Published ahead of print on August 14, 2006, doi:10.1164/rccm.200605-637OC
© 2006 American Thoracic Society doi: 10.1164/rccm.200605-637OC
Dosing Schedules of 6-Month Regimens and Relapse for Pulmonary TuberculosisTuberculosis and Chest Service, Center for Health Protection, Department of Health; Tuberculosis and Chest Unit, Grantham Hospital; and Hong Kong Tuberculosis, Chest, and Heart Diseases Association, Hong Kong, China Correspondence and requests for reprints should be addressed to Dr. Kwok Chiu Chang, M.B., M.Sc., Yaumatei Chest Clinic, 2nd Floor, Yaumatei Jockey Club Polyclinic, 145 Battery Street, Kowloon, Hong Kong. E-mail: kc_chang{at}dh.gov.hk
Rationale: The optimal approach for reducing tuberculosis relapse is open. Objectives: We examined the possibility of reducing relapse by increasing dosing schedules.
Methods: We conducted a systematic review of published clinical trials involving adult cohorts with pulmonary tuberculosis treated using 6-mo rifamycin-containing regimens, which were grouped under seven categories ordered by dosing schedules. Assuming cavitation and positive 2-mo culture were the driving forces for relapse, a static deterministic model apportioned observed numbers with and without relapse in each cohort into eight subgroups. Combining subgroups stratified by cavitation, 2-mo culture, and regimens enabled estimation of adjusted relapse risks. Results: We identified 200 cases of bacteriologic relapse out of 5,208 patients in 32 cohorts. A logistic risk model showed a significant dose–response relationship between dosing schedules and relapse, with the following odds (95% confidence intervals) of relapse relative to daily regimens: 1.6 (0.6–4.1) for daily initial phase (IP) plus thrice-weekly continuation phase (CP), 2.8 (1.3–6.1) for daily IP plus twice-weekly CP, 2.8 (1.4–5.7) for thrice-weekly, 5.0 (2.4–10.5) for daily IP plus once-weekly rifapentine, and 7.1 (3.3–15.3) for thrice-weekly IP plus once-weekly rifapentine. In the presence of cavitation, only 6-mo daily or daily IP plus thrice-weekly CP attained best-estimated relapse risks below 5%; they reached 6% when 2-mo culture was also positive. Conclusions: Cavitary tuberculosis is best treated with 6-mo regimens comprising daily IP and thrice-weekly CP, which may be extended when 2-mo culture is positive.
Key Words: logistic models • relapse • risk factors • treatment • tuberculosis
Relapse, defined as the situation in which a patient becomes and remains culture negative while receiving antituberculosis drugs but develops active tuberculosis again after completion of treatment (1), is the best estimate of treatment efficacy (2). Its occurrence has been attributed to the failure of eradicating "persisters," which are characterized by having brief spurts of metabolic activity interspersed with periods of dormancy, in contrast to treatment failure attributable to the inability of killing rapidly dividing bacteria (3). The risk of relapse for 6-mo rifampin-containing regimens is generally below 5% regardless of the dosing schedule, which may be daily, partly intermittent, or fully intermittent (4). The World Health Organization has recommended that 6-mo regimens may be given daily or thrice weekly during the initial (IP) or continuation phase (CP) (5). Thrice-weekly regimens, which facilitate treatment supervision and reduce drug costs, have been widely used with good results in high-burden countries (6, 7). The coexistence of certain factors can amplify the risk of relapse to 20% and above (1, 8). Among the multiple risk factors of relapse of pulmonary tuberculosis that have been studied (2, 8–20), the most important disease-related factors are probably initial cavitation and positive culture on completion of the 2-mo initial phase (2, 8, 12, 19, 20). On the basis of findings from observational studies (1, 15), the American Thoracic Society has recommended extending 6-mo regimens by 3 mo if the initial chest radiograph shows cavitation and the 2-mo sputum culture is still positive (1). The optimal approach for reducing relapse is still open. Although there are insufficient data in the published literature for comparing fully intermittent and daily short-course regimens (21–23), an observational study has suggested that treatment with 6-mo daily regimens in comparison with thrice-weekly regimens using the same drugs reduces relapse (20). This raises the possibility of reducing relapse by increasing dosing frequency. Although the unfavorable impact on treatment supervision and increased drug costs make 6-mo daily regimens unattractive, daily treatment during the IP followed by thrice-weekly treatment during the CP may strike the best balance between treatment efficacy and practical concerns. Unfortunately, no randomized controlled trials have compared the treatment efficacy of daily and partly intermittent regimens; furthermore, none has focused only on cavitary tuberculosis, and few have distinguished between cavitary and noncavitary tuberculosis in the evaluation of relapse (8, 24). Given the practical difficulty of conducting a randomized trial with sufficient statistical power and the lack of sufficient material for performing a meta-analysis, we assessed the effect of dosing schedules on the risk of relapse for pulmonary tuberculosis by conducting a systematic review supplemented by mathematical modeling and logistic regression analysis.
We conducted a systematic review of the published literature by searching MEDLINE for clinical trials involving adult cohorts with pulmonary tuberculosis treated using 6-mo rifamycin-containing antituberculosis regimens. Search terms included "tuberculosis" and "rifampicin," "rifampin," "Rifinah," or "Rifater" in the title or abstract. The search was supplemented by the bibliography of a review article (25). Only studies that comprised 6-mo rifamycin regimens supplemented by at least 8 wk of pyrazinamide were included. Studies were excluded if rifabutin was used, or if more than 5% patients were known to have any one of the following: multidrug-resistant tuberculosis, HIV infection, or treatment failure. Studies that lacked the following data were also ineligible: proportion of patients with cavitation on the initial chest radiograph, the number of patients who completed treatment, the number of relapse, and the duration of follow-up.
Assuming that pyrazinamide plays a pivotal role only in the first 2 mo of treatment (26), that streptomycin and ethambutol are supplementary rather than essential, and that once-weekly rifapentine–isoniazid therapy with the 600-mg rifapentine dose was significantly less effective than twice-weekly rifampin and isoniazid (27), we grouped cohorts under seven categories ordered by dosing schedules (see Table 1). The
Assuming that initial cavitation and 2-mo sputum culture were the major driving forces for relapse, observed numbers of patients with and without relapse in each cohort were apportioned into eight subgroups by a static deterministic model with the following data and parameters (see Figure 1): the number of patients included in analysis of bacteriologic relapse, the number of patients with bacteriologic relapse, the respective proportions of initial cavitation and positive 2-mo sputum culture, the risk of relapse among subjects with initial cavitation compared with the risk among those without initial cavitation (RR1), the risk of relapse among subjects with positive 2-mo sputum culture compared with the risk among those with negative 2-mo sputum culture (RR2), and the risk of positive 2-mo sputum culture among subjects with initial cavitation compared with the risk among subjects without initial cavitation (RR3). If the proportion of positive 2-mo sputum culture was missing, we used the weighted average value of similar cohorts. The online supplement contains the mathematical equations and steps.
Combining subgroups stratified by initial cavitation, 2-mo sputum culture, and regimen category enabled estimation of the adjusted risks of relapse for 6-mo regimens with different dosing schedules. We tested for any dose–response relationship between dosing schedules and relapse with the  2 test for trend and logistic regression analysis, which forced in four main predictor variables: initial cavitation, 2-mo sputum culture, treatment supervision, and regimen categories ordered by dosing schedules. Finally, we performed sensitivity analysis by substituting the best-estimated values of RR1, RR2, and RR3 with their upper and lower 95% confidence limits in eight combinations (see Table E1 of the online supplement). OpenOffice.org 2.0 (www.openoffice.org) and SPSS version 10 (SPSS, Inc., Chicago, IL) were used for statistical analysis. Approval was obtained from the institutional review board for conducting the review. Patient consent was not necessary because our review did not involve any intervention on patients.
A total of 36 studies were identified. Nineteen studies were excluded for the following reasons: 16 for lack of data on cavitation or relapse or both, one for tuberculosis–HIV coinfection, one for retreatment of failure, and one for being heterogeneous by risk of relapse. We identified 200 cases of bacteriologic relapse out of 5,208 patients with pulmonary tuberculosis in 32 cohorts extracted from 17 studies. Only one included study (8), the same study that provided the reference values for RR1 and RR2, provided data on the distribution of relapse according to initial cavitation and 2-mo sputum culture. Please refer to Tables E2 and E3 for details of included cohorts and excluded studies.
Dose–Response Relationship A logistic risk model of relapse controlling for initial cavitation, 2-mo sputum culture, and treatment supervision (see Table 2) also showed a significant dose–response relationship (p < 0.001) between dosing schedules of 6-mo regimens and relapse, with the following odds (95% confidence intervals) of relapse relative to daily regimens: 1.6 (0.6–4.1) for regimens with daily IP and thrice-weekly CP, 2.8 (1.3–6.1) for regimens with daily IP and twice-weekly CP, 2.8 (1.4–5.7) for thrice-weekly regimens, 5.0 (2.4–10.5) for regimens with daily IP followed by once-weekly rifapentine, and 7.1 (3.3–15.3) for regimens with thrice-weekly IP followed by rifapentine once weekly or less. The odds of relapse for 6-mo daily regimens are significantly lower than those of the other dosing schedules except for 6-mo regimens with daily IP and thrice-weekly CP. Initial testing of linearity for dosing schedules with the likelihood ratio test suggested a nonlinear effect (likelihood ratio statistic = 25.9, degree of freedom = 5, p < 0.05). However, repeating the test after excluding the single cohort of the twice-weekly regimen suggested linearity (likelihood ratio statistic = 3.0, degree of freedom = 4, p > 0.05).
Adjusted Risks of Relapse Table 1 summarizes the risks of relapse adjusted for initial cavitation and 2-mo sputum culture. For patients with cavitary tuberculosis (regardless of 2-mo sputum culture), the best-estimated adjusted risk of relapse exceeded 5% for all dosing schedules except for 6-mo daily regimens, 6-mo regimens with daily IP and thrice-weekly CP, and the 6-mo twice-weekly regimen. However, the last regimen, with only one cohort involving 57 patients with cavitary tuberculosis (see Table E4), was an outliner in the dose–response analysis. Random error was probably involved and the 95% confidence interval was very wide. In the presence of both initial cavitation and positive 2-mo sputum culture, the best-estimated adjusted risks of relapse were around 6% for both 6-mo daily regimens and 6-mo regimens with daily IP and thrice-weekly CP. In the presence of either risk factor alone, most regimens (except for regimens with rifapentine during the CP) gave acceptable best-estimated risks of relapse, with reasonable 95% confidence intervals. All regimens were acceptable for noncavitary tuberculosis with negative 2-mo sputum culture.
Power of Prediction and Sensitivity Analysis
Published clinical trials do not permit direct comparison of 6-mo regimens with different dosing schedules. Although a static deterministic model combined with logistic regression analysis may not be the ideal approach, there is probably no alternative in the absence of randomized controlled trials specifically designed for comparison of different regimens for different patient categories. We addressed missing data on confounding factors by apportioning observed data in a static deterministic model using parameters for major confounding factors identified from the published literature. The absence of major differences (e.g., HIV infection and resistance to rifampin) in the clinical profile and the absence of significant heterogeneity in the odds of relapse among cohorts grouped under the same regimen category lent support to combining subgroups stratified by 6-mo regimen category, cavitation, and 2-mo sputum culture. Minor heterogeneities, such as the duration of pyrazinamide in 6-mo daily regimens (26) and the presence of mild initial resistance (28), have been shown to be clinically insignificant. The robustness of our approach on sensitivity analysis and consistency with the published literature (1, 8, 20) substantiated the validity of our findings. How well our static deterministic model fit the observed data could not be evaluated directly because only one study (8) provided observed data on the distribution of relapse stratified by initial cavitation and 2-mo sputum culture. The area under the receiver operating curve (ROC) plotted from the logistic risk model is probably the best approach for assessing the goodness of fit of our static deterministic model, which fit the derived data best when both RR1 and RR2 reached the upper 95% confidence limit (power of prediction = 0.83), and worst when both RR1 and RR2 reached the lower 95% confidence limit (power of prediction = 0.71). Varying RR3 had negligible impact on the area under the ROC. Other factors that might have reduced goodness-of-fit included misclassification bias for initial cavitation and 2-mo sputum culture, limited sensitivity of 2-mo sputum culture, and the omission of other possibly less important risk factors, such as mild coexisting diseases that impair host immunity. In contrast to a Cochrane review that found no significant difference between the effectiveness of intermittent and daily rifampicin-containing short-course regimens in adults with pulmonary tuberculosis (23), our systematic review adopted an original approach and reaffirmed that 6-mo thrice-weekly regimens in comparison with 6-mo daily regimens significantly increased the risk of relapse (20). Furthermore, we found a significant dose–response relationship between dosing schedules and relapse. Reducing the dosing frequency in the IP from daily to thrice weekly increases the best-estimated odds ratio for relapse from 1.6 (0.6–4.1) for regimens with daily IP and thrice-weekly CP to 2.8 (1.4–5.7) for the thrice-weekly intermittent regimen. Reducing the dosing frequency in the CP of regimens with daily IP from thrice weekly to twice weekly and once weekly increases the best-estimated odds ratio for relapse from 1.6 (0.6–4.1) to 2.8 (1.3–6.1) and 5.0 (2.4–10.5), respectively. Our review showed no significant association between directly observed treatment and relapse. This might be explained by the clinical trial setting in which study subjects might be more compliant with treatment than patients treated in the service setting. It has been reported that treatment duration is more important than the total number of doses in determining the risk of relapse (29–31). Our review suggests that the total number of doses determines the risk of relapse when treatment duration is fixed. Both observations are complementary to the understanding of the biological mechanism of relapse. "Persisters," the putative cause of relapse, are susceptible to drug treatment only during occasional and brief spurts of metabolism that last for hours. Although the majority of rapidly dividing bacteria are inhibited during the "lag period" (32), a factor that establishes the scientific basis of intermittent regimens (4), only a minority of persisters in brief spurts of metabolism are eradicated after each dose of drugs; among these drugs, rifampin is the pivotal drug by virtue of its rapid onset of sterilizing action (33). The probability of eradicating persisters is thus largely determined by the total number of doses of rifampin, which can be increased by modifying dosing frequency, treatment duration, or both. The higher the dosing frequency or the longer the treatment duration, the lower the proportion of surviving persisters. Assuming that the number of surviving persisters on completion of treatment determines relapse, the risk of relapse is very significant for patients with high initial bacillary load treated intermittently for less than 6 mo, the optimal treatment duration for most patients (34). Table 3 summarizes our antituberculosis treatment recommendations based on this systematic review. All 6-mo regimens that we have examined under our study criteria are sufficient for treating noncavitary tuberculosis with negative 2-mo sputum culture. For cavitary tuberculosis with negative 2-mo culture, we recommend 6-mo regimens comprising a daily IP and thrice-weekly CP. In the presence of both initial cavitation and positive 2-mo culture, even the most efficacious 6-mo daily regimens may be barely adequate. Thus, from the observation that better efficacy is attainable by spreading fewer doses evenly over a longer treatment period (29–31), we recommend extending the thrice-weekly CP by 3 mo. This agrees with the recommended treatment for cavitary tuberculosis and strikes the best balance among treatment efficacy, drug costs, and convenience of supervision.
In summary, less frequent dosing for 6-mo regimens increases relapse with a significant dose–response relationship. Our findings further substantiate that "you get what you pay for in the treatment of tuberculosis" (35).
This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Originally Published in Press as DOI: 10.1164/rccm.200605-637OC on August 14, 2006 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form May 10, 2006; accepted in final form August 8, 2006
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
