Effectiveness and safety of clofazimine in multidrug-resistant tuberculosis: a nationwide report from Brazil

Discussion

The aim of this study is to evaluate the effectiveness, safety and tolerability of clofazimine used within a standardised MDR-TB regimen in a large cohort of Brazilian TB patients. The study results suggest that both regimens are rather effective in achieving treatment success, as the rates achieved are higher than those reported in the largest MDR-TB cohort ever published [2, 3]. However, both regimens used in Brazil can be considered rather weak as they include three active, or likely to be active, second-line TB drugs. Furthermore, no major differences are detected when comparing adverse events in patients treated with a clofazimine-containing regimen versus a clofazimine-free regimen. Interestingly, the cases treated with clofazimine were less exposed to important TB-related risk factors, such as drug abuse and alcohol and tobacco use, while having a lower prevalence of diabetes mellitus. These differences are likely to be due to the economic and demographic changes which occurred over time in Brazil, although information bias (improvement of the surveillance system over time) cannot be excluded. From a clinical point of view, the prevalence of MDR-TB, pre-XDR-TB and XDR-TB cases was similar in both groups, although the patients treated with clofazimine had experienced more previous TB treatments than those treated with pyrazinamide. DOT was performed to a significantly higher degree amongst patients on a pyrazinamide regimen than amongst those receiving clofazimine, as a result of programmatic improvements over time in Brazil and the decision, which was taken in 2005, to make it a strategic component of the anti-TB strategy.

The outcome in the clofazimine-treated group (60.9% treatment success) was slightly less positive than that reported in a recent systematic review (65% treatment success in 599 cases treated with clofazimine) [18]. This systematic review included a single large study, the Bangladesh study [20], and several small-scale studies which enrolled between 5 and 46 patients [18]. Unfortunately, several details important for the comparison of our study with the systematic review are not available. One study quoted in the systematic review reports the duration of treatment with clofazimine as being 608 days (i.e. 20 months) as opposed to the 18 months in our study, while information on the daily dose of clofazimine is missing in five out of the eight studies in the review. However, 84% of the cases analysed in the systematic review were prescribed clofazimine at a dose of 100 mg·day^–1, the same as administered in our study. Treatment outcome definitions varied across studies in the review [18] and demographic data were also different. The proportion of females ranged between 20 and 85.7%, the median being 34.2% in our study, while the mean age ranged between 26 and 50 years (the median age in our study was 38 years). The HIV prevalence, when tested, was very low (one case per series only was HIV-positive, this being 6% in our cohort).

In a recent randomised open trial performed in China, 53 cases treated with clofazimine (administered for 21 months at a dose of 100 mg·day^–1) were compared with 52 controls. Treatment success was significantly higher in the treatment group when compared to the control group (73.6% versus 53.8%; p=0.035) and culture conversion was significantly more rapid, as was cavity closure (evaluated by chest radiography) [17]. Importantly, the background regimen used in the Chinese trial was stronger than in our study, including prothionamide, pyrazinamide and a fluoroquinolone (moxifloxacin, gatifloxacin or levofloxacin) (100% of cases), capreomycin or amikacin (79.2% of cases), ethambutol (39.6% of cases), and clarithromycin (49.1% of cases) based on DST results. All the available data suggests that clofazimine is active in MDR-TB cases. In our study, the slightly better results of the pyrazinamide-containing regimen are explained by the higher resistance profile initially observed in Brazil, particularly to fluoroquinolones (>30%) [26], by the higher number of previous treatments, and by the higher DOT coverage in this group. While a delayed diagnosis might have played a role we cannot study this based on the available data. However, patients treated with clofazimine had significantly fewer risk factors or co-morbidities (drug or alcohol abuse, a tobacco smoking habit, or diabetes mellitus).

The initial annual mortality rate decreased over time in the clofazimine group (data not shown) and we believe that while this cannot be attributed to cardiological complications (prolonging of the Q-T interval) it can be attributed to a higher prevalence of resistance (of which we have evidence) and delayed diagnosis (which, as stated above, we cannot investigate). Increased MDR-TB mortality rates are commonly observed in the early stages of implementation of large-scale MDR-TB programmes due to improved notification (deaths which, although previously occurring, were not notified) and programmatic factors such as delayed diagnosis, delayed treatment initiation and insufficient drug availability (leading to the use of drugs generally available at hospital level such as ciprofloxacin/ofloxacin and/or amikacin). Therefore, given that systematic DST for all drugs was not possible, some of the MDR-TB cases might have in fact been pre-XDR-TB or XDR-TB cases, further contributing to the increase in mortality. The use of a weaker fluoroquinolone (ofloxacin) in the clofazimine regimen compared to that used in the pyrazinamide regimen (levofloxacin) might also have contributed to increased mortality.

Recent discussions have focussed on the mechanism of action of clofazimine, which is not yet well known [28, 29]. In fact, clofazimine reaches high concentration in macrophages and is able to tackle the inhibitory effect of M. tuberculosis derived factors on phagocyte intracellular killing. The potential synergistic effect of clofazimine with γ-interferon, pyrazinamide and clarithromycin deserves to be studied further [18]. In addition, clofazimine has a long half-life, which can also contribute to the final effect of the drug although its effect is evident after the second week of treatment [28–33]. Furthermore, a recent study suggests that clofazimine has excellent sterilising activity in the murine model [29]. Recently, the possibility of developing a rapid test to detect susceptibility to clofazimine has been demonstrated, allowing the correct identification of eligible individuals in the near future [31]. Furthermore, recent evidence indicates that the use of clofazimine does not favour the development of resistance against bedaquiline [32, 33]. In fact, cross-resistance with bedaquiline has been attributed to the transcriptional regulator Rv0678 with concomitant upregulation of the multi-substrate efflux pump MmpI.5. As such, the mutation in Rv0678 initially considered responsible for the cross-resistance should be considered a confounding factor [32, 33].

In terms of safety, the global evidence available before our study consisted of the previously mentioned systematic review and a clinical trial [17, 18]. The systematic review [18] concluded that although the optimal clofazimine dose is not yet known (being 100 mg·day⁻¹ in the majority of available studies), adverse events are in general minor and rarely life-threatening. Although important, the review is affected by relevant between-study heterogeneity and by the epidemiological observational nature, with potential selection biases. Gastro-intestinal intolerance was observed in 40–50% of cases, with 75–100% reporting brownish skin pigmentation while ichthyosis and skin darkness were notified in 8–20% of patients. In the Chinese trial [17], 94.3% of cases reported adverse events of the skin and 47.2% reported ichthyosis, while 11.3% had gastro-intestinal disturbances and, overall, <4% had neurological complaints. In our study, adverse events were even lower: gastro-intestinal complaints were recorded in 10.5% of cases and hyper-pigmentation in 50.2% of patients. Neurological disturbances were reported in 9–13% of cases.

Our study has several limitations, being based on programmatic information in a setting where it is not possible to perform a complete DST for each patient, and where DST, despite being performed in quality-controlled laboratories belonging to the WHO network, cannot be performed with a complete panel in all cases. As a consequence, no additional information can be provided as to whether any of the MDR-TB cases were, in reality, pre-XDR-TB or XDR-TB cases. This shortcoming can hinder the interpretation of the results and higher proportions of difficult-to-treat cases (i.e., pre-XDR-TB or XDR-TB) could, in fact, affect the effectiveness indicators. In terms of effectiveness, the difference in the prevalence of risk factors discussed above is probably biasing the comparison in favour of the pyrazinamide-containing regimen, which was introduced in a mature phase of the Brazilian MDR-TB programme. Safety was evaluated using programmatic data which could be associated with a high rate of under-reporting. Therefore, a complete safety assessment (e.g. an ECG analysis) was not performed. The risk of underestimating adverse events (those occurring but not being recorded) is potentially high and should be kept under consideration when safety data from routine surveillance programmes are generalised to a population context. Furthermore, it was not possible to homogenise for the main confounding variables of the two treatment arms. If the population-based dataset is an important scientific resource (showing a real-world scenario), the role played by multiple confounding variables and their interactions could hinder the reliability and comparison of the findings with other settings. An epidemiological design, based on inclusion and exclusion criteria, can reduce the background noise related to the confounding covariates. The logistic regression model cannot significantly preserve the inferentiality of the findings when the sampling source is represented by a population whose statistical units are consecutively recruited. Moreover, important variables such as individual follow-up of the enrolled patients were not collected. However, the database is quality-controlled (being used to authorise the prescription of second-line TB drugs) and contains a large, nationwide comprehensive record (over 6000 cases) from Brazil, one of the high TB-burden countries.