differences seen reflect these confounding differences more than any true benefit of individualised regimens.
The poor outcomes in XDR-TB patients is consistent with the findings of previous meta-analyses [6, 94, 95]. This finding demonstrates that 10 years after the first report of XDR-TB in South Africa [96], there has been no major advance in XDR-TB treatment.
The frequency of severe adverse events ranged from less than 1% to more than 10% with each drug. These findings are similar to those of other studies in individual cohorts [19, 21, 30, 43, 59], but given that these estimates are based on large numbers of patients treated at many centres, with narrow confidence intervals, they should be generalisable to most settings. However, the reporting of drug-related adverse events was remarkably inconsistent, even though adverse events with second-line drugs is considered a major limitation of current MDR-TB therapy [16, 97]. Although most studies reported adverse events, the majority of these did not define the methods of diagnosis, grading or attribution of these adverse events. If the management of the toxicity of these second-line drugs is to improve, we suggest that an international collaboration should define a standardised approach to definition, diagnosis, grading of severity and reporting of adverse events during MDR-TB treatment, similar to past efforts to standardise outcome definitions [15].
Conclusions
Interest and published experience in MDR-TB treatment has grown rapidly in the past 7 years, but efforts to synthesise this new body of evidence are seriously hampered by the methodological limitations of most studies and incomplete reporting. We suggest that collaborative international efforts are needed to standardise the reporting of diagnostics, treatment and adverse events, as was accomplished over a decade ago for outcome definitions [15]. This effort, plus the use of stronger study designs, including individual patient data meta-analyses, registry trials or other forms of randomised trials, will help to identify safer and more effective treatment for MDR-TB.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-00803-2016_Supplementary_material
Supplementary tables ERJ-00803-2016_Supplementary_tables
Disclosures
Supplementary Material
Z. Lan ERJ-00803-2016_Lan
M.L. Bastos ERJ-00803-2016_Bastos
D. Menzies ERJ-00803-2016_Menzies
Footnotes
This article has supplementary material available from erj.ersjournals.com
Support statement: This work was funded by a grant from the World Health Organization, which received funding from the United States Agency for International Development (USAID). Funding information for this article has been deposited with the Crossref Funder Registry.
Conflict of interest: Disclosures can be found alongside this article at erj.ersjournals.com
- Received April 22, 2016.
- Accepted January 10, 2017.
- Copyright ©ERS 2017
References
