Epidemiology and clinical management of XDR-TB: a systematic review by TBNET

Study design

The current study is a systematic review of published English-language literature on the clinical epidemiology, diagnosis, treatment approaches and outcomes of XDR-TB, according to the Cochrane guidelines 26.

Selection criteria

Two independent researchers used the same keywords and methodology to generate a list of original research articles published in peer-reviewed journals in clinical epidemiology and clinical management of XDR-TB. Using Medline and Embase databases, they searched for terms describing M. tuberculosis drug resistance (keywords: “tuberculosis”, “extensively drug-resistant tuberculosis”, “XDR-TB”, “multidrug-resistant tuberculosis” and “MDR-TB”). Potentially relevant articles were retrieved and the reference lists were reviewed, to identify studies that the search strategy may have missed.

Final inclusion criteria

Abstracts of journal articles were reviewed to retrieve only those containing: 1) a definition of XDR-TB consistent with the revised, current definition of XDR-TB (as stated in the document published by the Centers for Disease Control and Prevention in 2006 8 and in other relevant articles 7, 27); 2) a cohort of >100 MDR-TB cases from which data were gathered on XDR-TB cases; 3) evaluation of treatment regimens or of treatment outcomes (as proposed by Laserson et al. 28 or, alternatively, by the WHO 1); and 4) a publication date between March 1, 2006 and December 15, 2008. Each study included was systematically reviewed in terms of methodology, assumptions, results and conclusions, setting the current review apart from other recent overviews of XDR-TB.

Formal meta-analytic techniques could not be applied for a comparative analysis because the studies identified used different measures (i.e. mean or median, standard deviation or interquartile range) and some did not report parameter estimates.

Evaluation criteria

The following topics were reviewed and discussed by at least three independent researchers, to be then approved by the whole research panel.

Clinical and epidemiological features of cases enrolled

The clinical and epidemiological features of cases enrolled are shown in tables 1⇓ and 2⇓. All 13 studies included in the review used a definition of XDR-TB consistent with the current WHO definition (in vitro drug resistance to isoniazid and rifampicin plus to any fluoroquinolone and at least one of the injectable drugs (capreomycin, kanamycin or amikacin)) 7, 8. Out of the 13 studies, eight (62%) performed DST on all three injectable drugs included in the XDR-TB definition 7, 8 and five on only one of the three injectables. In those five studies, therefore, additional cases of XDR-TB may not have been identified because of the lack of comprehensive DST.

All 13 studies reported the number of MDR- and XDR-TB cases. Overall cohort size ranged from 1,989 MDR-TB cases (including 119 XDR-TB cases) in Peru 23 to 137 MDR-TB cases (including 11 XDR-TB cases) in several European countries 10. The proportion of XDR-TB cases among MDR-TB cohorts ranged from 24% in South Africa 9 to 4% in Germany 18, but a number of them were not representative of the overall MDR-TB population in the country.

The proportion of retreatment cases was reported in 10 (77%) out of 13 studies. The proportion of retreatment cases among MDR-TB cases ranged 49–98% (being higher in Tomsk (Russia), Peru and Korea 16, 19, 20, 23, 24 than in Germany, Italy, Estonia and Russia 10, 11). The proportion of retreatment cases among XDR-TB cases ranged 45–100%, being consistently higher in XDR- than in MDR-TB cases.

Only six (46%) studies reported the mean or median number of previous treatment regimens longer than 1 month. It was consistently higher among XDR-TB cases (range 2–4.2) than among MDR-TB cases (1.6–3.2).

Only five (39%) out of 13 studies reported the number of drugs to which strains of M. tuberculosis were resistant. Three studies reported the median number and two the mean number of drugs to which strains demonstrated in vitro resistance. In all of these studies, the number of drugs to which a strain of M. tuberculosis was resistant was higher in XDR-TB cases (range 7–9) than in MDR-TB cases (4–5.3). The sample was considered representative in eight (62%) studies, with a population coverage ranging 65–100%.

The report detailing the Tugela Ferry outbreak of XDR-TB in South Africa was the only study concerning HIV-infected patients 9, while the remaining reported HIV-uninfected subjects almost exclusively. HIV testing practices were not described in the majority of studies. The prevalence of HIV in the studied cohorts was <5% in the all studies except the one from South Africa (in which it was 100%) 9.

Laboratory diagnosis

Microbiological features of the studies reviewed are shown in table 3⇓. Out of the 13 studies included in the review, eight (62%) reported that DST was performed for all second-line drugs necessary to diagnose XDR-TB. In six (46%) of the studies, the diagnosis of MDR/XDR-TB was based on DST performed under SRL quality control or in SRLs themselves. No data were available on the methods used for cultures (solid versus liquid media) or on DST to assess the resistance of the strains reported in the studies. Altogether, 148 (35%) out of 428 XDR-TB patients and 1,720 (36%) out of 4,774 MDR-TB patients received a diagnosis based on quality-controlled DST.

Five (38%) studies reported the number of first- and the second-line drugs tested and their results: strains were resistant to a mean or median of five and eight drugs in patients diagnosed as MDR- and XDR-TB, respectively.

Characteristics of treatment

The characteristics of treatment are shown in table 4⇓. Although the majority of the studies reported the use of individualised treatment regimens 10–12, 16, 18–20, 22–24, two cohorts were prescribed standardised treatment regimens 9, 23. First-line drugs were reported to be available in all studies, and the vast majority of cohorts had access to all second-line drugs 9–12, 16–24. Linezolid was reported to be available in only three studies 12, 18, 21; therefore, data were insufficient to comment further on its efficacy and safety.

The rate and nature of adverse events during XDR-TB treatment were mentioned in some studies, none of them suggesting that adverse events are an important limiting factor for the treatment of XDR-TB compared with MDR-TB cases 12, 18, 19.

Treatment efficacy end-points

The treatment efficacy end-points are shown in table 5⇓. Out of the 13 studies included in the review, only three (23%) evaluated both median time to sputum smear and culture conversion, while four measured only the median time to culture conversion.

Longer treatment duration and delayed sputum smear conversion were reported in most studies among XDR-TB patients. Median time to sputum smear conversion ranged 41–56 days in MDR-TB patients, while it ranged 88–110 days in XDR-TB cases. The median time to culture conversion ranged 58–99 days in MDR-TB cases; it was substantially different in XDR-TB patients, ranging 60–195 days.

There were 11 studies reporting on treatment outcome in a series of XDR-TB cases 9–12, 16, 18–20, 22–24. Patients were followed up for a minimum of 18 months in the studies in which it was reported. Out of the 13 studies, four (31%) used the WHO definition of cure 9, 10, 22, 24, six (46%) used the Laserson et al. 28 definition of cure 12, 16, 18–20, 23 and two did not specify the definition used 17, 21. Although one study used the WHO definition 11, its cohort data were re-evaluated using the Laserson et al. 28 definition in order to improve comparability 13.

Analysis of treatment outcomes among HIV-infected individuals in the South Africa report revealed that 98% (52 out of 53) died with a median survival time of only 16 days 9. However, it is doubtful whether this finding can be generalised to all HIV-seropositive individuals with XDR-TB, and the outcome was certainly better in studies of predominantly HIV-seronegative individuals, where the proportion of treatment success exceeded 48% in all cohorts. Seven studies reported on the use of surgery, which varied from 4 to 56% of XDR-TB patients 12, 16–20, 24. In two studies, higher success rates were associated with the use of surgery among XDR-TB patients 16, 20.

The studies reviewed provide evidence against the statement that XDR-TB cases are virtually untreatable. In nine studies that provided figures on treatment success, the overall success rate (cured and completed) ranged from 34% 11 to 67% 20. The success rate of XDR-TB cases was compared with that of MDR-TB cases in 10 studies: no difference was found in five 11, 12, 16, 18, 20, while in the remaining five the success rate was significantly reduced among XDR-TB cases, by ∼1.5 times 23 up to 20 times 17.

The proportion of cases failing their treatment was only assessed in seven (54%) studies; it ranged 10.4–31% in the XDR-TB cases and 0.6–21% in MDR-TB patients. The majority of the nine studies reporting on death also demonstrated a high mortality rate among XDR-TB cases, ranging from 7% 19 to 36% 10, with the rate being >20% in five studies 10, 16, 22–24. Mortality was compared with that of MDR-TB cases in nine studies: three found similar rates 16, 19, 23 and the remaining six found increased mortality among XDR-TB cases: from a two-fold 12, 18, 22 up to a five-fold increase 10. The relative risk for death due to XDR-TB (compared with MDR-TB) was found in the study by Migliori et al. 11 to be 5.5, and was computed based on the data of Kim et al. 12, who found this relative risk to be 1.8.

One study reported that XDR-TB cases that were initially treated on the basis of the DST results (DST results available within 31 days of treatment initiation) had a significantly better outcome compared with XDR-TB cases for whom the DST results were available after >31 days of treatment 23. There were 14 XDR-TB patients in the first stratum (≤31 days) and 23 in the second (>31 days) 23. However, the reasons for this difference were not explained.

Multivariate analysis of factors (demographic, clinical and epidemiological) associated with treatment outcome, which included the XDR-TB status, was performed in three studies 16, 20, 24. Only one study 24 found a significant association between XDR-TB and poor treatment outcomes (treatment failure or death).

References

The reference section of only two studies reported all the previously published studies that were included in the present review 10, 11.

Summary

A summary of the variables analysed in the studies reviewed is shown in table 6⇓.

Methodological issues in the study of XDR-TB

A meta-analysis of the available data was not possible, owing to discrepancies between the studies with respect to definitions used and data elements collected. Since the study of XDR-TB as a phenomenon is relatively new, all the available studies have been affected by one or more of the following pitfalls: retrospective and observational design that does not permit collection of all useful variables; small sample size of XDR-TB cases (range 7–119); lack of quality-assured DST results; variable treatment regimens for XDR-TB (individualised or standardised) that prevent the possibility of drawing any conclusions on the effectiveness of single drugs or drug combinations 9–12, 16, 18–20, 22–24; and differences in outcome definitions that hamper the ability to compare results across studies (tables 1⇑–⇑3⇑).

Surprisingly, the proportion of retreatment cases in the cohort was not described in some studies and the number of previous treatment regimens longer than 1 month was not described in half of them. The same is true for the number of drugs to which isolates were resistant, which has been shown in some studies to correlate with treatment outcome 10, 11, 16, 18, 20. All of this information is necessary to describe the epidemiological context and to correctly interpret the data provided (table 4⇑).

Diagnosis of XDR-TB is one of the most problematic areas of its study. WHO recommendations for second-line DST call for testing one fluoroquinolone and all three injectable agents. However, it is not possible to do so at present in the vast majority of settings with suspected high levels of M. tuberculosis drug resistance, because laboratory capacity is lacking and in some cases second-line drugs are not registered for use in the country so DST is not performed. Not all drugs that are included in the current XDR-TB definition were tested in all reviewed studies, as noted. The main consequence is that in several studies the true XDR-TB prevalence may have been underestimated.

DST quality assurance is essential for the diagnosis of MDR- and XDR-TB and for surveillance at a national and international level: the WHO and the International Union Against Tuberculosis and Lung Disease are supporting an international network of 26 SRLs that support more than 120 national laboratories 3. Standardised methods to test second-line drugs were only recently made available, but it is known that current in vitro susceptiblity testing may not be reliable for selected second-line drugs (e.g. cycloserine). Again, this difficulty is hampering the implementation of a fully operational surveillance system for second-line drugs at national level (tables 1⇑–⇑3⇑).

The current authors assessed all the studies selected for the systematic review, recording how many of them reported DST data from quality-assured laboratories 29. At the time the reviewed studies were completed, second-line drug panel testing had only just been started among the SRLs, and not among national laboratories. So even for laboratories undergoing regular external quality assurance, this may not guarantee the accuracy of the second-line drug susceptibility results reported. Less than half of the studies included in the current review reported DST data obtained from quality-assured laboratories; in other words, only about one patient out of three received a diagnosis of MDR- or XDR-TB (and, most likely, a tailored treatment) based on a DST of proven quality. The current review highlights the many questions still open about the diagnosis and surveillance of drug resistance (table 3⇑).

Most of the studies reviewed reported that patients received individualised treatment based on DST results. However, data were not always available on the number of active drugs used in these regimens. While current expert opinion supports the use of three to five active agents in treating XDR-TB, at least one study reviewed showed successful outcomes with only two to three active agents 20. Again, more information on methodology would be needed to judge whether these results are reproducible (table 3⇑).

The relevant clinical, epidemiological and microbiological differences among the cohorts reviewed are also likely to lower the strength of any conclusions drawn from the treatment efficacy end-point results. Complete evaluation of treatment outcomes as proposed by Laserson et al. 28 or, alternatively, by the WHO 1 was not described in all studies, highlighting the risk of potential biases in a meta-analysis of the data. Although the WHO definition of cure (based on a negative bacteriological examination in the last month of treatment and on at least one previous occasion) is widely used, the definition of Laserson et al. 28 should be considered the most appropriate to be used in future studies. The latter is, in fact, based on several consistently negative cultures (five) for the final 12 months of treatment and achieved consensus from the main partners in defining cure among MDR-TB patients 28. As previously stated, the combination of the results of several studies addressing a set of related research hypotheses was not possible (table 2⇑).

When the prevalence of a condition (e.g. MDR- or XDR-TB status) needs to be determined, a sample is considered representative when all cases have the same probability of being included in the sample, and randomisation is the technique allowing minimisation of possible biases. As the main focus of the current review was to discuss the clinical and epidemiological evidence of XDR-TB, the studies enrolling two-thirds (>65%) of the cases notified in the country were considered representative. This cut-off is valid when discussing clinical aspects (e.g. treatment outcomes, adverse events, etc.), as it excludes extreme selection bias conditions, but it cannot replace the recommended 100% coverage when the prevalence of the condition needs to be determined (table 1⇑) 2–4, 6.

All of these methodological challenges must be considered when evaluating the results of the studies and formulating new questions for further research.

Treatment issues in XDR-TB

Treatment of XDR-TB cases relies on drugs that are less potent and much more toxic than those frequently used in the clinical management of disease caused by drug-susceptible or MDR strains. Most reported cases with TB due to XDR strains in the Americas, Europe and Asia have occurred in cases with previous treatment against TB, indicating that former treatment mismanagement may have substantially contributed to the development of drug resistance. Thus, there exists an opportunity for preventing further drug resistance through improved clinical practices.

The role of single drugs or drug combinations in effective XDR-TB treatment has not yet been ascertained. Information on the number and type of effective drugs used was usually not given in the studies reviewed. Most data point to the fact that at least three to four drugs to which the strain is sensitive need to be used for effective treatment, although, in one study from Korea, a high success rate was obtained in spite of a low number of effective drugs (median of two) used in XDR-TB regimens 20. The XDR-TB regimens used in Peru relied heavily on drugs such as capreomycin, para-aminosalycilic acid and cycloserine, which had not been previously used in the country 16, 23. This fact may partially account for the positive results of treatment, in contrast to countries where widespread use of most second-line agents is more common (tables 2⇑ and 3⇑).

In many settings, fluoroquinolones were used in XDR-TB regimens under the assumption that strains resistant to ciprofloxacin may still be sensitive to levofloxacin or moxifloxacin. This assumption has not been proven 31, but the issue of cross-resistance deserves further study to determine the possibility of additional treatment options for patients with early generation fluoroquinolone resistance. Similarly, some studies included isoniazid (at standard or high dose) and rifampicin in XDR-TB treatment regimens because an adequate number of effective drugs were not available 12, 32, but the effect of this strategy was not measured. Finally, most studies included some of the drugs regarded as third-line agents (amoxicillin-clavulanate, clarithromycin, clofazimine, imipenem and linezolid), but their respective contribution could not be ascertained 33.

Surgical resection is a potentially useful strategy for a condition that cannot be adequately controlled with medical interventions. In studies from tertiary hospitals in Korea, 48% and 56% of all XDR-TB patients received surgical interventions 12, 20 and, in one study, surgery contributed significantly to treatment success compared with cases that did not receive surgery. Even in community MDR-TB programmes like the one in Peru, surgery was used in 14% of the XDR-TB cases, with a trend towards a better outcome 16. Surgical resection was performed in 4% of MDR- and XDR-TB cases in Korea 24. Surgical treatment improved treatment success in MDR-TB patients (68.3% in surgically treated cases versus 44.2% in nonsurgically treated cases; p<0.001). The study was not sufficiently powered to detect a statistically significant difference in patients with XDR-TB (67% in surgically treated cases versus 28% in nonsurgically treated cases with XDR-TB; p = 0.2). In the multivariate logistic regression analysis, surgical treatment was the only positive predictor of treatment success among MDR-TB cases (p = 0.001), as measured in the overall cohort of MDR-TB cases (not specifically related to XDR-TB cases; table 5⇑).

A crucial question is whether optimal trials to compare efficacy of treatment regimens for XDR-TB can be designed and implemented. Regimens will always need to be individualised and, thus, there will be a large diversity of regimens within a given trial and between trials. Randomisation will not be possible, preventing minimisation of several relevant biases. At a minimum, studies will need to report on the number of effective drugs used on a single patient and stratify by categories of drugs used (first-, second- and third-line). In addition, documenting changes in regimens over time for a single patient can improve interpretation of future study results (table 5⇑).

Outcome issues in XDR-TB

In contrast to the results presented in the initial report from HIV-seropositive individuals with XDR-TB in South Africa 9, the results of the other studies reviewed suggest emphatically that XDR-TB is not a death sentence for all. With an individually tailored treatment regimen (including pulmonary surgery in selected cases), treatment success can still be achieved in one-half to two-thirds of patients with XDR-TB, as was shown in the majority of studies from the Americas, Europe and Asia.

However, the first reports on XDR-TB confirmed that these patients have, in general, a poorer prognosis and fewer treatment options than MDR-TB cases 10–12. The majority of the studies analysed in the current review confirmed that XDR-TB has a higher probability of death, failure, longer hospitalisation, longer treatment duration and delayed microbiological (sputum smear and culture) conversion compared with MDR-TB. The fact that mortality was almost invariably increased among XDR-TB cases may suggest that a fraction of truly incurable patients does exist, as demonstrated by the first report on this topic (table 4⇑) 10.

Only a few of the available studies (e.g. Mitnick et al. 16 in Peru) reported similar success rates in XDR- and MDR-TB patients. These were obtained despite the fact that XDR-TB patients were on average resistant to a larger number of drugs. A key element, however, is how many effective drugs could be used: this varied significantly within and between studies, although precise figures were not reported and this effect was not analysed (table 4⇑).

Experience from Peru and other countries suggests that strategies to support adherence, as well as psychological, nutritional and even financial interventions, may further contribute to improved outcomes in XDR-TB patients 16, 23, 24. The promising results obtained from Korea and Peru also suggest that strict and well-defined rules (e.g. management of XDR-TB cases in specialised reference centres) are essential to improve prognosis. High success rates can also be achieved without hospital care in community MDR-TB programmes 16–24. Indeed, hospital care may be associated with nosocomial transmission of XDR-TB if adequate infection control strategies are not implemented. The lingering question for further study is whether these encouraging results, achieved with significant effort, can be reproduced in the conditions faced by most TB control programmes worldwide, and what interventions would be needed to achieve high success rates.

With the exception of the South Africa study 9, the studies reviewed reported on cohorts with very low rates of HIV infection and good access to adequate care. Additional data for HIV-infected individuals is needed to determine the role of co-infection in XDR-TB treatment outcome and to evaluate interventions that may contribute to improved outcomes in HIV-infected XDR-TB patients. The high case-fatality rate in the Tugela Ferry outbreak 9 probably represents a combination of factors at play: not only host factors but also lack of access to adequate diagnosis and treatment (table 2⇑).

In short, there are a number of potential factors which may influence treatment outcomes among XDR-TB cases, but the currently available studies were not designed to control for many relevant variables. Further information from other MDR- and XDR-TB hot spots (e.g. former Soviet Union countries and the Philippines) is needed to confirm the findings reported in the present review through larger and better-designed prospective studies. An evidence-based approach to this delicate issue should replace the sensationalism that has surrounded XDR-TB in the first 2 yrs following its first description 34.

Conclusion

According to the criteria used in the current systematic review, 13 studies focusing on clinical epidemiology, management and treatment outcomes of XDR-TB were identified. The results of the systematic review suggest a number of productive areas for further investigation, using improved methods to yield better data.

As with all new topics of investigation, XDR-TB has posed challenges to investigators in optimising their approaches to its study. The current literature cannot form the basis for a meta-analysis because of methodological differences that make comparisons impossible. To strengthen the available body of evidence for the diagnosis and treatment of XDR-TB, and ultimately its prevention, future studies should: 1) be prospective in design; 2) adopt standardised, internationally accepted definitions; 3) use quality-controlled laboratory testing for all first- and second-line drugs that define XDR-TB; and 4) collect data on an agreed-upon set of standard variables that will allow for comparison of approaches and results across studies. These variables include, for instance, measures of disease severity (number of drugs to which isolates are resistant and clinical features), treatment history, number and type of active drugs used in treatment, time to sputum smear and culture conversion, treatment duration, adverse events related to treatment, and treatment outcome. Using a systematic approach to the investigation of XDR-TB will yield more useful results more quickly.

A number of important challenges remain, chief among them the speedy and accurate diagnosis of drug resistance. Although rapid, sensitive and specific detection of isoniazid and rifampicin resistance is currently possible, further research is necessary in order to identify new tools able to diagnose resistance to second-line anti-TB drugs in a simple, economical and reproducible manner 35.

Perhaps most importantly, XDR-TB has been shown in the current literature to be a treatable as well as curable condition. While this is an encouraging finding, complacency is not appropriate. XDR-TB patients do more poorly in terms of smear and culture conversion, treatment duration, treatment failure and death. With clear indications that XDR-TB results from mismanaged cases of drug-susceptible and MDR-TB, the first imperative is to treat susceptible TB appropriately to completion and to provide rapid diagnosis and aggressive, appropriate treatment of MDR-TB to avoid the unnecessary development of additional cases of XDR-TB 5.

More investigation is needed in order to determine optimal treatment regimens using first-, second- and third-line agents (including drugs potentially useful to treat severe XDR-TB cases, such as linezolid), indications for surgery, and other supportive interventions that can increase the chances for treatment success while we wait for new drugs. In addition, more research must be done on XDR-TB and HIV co-infection to avoid a repetition of the devastating results for patients in the Tugela Ferry outbreak 9. All of this work will require access to large databases from surveillance systems or ad hoc designed studies (e.g. the European-funded FP7 project, TB PAN-NET) with standardised variables on diagnosis, treatment, and outcomes, as double-blind randomised placebo-controlled trials cannot be conducted for ethical reasons.

New diagnostic techniques and new drugs are urgent priorities for adequate control of multidrug-resistant and extensively drug-resistant tuberculosis. At the same time, they can only be effective if introduced into health systems that function adequately to provide access to care and treatment for all of the people who need it, which is another priority that we must continue to address with intensified action. As we move forward on all of these fronts to address the challenges of extensively drug-resistant tuberculosis, early diagnosis and aggressive treatment of extensively drug-resistant tuberculosis can yield positive results for patients and public health.