Background
About a third of the world's population is infected with Mycobacterium tuberculosis. Tuberculosis remains one of the biggest killers among infectious diseases, with up to three million people dying from tuberculosis each year (Dye 1999). Diagnosis of tuberculosis generally relies on smear microscopy and culture of the sputum. The disease typically results in progressively destructive lung lesions but may affect almost any part of the body, usually with advanced wasting and death in more than half of cases in the absence of intervention. Despite the availability of increasingly effective treatment since the middle of the twentieth century the global burden of tuberculosis has continued to grow. This is partly because it is the commonest opportunistic infection in HIV-infected individuals and partly due to the practicalities of organizing complicated and prolonged treatment, with drug resistance often the price of failure (Dye 2000).
The discovery of effective antituberculous agents such as streptomycin, isoniazid, and para-aminosalicylic acid in the 1940s and early 1950s and their use in combination regimens to prevent drug resistance mutations arising in M. tuberculosis transformed the lives of tuberculosis sufferers. The introduction of the rifamycin class of antibiotics in the 1960s again revolutionized the treatment of tuberculosis and, as a component of a three- or four-drug regimen also including isoniazid and pyrazinamide, it was a crucial factor in reducing the duration of treatment from up to 18 to six months, raising rates of cure at six months to more than 90%, and reducing relapse to less than 5% (Fox 2001). Rifampicin, the first clinically useful rifamycin, has remained a central component of therapy. Its principal action appears to be in the later 'sterilizing' phase of treatment, and its postulated activity against semi-dormant organisms, which may form a significant component of the pool of persisting bacilli (Dickinson 1981), is probably the explanation for its efficacy (Mitchison 1992). The favourable pharmacokinetic and pharmacodynamic properties of rifamycins have also enabled treatment to be safely given as infrequently as twice a week rather than daily, helping to improve adherence to treatment in difficult situations. However, current regimens still require people to adhere to six months of treatment to reduce relapse rates to an acceptable level and new approaches to treatment clearly need to focus on improving the 'sterilizing' activity of the regimen (Gelband 2000).
New rifamycin derivatives with different properties have been synthesized, the first of which to reach clinical trials was rifabutin, a spiropiperidyl derivative of the parent compound rifamycin S (Marsili 1981). This drug was initially released on a compassionate basis in 1983 for the treatment of disseminated Mycobacterium avium intracellulare infection in patients with AIDS (O'Brien 1987). This initial experience suggested that it had good tolerability and safety, with the most prominent, and unusual, adverse effect being uveitis. At lower doses rifabutin also seemed to offer significant potential advantages over rifampicin for the treatment of M. tuberculosis (Kunin 1996). Several in vitro properties of rifabutin point to enhanced 'sterilizing' activity. Though the ratio of peak plasma concentration in humans to minimum inhibitory concentration (MIC) in the laboratory (Cmax/MIC) for rifabutin (7.5) is lower than that for rifampicin (67), it is less protein-bound (71% vs 85%), and is more strongly accumulated by cells (ratio of intracellular to extracellular concentrations 9 vs 5). Rifabutin is much more fat soluble than rifampicin resulting in a much larger volume of distribution (9.3 L/kg vs 1 L/kg) and higher tissue/plasma concentration ratios. It also has several theoretical pharmacokinetic advantages that include minimal induction of CYP3A4/5, its principal metabolizing enzyme in the liver, and absorption that is typically unaffected by food (Burman 2001). The in vitro MIC of M. tuberculosis is lower for rifabutin (< 0.06 mg/mL) than for rifampicin (0.15 mg/mL) (Heifets 1988). Furthermore, rifabutin may retain its activity against isolates resistant to rifampicin in 10% to 30% of cases, possibly due to differences in affinity for mycobacterial RNA polymerase or additional inhibition of DNA biosynthesis (Ungheri 1984; Cavusoglu 2004). Studies of pharmacodynamics in a mouse model supported these pharmacokinetic and pharmacodynamic data and suggested that M. tuberculosis infection was eradicated approximately twice as quickly with rifabutin as with rifampicin (Ji 1993).
This evidence from animal models or studies of relevant pharmacokinetic and pharmacodynamic properties in humans suggests that rifabutin may have the capability of accelerating the sterilization phase of treatment that could result in shorter treatment regimens for tuberculosis. Are these potential advantages of rifabutin realized in clinical research? In the treatment of human tuberculosis the drug has been used in three clinical situations: previously untreated disease; multidrug-resistant disease (Grassi 1996); and in HIV-associated tuberculosis where drug interactions between non-nucleoside reverse transcriptase inhibitors and particularly protease inhibitors and rifampicin have been a problem (CDC 2000). This review summarizes and evaluates the existing evidence from clinical trials comparing rifampicin- with rifabutin-containing regimens in these three situations.
Objectives
To compare combination drug regimens containing rifabutin with those containing rifampicin for treating pulmonary tuberculosis.
Methods
Criteria for considering studies for this review
Types of studies
Randomized and quasi-randomized controlled trials.
Types of participants
People being treated for pulmonary tuberculosis confirmed by sputum smear and/or culture.
Types of interventions
Intervention
Combination antituberculous drug regimens containing rifabutin given daily or two or three times weekly.
Control
Otherwise identical comparator regimen containing rifampicin.
Types of outcome measures
Primary
Cure (single negative M. tuberculosis culture at the completion of six months of therapy).
Secondary
- Relapse (single positive M. tuberculosis culture up to two years after the completion of therapy).
- Sputum smear and/or M. tuberculosis culture status two months after starting therapy.
- Sputum smear and/or M. tuberculosis culture status three months after starting therapy (added to protocol post-hoc).
- Median time to sputum smear and/or M. tuberculosis culture conversion on therapy.
- Hazard ratio of sputum smear and/or M. tuberculosis culture conversion on therapy.
Adverse events
- Serious adverse events (death, leading to hospitalization or continuation of hospitalization, life threatening, or persistent or significant disability).
- Adverse events leading to discontinuation of treatment.
- Other adverse events.
Search methods for identification of studies
We attempted to identify all relevant trials regardless of language or publication status (published, unpublished, in press, and in progress).
Databases
We searched the following databases using the search terms and strategy described in Appendix 1: Cochrane Infectious Diseases Group Specialized Register (July 2009); Cochrane Central Register of Controlled Clinical Trials (CENTRAL), published in The Cochrane Library (2009, Issue 3); MEDLINE (1966 to January 2007); EMBASE (1974 to January 2007); and LILACS (1982 to January 2007).
Conference proceedings
We searched the following conference proceedings for relevant abstracts from meetings held by the following organizations between 2000 and November 2006: International Union against Tuberculosis and Lung Disease; American Thoracic Society; British Thoracic Society; and the International Conference on Antimicrobial Agents and Chemotherapy.
Researchers, organizations, and pharmaceutical companies
We contacted Dr Andrew Vernon at the CDC Tuberculosis Clinical Trials consortium (January 2006), Drs Piero Olliaro and Phillip Onyebujoh at the Special Programme for Research and Training in Tropical Diseases (TDR) (September 2006), Dr Douglas Ross at Pfizer Inc (September 2006), Prof Dennis Mitchison (St George's Hospital Medical School), and Prof Andrew Nunn (MRC Clinical Trials Unit) regarding relevant unpublished or ongoing studies.
Reference lists and handsearching
We checked the reference lists of all the study reports retrieved by the above methods for any unidentified trials. We also handsearched the Indian Journal of Tuberculosis (1983 to July 2006; indexed on MEDLINE post-2006).
Data collection and analysis
Selection of studies
G Davies (GD) scrutinized the abstracts identified through the searches for potential relevance and retrieved the full-text versions of relevant abstracts. GD and S Cerri (SC) independently applied the inclusion criteria to the full-text versions using an eligibility form. Initial agreement was only fair (kappa = 0.25) but disagreements, mostly caused by the use of monotherapy as the intervention in some studies, and definition of study endpoints, were easily resolved after further discussion between the three authors.
Data extraction and management
GD and SC independently extracted data onto data extraction forms; GD then imported the data into Review Manager 4.2. Discrepancies were resolved by further discussion between all three authors. For each outcome, the number of participants randomized and the number analysed in each treatment arm were extracted to allow assessment of loss to follow up. For dichotomous outcomes at fixed time points (sputum smear/culture status at two, three, and six months), we extracted the number of participants experiencing the event and the number assessed in each arm (negative/positive/lost to follow up). Relapse data were also expressed as proportions at the end of follow up since no other measures were made available in the trial reports. Time-to-event outcome measures were intended to be summarized only by extracting median times to or hazard ratios for smear or culture negativity where available. For most of the included trials these details were not reported, with only quoted P-values from Cox modelling or the log-rank test, and we were unable to obtain either the results of further unpublished analyses or individual patient data from the trial authors.
Assessment of risk of bias in included studies
GD and SC independently assessed the methodological quality of the retrieved trials using a methodological quality form. We classified the generation of allocation sequence and allocation concealment as adequate, inadequate, or unclear (Juni 2001); and described who was blind to the interventions. We classified the inclusion of all randomized participants (proportion of participants included for which an efficacy endpoint was available) as adequate (if > 90%) or inadequate (if ≤ 90%). Disagreements were resolved by discussion between the two authors.
Data synthesis
GD analysed the data using Review Manager 4.2. The primary outcome and most of the secondary outcomes were analysed as a comparison of proportions using risk ratio (RR) as a measure of effect and presented with 95% confidence intervals (CI). Intention-to-treat analyses on an available-case basis were possible and are presented for all of these outcomes. Best-case and worst-case analyses were also carried out for the relapse outcomes due to the differing quality of follow up between the two largest included trials; they are referred to only in the text. Heterogeneity amongst the included trials was sought by inspection of the forest plot and by formal testing using both the chi-squared statistic with a significance level of 5% and the I
Results
Description of studies
See: Characteristics of included studies; Characteristics of excluded studies.
Trial selection
We identified eight trials with the search strategy. Five trials and a total of 924 participants were included in the review (see 'Characteristics of included studies'). We excluded three studies: two were monotherapy studies of early bactericidal activity and did not report on outcome measures relevant to the review (Chan 1992; Sirgel 1993); and one was a cohort study of rifabutin therapy in HIV-positive patients on antiretroviral therapy that did not involve a rifampicin control arm (Burman 2006); see 'Characteristics of excluded studies'.
Trial characteristics
The included trials were conducted between 1992 and 1996. Two trials were moderately large with a total of 818 participants (Gonzalez 1994; McGregor 1996), while the three other trials were smaller with 106 participants in total (HKCS/BMRC 1992; Rowinska 1992; Schwander 1995). Two were conducted in Africa (one in each of Uganda and South Africa), one in Hong Kong, one in Poland, and one was a multicentre trial involving participants in Argentina, Brazil, and Thailand.
Participants
The trials involved diverse groups of participants: Rowinska 1992 involved Polish people with new or chronic disease; HKCS/BMRC 1992 was a paired study of Chinese people with multidrug-resistant tuberculosis (and was designed to detect response to rifabutin in the presence of established rifampicin-resistance); Gonzalez 1994 and McGregor 1996 included people with previously untreated tuberculosis in Africa, South-East Asia, and South America; while Schwander 1995 was restricted to HIV-positive Ugandan people with previously untreated disease. Overall, 90% of participants in the trials included were believed to be HIV negative, and none of the trials provided antiretroviral therapy.
Interventions
The trials employed regimens with three different doses of rifabutin (150 mg, 300 mg, or 600 mg) and also differed with respect to whether the dose was adjusted according to bodyweight, as summarized in Appendix 2.
Outcome measures
Gonzalez 1994 and McGregor 1996 were the only two trials to report on cure and relapse. Smear conversion only at two months was reported in three trials (HKCS/BMRC 1992; Rowinska 1992; Schwander 1995), and culture conversion at two months was reported only in McGregor 1996. McGregor 1996 and Gonzalez 1994 also reported on culture conversion at three months; since this outcome measure also appeared potentially informative, we included it in the review. Also, three trials carried out some form of time-to-event analysis using either smears (Schwander 1995) or cultures (Gonzalez 1994; McGregor 1996), though we could extract relevant time-to-event outcomes from only Gonzalez 1994.
Adverse events were presented in four of the trials (HKCS/BMRC 1992; Rowinska 1992; Gonzalez 1994; McGregor 1996) as total numbers of adverse events, serious adverse events, and proportion of participants experiencing them during study follow up, rather than as rates. Only two trials, Gonzalez 1994 and McGregor 1996, specified whether adverse events resulted in discontinuation of treatment. The fifth trial, Schwander 1995, did not present data for adverse events, stating only that "no major differences between regimens were detectable"; no further information could be obtained.
Risk of bias in included studies
Generation of allocation sequence
One trial reported an adequate method of randomization (Schwander 1995); the method was unclear in the other trials.
Allocation concealment
Allocation concealment was adequate only in the three smaller trials and used either central randomization (HKCS/BMRC 1992; Rowinska 1992) or sealed envelopes (Schwander 1995). No details concerning allocation concealment were given in Gonzalez 1994 or McGregor 1996.
Blinding
Four of the included trials used blinding for the assessor only. Though Schwander 1995 reported blinding for the investigator and assessor, protection was weak since the drugs were formulated differently and no placebos were used. While the assessment of bacteriological outcomes in the laboratory appeared adequately blinded in all of the trials, this was not true for assessment of adverse event outcomes.
Inclusion of all randomized participants
The proportion of participants included for which an efficacy endpoint was available in the two trials that reported on cure and relapse, Gonzalez 1994 and McGregor 1996, was inadequate (≤ 90%). HKCS/BMRC 1992 was also assessed as inadequate, while Rowinska 1992 and Schwander 1995 both included more than 90% of all randomized participants and were assessed as adequate.
The reports of the two larger trials, Gonzalez 1994 and McGregor 1996, identified bacteriological conversion as the (composite) primary outcome measure. This was not predefined in the other trials, and no trial presented a power calculation.
Effects of interventions
Cure and relapse
Two trials reported these outcome measures (Gonzalez 1994; McGregor 1996). On the basis of available cases, both rifampicin- and rifabutin-containing regimens achieved acceptable cure rates (≥ 95%). Relapse rates reported in McGregor 1996 (8% to 11%) were more than four-fold higher than those in Gonzalez 1994 (0.8% to 1.8%). However, this heterogeneity was not statistically significant as assessed by the chi-squared and I
There was no statistically significant difference in cure (RR 1.00, 95% CI 0.96 to 1.04; 553 participants, 2 trials, Analysis 1.1) or relapse (RR 1.23, 95% CI 0.45 to 3.35; 448 participants, 2 trials, Analysis 1.2) despite there being numerically more relapses for the rifabutin-based regimens in both trials. These results were unaffected by rifabutin dose (150 mg or 300 mg). Follow up to 24 months was only 38% in McGregor 1996, with 'best-case' estimates of 3% and 4.1% relapse for the rifampicin- and rifabutin-containing regimens respectively. While Gonzalez 1994 achieved 68% follow up, this included only 75% of the cohort at the time of the trial report with 'best-case' estimates of 0.6% and 1.2% for the rifampicin- and rifabutin-containing regimens.
Smear and culture conversion
All of the included trials reported one or more of these outcome measures. There were no statistically significant differences in sputum culture conversion at two months (214 participants, 1 trial, Analysis 1.3) or at three months (654 participants, 2 trials, Analysis 1.4). Results of survival analysis and median conversion time based on culture were reported only in Gonzalez 1994. Only the outcome of the analysis was reported, and this did not support any statistical differences between the treatment groups (median time to negative culture 34 versus 37 days for the rifampicin- and rifabutin-containing regimens, logrank test P = 0.59). McGregor 1996 reported only the mean time to culture conversion (99 versus 100 days for the rifampicin- and rifabutin-containing regimens), so we could not combine these outcomes.
The three small trials relied on outcomes based primarily on sputum smear conversion and used different and incompletely reported measures of effect. Of these, only Schwander 1995 reported the results of a time-to-event analysis, which included Cox regression. After an adjustment for radiological cavitation, an apparent advantage for the rifabutin-containing regimen (P < 0.05) did not reach statistical significance (P = 0.1). Median times to smear conversion for the regimens were not reported.
Adverse events
Four trials reported information on adverse events (HKCS/BMRC 1992; Rowinska 1992; Gonzalez 1994; McGregor 1996). Overall reported proportions of participants experiencing any adverse event varied between trials, from 4% (McGregor 1996) to 70% (HKCS/BMRC 1992); the latter trial used higher doses of rifabutin in the context of different and more toxic companion drugs for multidrug-resistant tuberculosis and hence was not directly comparable with the other three trials. Even within the three trials evaluating first-line therapy, the incidence of adverse events varied widely; for example, 19% of participants receiving 300 mg rifabutin in Gonzalez 1994 experienced an adverse event compared to 4% in McGregor 1996, which used the same dose. Furthermore, dose adjustment by weight was not included in the protocol of either of these trials. On an available-case basis, there was no significant difference in the risk of adverse events between rifampicin- and rifabutin-containing regimens (RR 1.42, 95% CI 0.88 to 2.31; 714 participants, 3 trials, Analysis 1.5, though the RR increased from a dose of 150 mg rifabutin (RR 0.98, 95% CI 0.45 to 2.12; 264 participants, 2 trials, Analysis 1.5) to 300 mg (RR 1.78, 95% CI 0.94 to 3.34; 450 participants, 2 trials, Analysis 1.5). However, in no trial did there appear to be an increased incidence of specific relevant serious adverse events such as leucopoenia or hepatitis (common to all rifamycins) or of uveitis (specific to rifabutin), with no cases of the latter being reported in any of the included trials. In Gonzalez 1994, though it was claimed that adverse events tended to be classified as "severe" more often in the control arm, only 0.5% of controls discontinued whereas 3% of participants in the rifabutin 300 mg arm ultimately discontinued study medication. In McGregor 1996 the proportion discontinuing was 0.01% in both arms.
Discussion
We identified five trials directly comparing regimens containing rifabutin with rifampicin for treating tuberculosis, all of which date from the period shortly after licensing of the drug. None of these trials was of high methodological quality, they were conducted in diverse patient populations, and the outcome measures chosen by the investigators varied. The included trials comprised less than 1000 participants and were dominated by two moderately sized multicentre trials, conducted on three continents, and in which few HIV-positive individuals are likely to have been included. We assessed follow up in both of these trials as inadequate.
There was no evidence that rifabutin- and rifampicin-containing regimens differed in terms of efficacy as assessed by sputum culture conversion at two, three, or six months on treatment. The proportion of participants who relapsed after treatment was different in the two major trials reporting this endpoint (8% to 11% in McGregor 1996 and 0.8% to 1.8% in Gonzalez 1994). While treatment was administered daily throughout the trial in Gonzalez 1994, it was given twice weekly in the continuation phase in McGregor 1996. In addition, this finding may reflect the more intense transmission environment in South Africa and the fact that in neither study were relapses distinguished from reinfections using molecular methods. Though relapses occurred more frequently on rifabutin-containing regimens at either 150 mg or 300 mg, this was not a statistically significant finding, and the overall estimates of absolute relapse rates would be considered acceptable by current standards. In the context of multidrug-resistant tuberculosis, the only published study did not support a role for rifabutin in the treatment of people harbouring rifampicin-resistant organisms (HKCS/BMRC 1992). None of the trials stated whether demonstration of equivalence, non-inferiority, or superiority was the purpose of the primary comparison. None of the secondary outcome measures defined in the review give any substantial support to superiority of rifabutin and were not designed to provide information relating specifically to reducing the duration of treatment. No power calculations were presented and the size of the trials is certainly too small to evaluate superiority. There does not therefore currently appear to be any case for replacing rifampicin with rifabutin in the first-line regimen on the basis of efficacy alone, though it seems reasonable on the limited evidence available to assume that rifabutin-containing regimens are likely to be similar in efficacy for practical purposes to those containing rifampicin. However, given that the CI for relapse currently includes a RR as high as three and the poor quality of follow up in the included trials to date, new, higher quality, equivalence trials would be needed to provide further reassurance on this point.
Participants taking rifabutin-containing regimens were reported to have a similar number of adverse events as those taking rifampicin-containing regimens in the three trials of first-line treatment that reported adverse event data. Higher doses of rifabutin (300 mg) were associated with an increasing proportion of participants experiencing any adverse event, but at neither dose level did this proportion differ significantly from the standard dose of rifampicin (600 mg). In the largest trial, discontinuation rates in the rifabutin arm were as high as 3% on the higher dose. Notably, however, few of these adverse events were serious and did not include any of those of particular concern such as leucopoenia, hepatitis, or uveitis. Furthermore, the absence of any dose adjustment by weight in the two largest trials could have increased the frequency of adverse events in the rifampicin and rifabutin 300 mg arms. However, given that the review provides no evidence to prefer the higher dose of rifabutin in terms of efficacy, a case can be made for using lower doses in future trials or at the very least ensuring that adjustment of dose according to weight is used.
This review identified only one trial comprising a small number of HIV-positive participants who were not receiving antiretroviral therapy (Schwander 1995). Though this trial was methodologically sound it did not report outcome measures that could easily be compared with those of the other trials. HIV-positive people, who constitute the majority of tuberculosis patients in sub-Saharan Africa, are therefore currently under-represented in this review. Future trials in this group will be more complex to conduct than those included in the review to date, since their design must necessarily also include evaluation of the effect of the antiretroviral regimen selected. However, there is a clear need for more information about the use of rifabutin in these people, since it is precisely this group in whom the greatest practical benefits of substituting rifabutin for rifampicin can be envisaged.
Authors' conclusions
Implications for practice Rifabutin-containing regimens perform as well as rifampicin-containing regimens in achieving cure and preventing relapse, but higher doses of rifabutin may be associated with more adverse events and discontinuations. There is no evidence currently to support the replacement of rifampicin by rifabutin for the treatment of new cases of tuberculosis on the basis of efficacy. However, HIV-positive people were under-represented in the included trials and are the group most likely to benefit from substitution of rifampicin with rifabutin due to its lack of interaction with antiretroviral drugs. |
Implications for research Attempting to establish superiority of rifabutin alone in larger trials may not be a worthwhile goal, but well-designed and executed equivalence trials with more precise confidence limits would be useful. Further trials evaluating the use of rifabutin and rifampicin in conjunction with antiretroviral therapy for people with HIV-related tuberculosis must be a priority since no trials have yet evaluated this combined intervention, which has the potential to greatly simplify their care. |
Acknowledgements
The editorial base for the Cochrane Infectious Diseases Group is funded by the UK Department for International Development (DFID) for the benefit of developing countries.
Data and analyses
- Top of page
- Background
- Objectives
- Methods
- Results
- Discussion
- Authors' conclusions
- Acknowledgements
- Data and analyses
- Appendices
- What's new
- History
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Appendices
Appendix 1. Search methods: detailed search strategies
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Appendix 2. Rifabutin dose adjustment according to weight
| |||||||||||||||||||||||||||||||||
E: ethambutol; H: isoniazid; R: rifampicin; Rb: rifabutin; Z: pyrazinamide.
What's new
Last assessed as up-to-date: 5 July 2007.
| |||||||||||||||
History
Protocol first published: Issue 1, 2005
Review first published: Issue 4, 2007
Contributions of authors
G Davies formulated the review concept, wrote the protocol, extracted data, carried out the analysis, wrote the first draft, and co-authored the final draft of the review. S Cerri assisted with protocol development, extracted data, and co-authored the final draft of the review. L Richeldi assisted with protocol development, arbitrated during data extraction, and co-authored the final draft of the review.
Declarations of interest
G Davies is a co-applicant for research funding of a large randomized controlled trial evaluating the use of rifampicin and rifabutin in conjunction with antiretroviral therapy. None known for S Cerri and L Richeldi.
Sources of support
Internal sources
- No sources of support supplied
External sources
- The Wellcome Trust, UK.
Differences between protocol and review
2007, Issue 4 (first review version): We included three-month sputum culture conversion results as a secondary outcome measure since it also appeared potentially informative.
Index terms
Medical Subject Headings (MeSH)
Antibiotics, Antitubercular [adverse effects; *therapeutic use]; Randomized Controlled Trials as Topic; Rifabutin [adverse effects; *therapeutic use]; Rifampin [therapeutic use]; Tuberculosis, Pulmonary [*drug therapy]; Uveitis [chemically induced]
MeSH check words
Humans