Bárbara Flor de Lima conceived the study, participated in the collection of data and wrote the paper. Margarida Tavares participated in the design of the study and reviewed the manuscript. Both authors read and approved the final manuscript.
Conflict of interest
The authors have stated explicitly that there are no conflicts of interest in connection with this article.
Extensively drug-resistant tuberculosis (XDR-TB) is emerging as a global public health problem. Its treatment is more expensive and difficult, and the outcomes much severe. The identification of risk factors for XDR-TB is of paramount importance to design effective TB control strategies.
To review published articles on risk factors for XDR-TB.
We identified 249 English articles on PubMed, and 182 were excluded by the abstract. The remaining articles were retrieved for full-text detailed evaluation by authors, and 27 relevant articles were selected for final review.
Some risk factors were consistently present, mainly previous TB treatment and its length. Other conditions often associated were immigration, alcoholism and HIV coinfection. Pre-XDR-TB points to an increased risk of XDR-TB.
The information regarding determinants of XDR-TB is relatively weak. However, special emphasis should be given to minimize the risks of TB retreatment to prevent the emergence of highly resistant TB.
The term extensively drug-resistant tuberculosis (XDR-TB) was first proposed by the US Centers for Disease Control and Prevention in March 2006 . The definition was revised in October 2006, being defined as TB with resistance to at least isoniazid and rifampicin, and resistance to a fluoroquinolone and a second-line injectable drug (SLID) (i.e. amikacin, kanamycin or capreomycin) .
By 2010, according to the World Health Organization (WHO), XDR-TB patients were reported from 46 countries, and a total of 58 countries have reported at least one case. The proportion of XDR-TB among multidrug resistant tuberculosis (MDR-TB) was 5.4% from combined data. Eight countries and settings reported XDR-TB patients in more than 10% MDR-TB isolates, being the majority of them located in Eastern Europe and Central Asia, with the highest proportions found in Japan (30.9%), Tajikistan, Dushanbe city and Rudaki district (21.0%). Forty-one countries notified less than 10 patients. Of the 27 high MDR-TB burden countries, only in Estonia and Latvia MDR-TB patients were routinely tested for second-line drug susceptibility, and 11 countries have not reported a case of XDR-TB. In certain countries with low burden of TB and MDR, such as Czech Republic, Ireland or Israel, higher proportions of XDR-TB patients are found, but meaning few absolute patients [3, 4].
The emergence of resistant tuberculosis is related with poor quality of tuberculosis programs. Directly observed therapy (DOT) was implemented to improve control over TB treatment. In 1999, WHO established the Working Group on DOT-Plus for MDR-TB to explore the effectiveness of treatment. By the end of 2004, 83% of the world's population is living in countries covered by DOT, with a TB treatment success of, on average, 82%. Treatment success was below in African region (72%) because of HIV coinfection, and in the European region (75%) partly because of drug resistance. Many developing countries have not adopted the principles of international standards of care with DOTs because of resource-limited conditions, thus contributing to drug-resistant TB .
At an individual level, TB resistance can be primary (transmission of resistant strains of Mycobacterium tuberculosis to a susceptible host) or acquired (in the context of treatment) .
XDR-TB is emerging worldwide as a major public health problem , and may be related to different regional epidemiological aspects and challenges to deliver appropriate care and treatment of TB. In a study conducted in KwaZulu Natal, South Africa, from 1428 presenting to the district hospital with signs and symptoms of TB, the probability of having MDR-TB was 13% and XDR-TB 2% .
XDR-TB is more expensive and difficult to treat than MDR-TB, and outcomes much worse with higher rates of treatment failure and death [6, 8]. These patients also presented longer duration until culture conversion, a median of 26 months or, accordingly to other series, a median of 183 days until conversion [8, 9]. Recent data showed that patients with XDR-TB were 13 times more likely to remain culture-positive at 6 months of treatment , reflecting the lack of effective drug schemes, and more frequently presenting a fatal outcome, with 23% of a series of patients dying during the first year of treatment . Accordingly to a South African series of HIV coinfected patients, 98% of XDR-TB patients died within a median of 16 days from prompt diagnosis . In one study conducted in South Korea, TB-related mortality was 48%, with a median survival time of 51 months . The most relevant characteristics associated with treatment failure of XDR-TB were male gender, HIV infection and resistance to all second-line drugs . Others report that MDR-TB and drug resistance to first-line drugs beside rifampicin and isoniazid are predictors of worst outcomes . Actually, many series have more encouraging results, with 60% of the patients completing treatments being cured .
Recently, Velayati et al. introduced the term super-XDR-TB, or totally drug-resistant tuberculosis, defined as MDR-TB resistant to all second-line drug classes tested (i.e. aminoglycosides, cyclic polypeptides, fluoroquinolones, thioamides, serine analogues and salicylic acid derivates) .
The present study review available data on XDR-TB to identify its determinants and highlight modifiable factors.
Data source/study selection
This study is a review of published English articles on the risk factors for XDR-TB, developed following standard guidelines . We searched PubMed database from January 2006 to June 2010. Search terms were ‘risk factors’, ‘XDR TB’ and ‘extensively drug-resistant tuberculosis’. References cited by review articles were additionally screened.
Two hundred forty-nine articles were identified, and 182 were excluded by title and abstract. The remaining 67 articles were retrieved for detailed evaluation, with 27 relevant articles being selected for final review (Fig. 1).
The criteria for inclusion were XDR-TB case definition as WHO 2006, explicit data about XDR-TB patients and XDR-TB patients not included in MDR-TB group.
Demographic (sex, age, migrant status) and clinical characteristics (previous TB diagnosis and treatment regimens, HIV status) of the enrolled patients were revised.
Meta-analyses were not possible because of inconsistent definitions and methodologies, with different parameters studied.
Country general TB epidemic characteristics are relevant for the occurrence of XDR-TB. Table 1 presents the epidemiological features of the countries of origin of the populations evaluated , and Table 2 describes the social and demographic characteristics of patients.
Table 1. Characteristics of the countries of the studies
The highest proportion of XDR-TB among MDR-TB patients was found in the study conducted in South Africa (58.4%) and in Portugal (52.3%) [16, 17], and the lowest was seen in United Kingdom (0.9%) .
In the majority of studies, the male gender was predominant. In Jeon et al.'s study, female gender was significantly associated with XDR-TB, but only in univariate analysis [odds ratio 3.2, 95% confidence interval (CI) 1.1–8.3] , similar to the lack of a gender effect observed in other studies [9, 19, 20]. In one study, super-XDR-TB patients (n = 15), the male to female ratio was more than threefold, with statistical relevance (P < 0.05) .
In United States, the 83 XDR-TB patients were significantly less likely to be in the older groups (PR, 0.45; 95% CI, 0.26–0.78 for age 45–64 years; PR, 0.28; 95% CI, 0.14–0.57 for age >65 years) compared with drug-susceptible ones , which is concordant with another study . The same was verified in the study conducted in Japan, with patients with MDR/XDR-TB significantly more likely to be younger (OR 5.69 95% CI 2.63–12.45 for age 21–40 years; OR 4.11 95% CI 1.93–8.85 for age 41–60 years) . Bonilla et al. found no age difference between XDR and MDR patients (P = 0.316) .
In the study conducted in Estonia, immigrants (26%) were insignificantly associated with XDR-TB . High proportion of immigrants were seen in the studies conducted in Germany and California, and may be associated with other clinical characteristics [12, 21].
In the study conducted in the United States, compared with drug-susceptible patients, XDR-TB patients were more likely to be Hispanic (PR 2.16, 95% CI 1.19–3.93) . In the study conducted in Japan, MDR/XDR-TB patients were more likely to be immigrants (OR 6.41 95% CI 2.69–14.72) . Another study states that non-immigrant status was predominant, without significant differences . Others reported that 83% of 18 patients of XDR-TB at the time of the presentation were foreign-born, although not statistically different from MDR-TB patients. These patients were mainly from Mexico (46.7%), South Korea (20%) and Philippines (13.3%) (P = 0.02). Six patients were reported within 6 months after patient's arrival in the new country, with an average of 10.8 months (0–13.5 years) to the TB report (P = 0.04). Consequently, they were more likely to be recent immigrants, with an average of 3.2 years to TB diagnosis, compared with 5.1 years in the MDR-TB (P = 0.007) . In addition, Velayati et al. reported 50% and 40% immigrants among XDR and super-XDR-TB patients, respectively. They were mainly from Afghanistan and have been frequently travelling in and out the country. They were much younger than Iranian TB patients (average 34.5 vs 61.2 years), with statistical relevance (P < 0.05) .
In one study conducted in Russia, 52% (n = 14) had previous history of imprisonment, with an association with XDR-TB (PR 1.86 95% CI 0.88–3.94), although without statistical relevance (P = 0.10) . In the study conducted in California, 5.6% (n = 18) had been imprisoned, not different from MDR-TB group, (P = 0.34), but statistically significant from drug-susceptible patients (P = 0.04) . In the study conducted in the United States, compared with drug-susceptible patients, XDR-TB ones were more likely to reside in a correctional institution (PR 2.35, 95% CI 1.08–5.10) .
In the study conducted in Japan, patients with XDR-TB were more likely to have diabetes mellitus (OR 2.0 95% CI 0.34–11.85) and a history of malignancy (OR 5.25 95% CI 0.52–61.86), although not statistically relevant . Lai et al. showed diabetes mellitus to be the most frequent underlying disease (60%), followed by chronic pulmonary disease (20%) and lung cancer (10%), along with end-stage renal disease (10%) . Data were consistent with another study conducted in Taiwan. However, this was not different from the clinical characteristics observed in MDR-TB patients . In another study, diabetes mellitus is still the condition that is more prevalent (18.7%) .
The 53% prevalence of HIV among 69 XDR-TB patients, although similar to the one found among MDR-TB patients, was more than threefold higher than in drug-susceptible TB patients (PR 3.76, 95% CI 2.25–6.30). HIV had a predictive value for the diagnosis of XDR-TB, being 2.5-fold higher . HIV coinfection was found in 65.3% of XDR-TB Portuguese patients compared with 42.9% of non-XDR-TB ones (P = 0.01), and HIV coinfection had predictive value for the diagnosis of XDR-TB (OR 2.5, 95% CI 1.24–5.05) .
In South Africa, all patients tested for HIV were positive [7, 26], with 34% on antiretroviral therapy and a median CD4 count of 63/mm3 . The same was reported by Chakraborty et al. (n = 4) . In another study conducted in Tugela Ferry (n = 139), the HIV was screened in 84% of patients, being 98% coinfected. The median CD4 cell count at TB diagnosis was 66 cells/mm3, with 22% of patients on antiretroviral therapy .
However, in many studies, no HIV-positive patients were found [8, 9, 11, 13, 14, 28, 29]. In one study, the population was specifically HIV-negative . Not every study accessed HIV status, and screening was not done to all participants [18, 24, 31].
Prior TB diagnosis and treatment
The highest proportion of prior TB diagnosis was reported in Germany (86%) and Taiwan, with 90% of patients with a history of TB, and around 50% receiving fluoroquinolones for more than 1 month before diagnosis of XDR-TB [11, 23]. Another study conducted in Taiwan (n = 10) reports that patients with XDR-TB were more likely to have previous history of TB (P = 0.017) .
Shah et al.'s study suggested that prior TB diagnosis was significantly associated with XDR-TB (PR 2.78, 95% CI 1.43–5.38, compared with drug-susceptible patients) and showed a 1.2-fold increase in XDR-TB . In the Portuguese study, the characteristics with predictive value for the diagnosis of XDR-TB was increased duration of previous treatments (OR 1.2, 95% CI 1.11–2.30) .
In the Indian study on XDR-TB patients (n = 5), two patients had a definitive history of previous treatment with second-line drugs; the remaining three patients had multiple previous courses of antituberculosis therapy, without detailed records . In addition, in another study from Eastern India, three out of four patients had a history of interrupted TB treatment . In a study conducted in Iran, seven out of eight XDR-TB patients had previous TB treatment, and all the super-XDR-TB patients (n = 15) had previous TB . In the study conducted in the United Kingdom, 50% of XDR-TB patients (n = 6) had previous TB, while previous TB history was unknown in two other patients .
In the study of Murase et al., the MDR/XDR-TB patients (n = 55) were more likely to have had previous TB treatment (OR 10.55 95% CI 5.93–18.83) .The study of seven XDR-TB patients showed that 86% had a previous TB treatment for more than 1 month compared with 53% of 177 MDR-TB patients . Also, Vilariça et al.'s study showed 4.2 months of previous treatment in XDR-TB patients (P = 0.03) compared with 3.8 months in the non-XDR-TB group. Among XDR-TB patients, 26.1% were naive, and the remainder had retreatment because of treatment failure, interrupted treatment or relapses (P = 0.003) .
In another study, XDR-TB patients had a greater number of previous TB treatments than those with MDR-TB (2.0 and 1.6, respectively, P = 0.0026) and underwent two or more previous treatments. The proportion of XDR-TB patients never treated with SLID (40.5%) were lower than MDR-TB patients (63.3%) (P < 0.0001) . Accordingly to Mitnick et al., XDR-TB patients had undergone more treatment regimens than MDR-TB patients (4.2 ± 1.9 vs 3.2 ± 1.6; P < 0.001). The former patients had isolates resistant to more than eight agents tested against five from MDR-TB group, P < 0.001, and received an average of 5.3 ± 1.3 anti-TB agents . In opposition, in the study conducted in Russia, although patients with XDR-TB patients were more likely to have history of previous TB treatment (PR 4.85 95% CI 0.64–36.92), it was not statistically significant (P = 0.14) . In another study conducted in Siberia among patients who were receiving MDR-TB treatment (n = 536), 6.3% developed XDR-TB. In the multivariable analysis, the most relevant factors for the development of XDR-TB were prior TB treatment with SLID (adjusted HR, 3.65; 95% CI, 1.81–7.37) and cumulative non-adherent months (adjusted HR, 1.17; 95% CI, 1.01–1.35) . In opposition, in the study conducted in Russia, although patients with XDR-TB patients were more likely to have history of previous TB treatment (PR 4.85 95% CI 0.64–36.92), it was not statistically significant (P = 0.14) . In the study of Jeon et al., the number of SLID taken in the past were significantly associated with XDR-TB (OR 1.3, 95% CI 1.1–1.5). In the multivariate analysis, those patients with a cumulative treatment of 18–34 months were 5.8-fold (95% CI 1.0–61) more likely to have XDR-TB compared with those treated for ≤8 months, suggesting an increasing prevalence of drug resistance associated with a long treatment history (P = 0.012) .
In one study of 18 XDR-TB patients, 60.9% had received previous treatment under DOT strategy (P = 0.005) . Eker et al. and Murase et al. also reported one case with previous TB treatment under DOT [11, 20], and Shah et al. also reported 28 and 30 patients under complete or partial DOT, respectively . In other case report from Burkina Faso, patient one had DOT during the first 2 months of standard therapy, and patient two had previous diagnosis of MDR-TB and had 13 months of DOT therapy of SLID .
In South Africa, most XDR patients had no previous TB diagnosis, but it may reflect lack of reporting or high chances of nosocomial transmission . In another study conducted in Tugela Ferry, 69% of patients had previous TB treatment, with 56% on the past year. More than half of the patients (57%) had been hospitalized in the previous 2 years .
Many potential factors might contribute to treatment failure, the most relevant being male gender and HIV coinfection . Reported factors for treatment non-compliance among XDR-TB patients were alcoholism (10.1%), drug addiction (42%) or both (47.8%) .
Pre-XDR-TB and super-XDR-TB
The term pre-XDR-TB has been introduced recently, and is defined as resistance to rifampicin, isoniazid and either to a fluoroquinolone or a SLID . A higher proportion of such patients were found in South Africa [26, 35]. Accordingly to Kim, 11.3% (n = 159) had ofloxacin-resistant pre-XDR-TB, and 8.3% (n = 117) had SLID-resistant pre-XDR-TB. A previous history of treatment with SLID was more common in patients with XDR-TB (35.6%) and ofloxacin-resistant pre-XDR-TB (32%), P < 0.05. At least two previous TB treatments were more common in patients with XDR-TB (47.7%), ofloxacin-resistant pre-XDR-TB (38.4%) and SLID-resistant pre-XDR-TB (27%) than in other forms of MDR-TB (24.9%), P < 0.05 .
Recently, the term super-XDR-TB was introduced . Some authors already described strains of XDR-TB resistant to all second-line drugs tested [11, 19, 27, 32]
This review is quite representative of the worldwide distribution of XDR-TB, with studies included from varied countries, although some of them are enrolled in specific settings and with small samples.
One study conducted in South Africa had much higher proportion of XDR-TB cases (58.4%) comparing with others conducted in the same country (24% or 6%), which may represent the local epidemic specificities of Tugela Ferry [7, 16, 25]. The same succeeds in the study. This may be explained by the fact that first represent a epidemic outbreak in this region and the latest patients were selected from a referral hospital .
Identifying the risk factors for the development of XDR-TB are of paramount importance to understand XDR-TB, and to the effectiveness of TB control strategies, and may have prognostic relevance; as stated by Kim et al., previous TB treatment with SLID or more previous TB treatment could be associated with poorer long-term survival .
The majority of these patients were male and were younger, quite transversal to all TB patients, and not different from the MDR-TB ones.
In European prisons, there is 17 times higher frequency of TB compared with the general population, being higher in Western Europe. Risk factors for TB outbreaks in prisons are late diagnosis, inappropriate treatment, overcrowding, HIV coinfection, illegal immigrants and drug addiction. Homeless, unemployment, alcoholism, drug addiction and difficulty in complying with TB therapy were stated as major barriers to completion of TB treatment among prisoners, therefore potentially contributing to the emergence of XDR-TB . However, to the best of our knowledge, there are no data on the prevalence of XDR-TB in prisons, although in specific settings this may contribute to a higher risk of developing XDR-TB .
Immigrants may face particular barriers to timely access health-care services; therefore, they may have higher chances of developing TB. The risk of TB decreases since the entry in the new country but remains higher than general population . In many studies, a great proportion of XDR-TB patients were immigrants, and in such cases it imposes a reactivation of latent TB, although the true risk for developing XDR has not been established.
In many studies, diabetes mellitus was an important condition associated with TB. Although some authors say that this could result in higher susceptibility to the infection of resistant strains, the true mechanisms are not known .
Other conditions, such as homelessness, drug and alcohol abuse, and unemployment, may be important, not directly, but involved through HIV coinfection or contributing to the failure of previous TB treatment. However, such conditions are not accessed in the majority of the studies.
Africa is facing the worst TB epidemics [3, 7], with HIV being attributed as the major risk factor, by reactivating latent TB by 100-fold or because of immunosuppression-promoting TB infection. It would be expected higher proportions of HIV coinfection in these studies, attending to HIV prevalence in those countries. However, in some studies all patients were HIV positive [10, 26]. In one study, patients with HIV coinfection had worse 1-year mortality (83%) and 30-day mortality (956 deaths per 100 patient-years) rates, with higher degree of drug resistance (P < 0.001) . HIV had a predictive value for the diagnosis of XDR-TB, being 2.5-fold higher in one study . However, HIV screening is not always done, thus not being able to assess the true risk, and overall limiting the management of such patients.
The proportion of resistance among second-line drugs has been increasing over the years, and is higher in previous TB patients. Some of the authors concordantly agreed that the higher the previous use of drugs, specially SLID, the higher the resistance . However, this can be explained not only by increased resistance but also by higher rates of screening of susceptibility to second-line drugs . Few diagnostic tests have had their performance evaluated with second-line drugs, just making possible the diagnosis of XDR-TB at the level of national/regional reference laboratories, limiting the prompt diagnosis and management in underdeveloped countries .
One problem that is emerging is resistance to fluoroquinolones, widely used for other conditions. In a study conducted in South Korea, ofloxacin resistance increased 5.1-fold among persons who reported purchases of antibiotics unrelated to TB treatment. In addition, Huang found, after analyzing 141 isolates of Mycobacterium tuberculosis, that minimum inhibitory concentrations of fluoroquinolones were increased only in MDR-TB isolates, and from 10 fluoroquinolones-resistant MDR isolates, 50% possessed mutations other than S95T in the gyrA gene. No gyrB mutation was found in any isolates of MDR-TB. Thus, fluoroquinolone resistance is the result of previous MDR treatment rather than its use in the comunity. This reinforces the need for routine fluoroquinolone susceptibility testing .
Pre-XDR-TB is an interesting condition that needs to be further assessed but may point to an increased risk of developing XDR-TB. Even these cases may have poorer outcomes (P < 0.05). Streptomycin-resistant pre-XDR-TB was associated with long-term mortality (HR 2.69, 95% CI 1.40–5.16, P = 0.003) .
The term super-XDR-TB has recently been introduced and may imply even worst outcomes. It could be considered a subset of XDR-TB. This needs also to be further assessed.
To sum up, there is limited information regarding the determinants of XDR-TB, concordant with a systematic review . This fact limited the scope of our inferences because the analysis is lacking from nationwide surveys. The described quantitative risk for these factors is probably underestimated because of several limitations of individual studies, namely concerns with prompt diagnosis, laboratory quality control for testing drugs resistance, data collection on patients characteristics and absence of standard procedures.
A major finding of this review is the special emphasis that should be given to minimize the risks of TB retreatment on the emergence of highly resistant TB.
Actually, emerging resistance imposes serious limitations in terms of efficient TB treatment. Knowledge of its risk factors is of major importance in order to prevent its occurrence.
In a 5-year view, it is expected that the outgrowth of primary resistance and XDR-TB will no longer be a problem of those non-adherent to therapy, with important consequences for public health.