Comparative efficacy and safety of high-dose rifamycin regimens for tuberculosis treatment: a Bayesian network meta-analysis

Abstract

Objectives

High-dose rifamycin (HDR) regimens have demonstrated significant potential in tuberculosis (TB) treatment. This study aims to evaluate the efficacy and safety profile of different HDR regimens.

Design

Using a systematic review and Bayesian network meta-analysis (NMA).

Data sources

PubMed, Web of Science, Cochrane Library and Embase were searched up to 2 November 2024.

Eligibility criteria for selecting studies

Randomised controlled trials that compared the efficacy and safety of HDR regimens (rifampin 15–30 mg/kg/day and rifapentine 7.5–20 mg/kg/day) to standard-dose rifampin in patients with pulmonary drug-susceptible TB were included.

Data extraction and synthesis

The risk of bias was assessed using Cochrane tools. We conducted NMA with GEMTC in R. The simulation was performed using the Markov Chain Monte Carlo technique set on four parallel chains, with 20 000 burn-in iterations, 50 000 inference iterations and a thinning factor of n=2.5. To check for model convergence, Gelman and Rubin diagnostic plots and density plots were applied. We assessed heterogeneity using the I² test, evaluated transitivity by comparing effect modifiers across studies and examined consistency via node-splitting analysis. The confidence in network meta-analysis online tool and Cochrane Risk of Bias 2.0 Tool were used to assess evidence certainty and risk of bias, respectively. Higher surface area under the cumulative rank curve scores indicated a higher probability of top-ranking treatments.

Results

Out of 15 766 citations screened, 15 randomised controlled trials were included, encompassing 6456 subjects. The risk of bias was low in 14 studies, with some concerns in one. Patients receiving rifapentine 20 mg/kg/day (risk ratio, 1.09; 95% credible interval, 1.03 to 1.17) had higher culture conversion rates at 8 weeks in solid culture compared with the control. There was no significant difference in primary efficacy within all HDR regimens. Rifapentine 20 mg/kg/day was ranked as the most effective intervention for primary efficacy. No statistical difference in the incidence of serious adverse events was found between all regimens.

Conclusions

Rifapentine 20 mg/kg/day may be the most effective for achieving the strongest anti-TB activity. All HDR regimens demonstrated good safety.

PROSPERO registration number

CRD42024504575.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This study uses the Bayesian network meta-analysis study to evaluate the efficacy and safety of various rifamycin dosing regimens for drug-susceptible tuberculosis.

• The high quality of the included trials and the rigorous analysis methods used to assess bias enhanced the credibility of the findings.

• Due to the limited number of studies involving people living with HIV, individuals with low body weight or those with diabetes mellitus, we were unable to conduct subgroup analyses.

Introduction

Tuberculosis (TB) is a major global healthcare challenge and a leading cause of death from infectious pathogens.¹ In 2023, 10.8 million people contracted TB and 1.25 million died from the disease.² The WHO has developed the End TB Strategy, which outlines a global approach for preventing and controlling TB. The goal of the strategy is to reduce TB incidence by 90% by 2035 compared with 2015.³ To achieve this ambitious goal and end the TB epidemic, effective and safe treatment is critical.⁴

There is growing evidence that the currently recommended doses of rifampin and rifapentine may be suboptimal.^{5 6} For rifampicin, the current WHO guideline recommending 10 mg/kg has remained unchanged since its introduction in 1971.⁷ This dosage was initially based on toxicological and financial concerns, with limited pharmacokinetic data available at the time.⁷ Subsequent researchers found adverse events (AEs), like influenza-like reactions, were associated with intermittent administration, suggesting that intermittent use may have distorted the potential benefits of high-dose rifampicin in early studies.^{8 9} Growing evidence demonstrated that 10 mg/kg of rifampicin was at the low end of the dose-exposure curve, and inadequate drug exposure was significantly associated with treatment failure and drug resistance.^{10 11} Rifapentine, as a cyclopentyl derivative of rifampicin, has shown favourable pharmacokinetic and safety profile in several phase I and II clinical trials at higher doses.^{12 13} The use of high-dose rifapentine has been reported to achieve higher culture conversion rates at 8 weeks of treatment with a good safety profile.^{13 14}

There has been an arising debate on the optimal doses of rifampin and rifapentine that would achieve the strongest bactericidal activity while maintaining safety.¹⁵ Different doses of high-dose rifampin and rifapentine regimens have been explored, including rifampin 13–15 mg/kg/day (R15), rifampin 20 mg/kg/day (R20), rifampin 25 mg/kg/day (R25), rifampin 30 mg/kg/day (R30), rifampin 35 mg/kg/day (R35), rifapentine 7.5 mg/kg/day (P7.5), rifapentine 10 mg/kg/day (P10), rifapentine 15 mg/kg/day (P15) and rifapentine 20 mg/kg/day (P20). However, there is a paucity of head-to-head comparisons of different regimens in randomised controlled trials (RCTs), and the comparative efficacy and safety of these interventions remain unclear.

Furthermore, compared with standard-dose rifampin, the superiority of high-dose rifamycin (HDR) remains a debate. The choice between high-dose rifapentine and rifampin continues to be a focus of clinical and research interest and has yet to be fully resolved. One meta-analysis has explored the potential benefits of HDR regimens, with some reporting higher sputum culture conversion rates at 8 weeks when rifampin was administered at ≥20 mg/kg/day compared with the standard 10 mg/kg/day dose.¹⁶ Conversely, another study found no significant differences in efficacy or safety outcomes between high-dose and standard-dose regimens.¹⁷ Recently, several RCTs comparing HDR to standard-dose rifampin have been published.¹⁸,²⁰ A recent network meta-analyses (NMAs)²¹ suggested that optimal doses of rifampicin may be between 25 and 35 mg/kg/day, but this analysis did not include any data related to rifapentine. Besides, Peng et al performed an NMA of short-course rifamycin-based regimens for latent TB infection, which differs in population and treatment context from our study.²² Thus, we included the new evidence on efficacy and safety of various rifamycin regimens for drug-susceptible TB (DS-TB) treatment in this updated NMA.

Methods

Study design

A systematic review and NMA of RCTs comparing HDR with standard-dose rifampin in patients with pulmonary DS-TB was conducted. The study was designed and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for NMA.²³ This systematic review was registered in the International Prospective Register of Systematic Reviews (CRD42024504575).

Search strategy

We searched PubMed, Web of Science, Cochrane Library and Embase for original articles comparing different doses of rifamycin in DS-TB, from each database’s inception to 2 November 2024. Detailed search strategies are provided in online supplemental table S1. Briefly, the search strategy combined terms for TB, rifamycin and RCTs, with filters for human studies. Additionally, we scanned the reference lists of all full-text articles to identify other eligible studies. When additional information was required, we contacted the corresponding author via email.

Study selection and eligibility criteria

Two researchers (ZF and HWa) independently screened the titles and abstracts of studies and reviewed full-text articles for inclusion. Any disagreements were resolved through discussion with a third investigator. The following inclusion criteria were used: (a) providing original data from interventional studies, (b) investigating the efficacy or safety of standard-dose rifampin versus HDR in patients with pulmonary DS-TB, (c) reporting at least one of the following results: sputum culture conversion, treatment success (ie, cured, completed), relapse, incidence of serious AEs (SAEs), or grade 3 or higher AEs, and occurrence of grade 3 or higher adverse drug reactions (ADRs) and (d) publishing as a complete paper. Papers were excluded if they were: non-interventional studies, reviews, guides, case reports, conferences, in vitro studies, studies without sufficient information about outcomes of interest, studies without rifamycin or studies before 1990 (because around 1990, international guidelines began to recommend the current standard regimen as the first-line treatment for TB). We excluded studies with overlapping data and studies with missing or insufficient data after a reasonable attempt at contacting corresponding authors. Full-text articles unavailable after exhaustive searches were also excluded.

Data extraction and quality assessment

Two investigators (ZF and LS) independently extracted data from each eligible study using a predefined form. Any discrepancies would be resolved through discussion. If there were still differences after the discussion, a third investigator (YL) would be consulted to make the final decision. The following data were extracted: first author, year of publication, country of study, sample size, enrolment period, study design, regimen, number of patients for each regimen, type and dose of rifamycin, length of treatment, follow-up period, safety and efficacy outcomes of interest. When several researchers used the same database, all articles were examined to collect data completely. Two authors (ZF and QL) independently assessed the risk of bias using the Risk of Bias assessment tool RoB2 (revised version) for RCTs.²⁴ In the case of disagreement, a third author (YL) decided.

Definition and outcomes

We defined HDR as rifampicin mean dose >600 mg/day or >10 mg/kg/day, rifapentine mean dose >1200 mg/week, and rifabutin mean dose >300 mg/day or >5 mg/kg/day.¹⁷ A drug was defined as daily given if administered at least 5 days per week.

Modified intention-to-treat (ITT) data were used for evaluating efficacy where available. For primary efficacy outcome, we included the culture conversion rate at 8 weeks in solid culture after the first dose of treatment. The second efficacy outcome was culture conversion rate at 8 weeks in liquid culture after the first dose, treatment success (ie, cure, completed), relapse and the culture conversion rate at 12 weeks in liquid and solid culture after the first dose.

When applicable, we used ITT data for safety analysis. The primary safety endpoint of this NMA was the incidence of SAE. SAEs were defined as any event leading to death, life threatening, prolonged hospitalisation, permanent or serious disability, malformation or congenital defect of the offspring, or required hospital admission for management. When this information was unavailable, we used SAEs defined by each study as SAE. The secondary safety endpoint included the incidence of grade 3 or higher ADRs, grade 3 or higher AEs, grade 3 or higher increase of transaminases, AEs, occurrence of treatment discontinuation from trial for toxicity and trial withdrawal. We also defined mortality as a separate secondary safety endpoint, in addition to including it under the broader category of SAEs. The grade of AE was defined by Common Terminology Criteria for Adverse Events, Division of AIDS Table for Grading the Severity of Adult and Pediatric Adverse Events, or other criteria from international authorities.

Data synthesis and statistical analysis

The NMA was performed with R (V.4.3.1, R Foundation for Statistical Computing, Vienna, Austria). We generated plots depicting the network geometry, relative effects and risk of bias summary using Microsoft Excel and GeMTC R package (Gemtc V.1, Repository CRAN).^{25 26} Since no loops or design inconsistencies were present, the NMA was conducted using the Bayesian consistency model.²⁷ The simulation was performed using the Markov Chain Monte Carlo technique set on four parallel chains, with 20 000 burn-in iterations, 50 000 inference iterations, and a thinning factor of n=2.5. To check for model convergence, Gelman and Rubin diagnostic plots and density plots were applied.²⁸ A random-effects Bayesian NMA was carried out to summarise direct and indirect (ie, mixed) evidence. I² test was used to quantify the heterogeneity in the model. The heterogeneity was considered high if I² exceeded 75%. The transitivity assumption was evaluated by comparing the distributions of age, sex, body mass index (BMI) and clinical presentation across studies. Consistency was explored by using node split methods to examine whether indirect treatment effects (ie, treatment effects that were not directly compared in included studies) were similar or different from direct treatment effects (ie, treatment effects that were directly compared in included studies). Additionally, sensitivity analyses included omission of trials at moderate and high risk for bias, and leave-one-out analysis was performed for primary outcomes.

Risk ratio (RR) was used as the measure in the meta-analysis. We used RR with 95% credible interval (95% CrI) to represent the risk measure for dichotomous data. Besides, different doses of rifamycin were ranked using surface area under the cumulative rank curve (SUCRA value). Higher SUCRA scores indicated a higher probability of top-ranking treatments.²⁹ Therefore, we can evaluate multiple treatments and determine the likelihood of each being ranked first or second. In addition, we chose the standard-dose rifampin (10 mg/kg/day) as a common comparator to include all trials within one framework.

To assess the certainty of evidence, we used the confidence in network meta-analysis (CINeMA) to grade the quality of each comparison.^{30 31} This online tool was designed by the Cochrane Comparing Multiple Interventions Methods Group as an adaptation of the Grading of Recommendations, Assessment, Development and Evaluation for NMAs.³⁰

Results

Study selection

Through a comprehensive and systematic search of different databases, 15 766 studies were identified in the title and abstract screening. After preliminary screening, a total of 534 articles were retrieved for full-text evaluation, of which 15 articles fulfilled the inclusion criteria and were included in this study (figure 1).

Figure 1. Flowchart according to PRISMA guidelines. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT, randomised controlled trial.

Characteristics of the included studies

The characteristics of the included RCTs are outlined in table 1. Overall, 15 trials^{510 13 18},^{20 32} were included, involving a total of 6456 patients. Currently, no RCT concerning high-dose rifabutin in DS-TB has been published yet, thus different types of rifamycin only included rifampin and rifapentine (figure 2). Online supplemental table S2 summarises key demographic and clinical characteristics of study participants in each trial. Among all included trials, the weighted mean age and BMI were 32 years and 18.8 kg/m², respectively. The weighted proportion of men was 70.5%, suggesting the assumption of transitivity.

Table 1. Description of 15 studies included in the network meta-analysis.

First author, year	Country	Sample size	Study design	Regimen	No pts	Rifamycin dosage (mg/kg/day)	Follow-up period	Overall RoB assessment
Ruslami, 200732	Indonesia, Netherlands	47	Double-blind RCT	Control: 2HRZE/4HR	24	9.5	26w	Low risk
				R15: 2HRZE/4HR	23	12.9	26w
Dorman, 201233	America, South Africa, Uganda, Spain, Brazil, Peru and Vietnam	389	Double-blind RCT	Control: 2HRZE/4HR	183	10	8w	Low risk
				P10: 2HPZE/4HR	206	10	8w
Boeree, 201510	South Africa	68	Dose-Ranging Trial	Control: 2HRZE/4HR	8	10	12w	Low risk
				R20: 2HRZE/4HR	15	20	12w
				R25: 2HRZE/4HR	15	25	12w
				R30: 2HRZE/4HR	15	30	12w
				R35: 2HRZE/4HR	15	35	12w
Dorman, 201513	North America, Africa, South America, Asia, Europe	334	Dose-Ranging Trial	Control: 2HRZE/4HR	85	10	8w	Low risk
				P10: 2HPZE/4HR	87	10	8w
				P15: 2HPZE/4HR	81	15	8w
				P20: 2HPZE/4HR	81	20	8w
Conde, 201634	Brazil	121	Open-label RCT	Control: 2HRZE/4HR	59	10	8w	Low risk
				P7.5: 2HPZM/4HR	62	7.5	8w
Jindani, 201635	Bolivia, Nepal and Uganda	300	Single-blind RCT	Control: 2HRZE/4HR	100	9.6	16w	Low risk
				R15: 2HRZE/4HR	100	15	16w
				R20: 2HRZE/4HR	100	18.8	16w
Aarnoutse, 20175	Tanzania	150	Double-blind RCT	Control: 2HRZE/4HR	50	10.7	12w	Low risk
				R15: 2HRZE/4HR	50	16.7	12w
				R20: 2HRZE/4HR	50	21.4	12w
Boeree, 2016	Tanzania and South Africa	365	Open-label RCT	Control: 2HRZE/4HR	123	10	12m	Low risk
				R35: 3HRZE/3HR	63	35	12m
				Control: 3HRZQ/3HR	59	10	12m
				R20: 3HRZQ/3HR	57	20	12m
				R20: 3HRZM/3HR	63	20	12m
Velasquez, 201837	Peru	180	Triple-blind RCT	Control: 2HRZE/4HR	60	9.6	12m	Low risk
				R15: 2HRZE/4HR	60	13.7	12m
				R20: 2HRZE/4HR	60	18.8	12m
Atwine, 202038	Uganda	97	RCT	Control: 2HRZE/4HR	33	10	28w	Low risk
				R20: 2HRZE/4HR	64	20	28w
Maug, 202039	Bangladesh	698	RCT	Control: 2HRZE/4HR	346	10	12m	Some concerns
				R20: 2HRZE/4HR	352	20	12m
Dorman, 202140	Brazil, China (Hong Kong), Haiti, India, Kenya, Malawi, Peru, South Africa, Thailand, Uganda, the USA, Vietnam and Zimbabwe	2516	Open-label RCT	Control: 2HRZE/4HR	829	10	12m	Low risk
				P20: 2HPZM/2HPM	838	20	12m
				P20: 2HPZE/2HP	849	20	12m
Jindani, 202318	Uganda, Guinea, Peru, Nepal, Botswana and Pakistan	578	Open-label RCT	Control: 2HRZE/4HR	191	10	18m	Low risk
				R20: 2HRZE/2HR	192	20	18m
				R35: 2HRZE/2HR	195	35	18m
Paton, 202319	Indonesia, the Philippines, Thailand, Uganda and India	485	Open-label RCT	Control: 2HRZE/4HR	181	10	96w	Low risk
				R35: 2HRZELzd	184	35	96w
				R35: 2HRZECfz	78	35	96w
				P20: 2HPZLfxLzd	42	20	96w
Sekaggya-Wiltshire, 202320	Uganda	128	Open-label RCT	Control: 2HRZE/4HR	67	10	24w	Low risk
				R35: 2HRZE/4HR	61	35	24w

Cfz, clofazimine; DTG, dolutegravir; E, ethambutol; EFV, efavirenz; H, isoniazid; Lfx, levofloxacin; Lzd, Linezolid; m, month; M, moxifloxacin; P, rifapentine; Q, SQ109; R, rifampicin; RCT, randomised controlled trial; RoB, risk of bias; w, week; Z, pyrazinamide.

Figure 2. Network plots for negative culture conversion at 8 weeks in solid culture (A) and serious adverse event (B). Different nodes refer to different doses of rifamycin treatments. Lines connect different rifamycin treatments that have been investigated in head-to-head (direct) comparisons in the included studies. The width of the lines is proportional to the number of trials that directly compared the two rifamycin treatments.

In the NMA, all studies used rifampin 10 mg/kg/day (control) as a reference intervention (2398 (37.1%) patients). Patients were distributed across the investigated arms as follows: 233 (3.6%) in the R15 arm, 953 (14.8%) in the R20 arm, 15 (0.2%) in the R25 arm, 15 (0.2%) in the R30 arm and 596 (9.2%) in the R35 arm. In addition, 62 (1.0%), 293 (4.5%), 81 (1.3%) and 1810 (28.0%) patients were allocated to the P7.5, P10, P15 and P20 arm, respectively.

For the NMA on the efficacy of treatments, a total of 11 articles513 18 20 33,^{38 40} comparing culture conversion rate at 8 weeks in solid culture of different treatments in a direct analysis were included (online supplemental figure S1). For safety, a total of 12 articles^{510 13 18},20 32 33 35 reported SAEs (online supplemental figure S2).

NMA results for efficacy

Regarding the primary efficacy outcome, P20 showed a clear benefit over the control, with a significantly higher rate of culture conversion at 8 weeks on solid culture (RR=1.09, 95% CrI 1.03 to 1.17) (figure 3A). CINeMA assessment rated the quality as low (online supplemental table S3). No statistically significant difference was observed between any two HDR regimens in the network for the primary efficacy outcome (figure 4).

Figure 3. Forest plots of relative effects. Each horizontal line on the forest plot represents the pooled risk ratio of different doses of rifamycin treatment (compared with the control group), with the risk ratio plotted as a circle and the 95% credible interval plotted as the line. (A–C) Network meta-analysis for efficacy (negative culture conversion at 8 weeks in solid culture, negative culture conversion at 8 weeks in liquid culture and treatment success). (D–F) Network meta-analysis for safety (serious adverse event, grade 3 or higher adverse event and grade 3 or higher adverse drug reactions).

Figure 4. League table of outcome analysis. Comparison of the included interventions: risk ratio (95% credible interval). Each box represents the comparison of the column-defining treatment versus the row-defining treatment. Bold indicates statistical significance. NA, not applicable.

A similar trend was observed in liquid culture results, where P20 showed a potential increase with the higher culture conversion rate at 8 weeks (RR=1.2, 95% CrI 1.0 to 1.6) (figure 3B). In terms of other secondary efficacy outcomes, compared with the control, individuals receiving R35 had a significantly lower risk of treatment success (RR=0.94, 95% CrI 0.88 to 0.99) (figure 3C). There was no statistical difference among all regimens in the risk of relapse (online supplemental figure S3). All HDR regimens had no statistical association with a higher rate of culture conversion at 12 weeks (online supplemental figure S3).

Figure 5 and online supplemental table S4 present the relative ranking of all rifamycin dosing regimens based on their SUCRA scores. The SUCRA scores were consistent with the relative effect results, indicating that P20 was the most effective regimen with bactericidal activity at 8 weeks in solid culture. No statistically significant evidence of inconsistency was reported in the node splitting test for negative culture conversion at 8 weeks in solid culture (online supplemental figure S4).

Figure 5. Ranking probabilities of all regimens for (A) negative culture conversion at 8 weeks in solid culture and (B) serious adverse event. The x-axis represents the rank position, and the y-axis indicates the probability of each treatment occupying each rank. Each coloured bar denotes a different rifamycin regimen.

NMA results for safety

We found no statistical differences between all regimens for the incidences of SAEs (figure 3D). CINeMA assessment rated the quality as very low (online supplemental table S3). No statistically significant evidence of inconsistency was reported in the node splitting test for SAEs (online supplemental figure S5). Regarding secondary safety outcomes, compared with the control, all HDR regimens also showed no significant difference in the incidence of grade 3 or higher AEs, grade 3 or higher ADRs, AEs, grade 3 or higher increase of transaminases, death and trial withdrawal (figure 3E–F, online supplemental figure S3). R15 was associated with a higher incidence of treatment discontinuation due to toxicity than the control (online supplemental figure S3).

Furthermore, no statistically significant differences in primary safety outcomes were observed across all regimens (figure 4). Relative ranking of all rifamycin dosing regimens based on their SUCRA scores is shown in figure 5 and online supplemental table S4. According to the SUCRA scores, P15 was ranked as the safest regimen for SAE, followed by R15 (figure 5). Besides, SUCRA rankings for secondary safety outcomes further indicated that the lowest rates of AEs were observed in different regimens across specific endpoints. P15 had the lowest incidence of grade ≥3 AEs and trial withdrawals; P10 was associated with the lowest rate of treatment discontinuation due to toxicity. The control group showed the lowest risk of grade ≥3 ADRs and mortality, while R20 had the lowest overall AE rate. For grade ≥3 transaminase elevations, P20 ranked highest in safety (online supplemental table S4).

Risk of bias and sensitivity analysis

All the 15 included trials (100%) were free of high risk of bias. The majority of included studies had a low risk of bias; only one trial was evaluated as ‘some concerns’.³⁹ The risk of bias assessment of included studies was provided in online supplemental figure S6. The sensitivity analysis excluding the trial of ‘some concerns’ confirmed the solidity of the results of the main analysis (onlinesupplemental figures 7 8). A leave-one-out sensitivity analysis of all study outcomes did not identify major sources of confounding for culture conversion rate at 8 weeks in solid culture and SAEs (onlinesupplemental figures 9 10).

Discussion

In this systematic review and NMA, we evaluated the efficacy and safety of different rifamycin dosing regimens for DS-TB. Our findings indicated that, compared with the control, P20 was associated with a higher 8 weeks culture conversion rate in solid culture. In liquid culture, the effect was modest (RR = 1.20, 95% CrI 1.00 to 1.60), suggesting a potential benefit while acknowledging the borderline statistical significance. However, no significant differences were observed in culture conversion at 12 weeks in either solid or liquid culture, nor in relapse rates across different groups. Additionally, R35 was linked to a lower rate of treatment success. Regarding safety, R15 was associated with a higher risk of treatment discontinuation from trial for toxicity compared with the control, but this finding exhibited substantial heterogeneity. No significant differences were observed in SAEs, grade 3 or higher AEs, grade 3 or higher ADRs, AEs, grade 3 or higher increase of transaminases, death or trial withdrawal.

For 8 weeks culture conversion rate in solid media, P20 demonstrated a significant improvement over the control and was ranked as the most effective intervention. Interestingly, its benefit in liquid media was less pronounced, with only borderline statistical significance. Based on SUCRA values, R35 ranked highest in liquid culture, followed by P20, suggesting variability in performance across different media types. This discrepancy may be attributed to the limited sample size of RCTs reporting outcomes in liquid media. Prior research⁶ demonstrated that the liquid media facilitated the recovery of Mycobacterium tuberculosis from sputum more effectively than solid media, potentially explaining the observed differences. Onorato et al, focusing exclusively on higher doses of rifampicin, reported an increased sputum culture conversion rate at 8 weeks with high-dose rifampicin (≥20 mg/kg) in solid media.¹⁶ However, a meta-analysis did not find significant differences in 2-month culture conversion rates in liquid culture between HDR and the control.¹⁷ Our study, incorporating more recent evidence and clearly distinguishing between solid and liquid media, offered a more comprehensive and nuanced analysis of treatment responses, indicating the trend toward stronger anti-TB activity with higher-dose regimens, particularly P20 and R35.

The reliability of 8-week sputum culture conversion as a surrogate marker for end-of-treatment outcomes remains debatable.⁴¹ Studies⁴²,⁴⁴ investigating 4-month TB treatment regimens showed that higher culture conversion rates at 2 months did not necessarily correlate with better long-term outcomes. Recently, high-quality phase III clinical trials have emerged with long-term treatment outcomes,¹⁸,²⁰ allowing for a more robust analysis of treatment success and relapse rates. This study found that R35 was associated with a lower treatment success rate, and no significant differences in relapse were observed across different dosing regimens.

This discrepancy highlights the clinical and conceptual complexity of using early bacteriological response as a surrogate for durable treatment outcomes. The lower rate of treatment success R35 might be linked with the non-inferiority trial comparing the short-term R35 regimen for 8 weeks and the standardised regimen for 6 months.¹⁹ Moreover, potential confounding factors—such as background drug regimens, treatment adherence and drug quality—may also have influenced treatment outcomes.⁴⁵ This suggested that while higher doses of rifamycin might enhance early bactericidal activity, their long-term outcomes in short course therapy required further investigation through prospective, multicentre studies.

Safety considerations are paramount when initiating anti-TB therapy.⁴⁶,⁴⁸ Previous meta-analyses have reported that compared with the control, HDR regimens showed no significantly higher rate of SAE, mortality, treatment discontinuation from trial for toxicity or trial withdrawal.^{16 17} Our findings are consistent with these observations, with the exception of R15, which had a higher rate of treatment discontinuation for toxicity than the control. However, the result showed high heterogeneity. Moreover, this study provided a comprehensive analysis of other safety outcomes, including grade 3 or higher ADRs, grade 3 or higher AEs, grade 3 or higher increase of transaminases and AEs. No significant differences were observed in these outcomes. Lower-dose rifamycin generally showed higher SUCRA scores for safety. Overall, HDR appeared to be safe compared with standard-dose rifampin, but the application of HDR regimen should still be accompanied by careful monitoring for ADRs. Multicentre, prospective studies are still needed to further explore the tolerability and safety of HDR therapy in the future.

Taking both safety and efficacy into account, P20 emerged as the most likely optimal regimen based on SUCRA results. In May 2022, WHO consolidated TB treatment guidelines recommended the combination of 1200 mg rifapentine (approximately 20 mg/kg), isoniazid, pyrazinamide and moxifloxacin,⁴⁹ which has risen considerable attention worldwide as this highlighted the potential of this regimen to shorten treatment duration without compromising efficacy or safety. This aligned with the results of this NMA.

Our study held several notable strengths. To our knowledge, it is the first NMA study to evaluate the efficacy and safety of various rifamycin dosing regimens for DS-TB, providing valuable evidence for clinicians in the absence of direct comparisons from RCTs. Besides, the high quality of the included trials and the rigorous analysis methods used to assess bias enhanced the credibility of our findings.⁹ Additionally, our approach allowed for simultaneous estimation of the relative effects of interventions within the evidence network and provided a ranking score of interventions. Furthermore, our separate analyses of culture conversion rates in solid and liquid media offered a more detailed exploration of optimal rifamycin dosing strategies.

However, the results of this NMA should be interpreted with caution due to several limitations. First, some included articles were dose-ranging trials and open-label RCTs, which may introduce bias. Second, because of the limited number of studies involving people living with HIV, individuals with low body weight, or those with diabetes mellitus, we were unable to conduct subgroup analyses. Furthermore, variability in the definitions of safety and efficacy endpoints across studies may also affect the applicability of our results to diverse clinical settings. In addition, no study examined high-dose rifabutin, thereby limiting the conclusion we can draw about it.

From a public health perspective, the adoption of HDR regimens—if proven efficacious—may necessitate the reformulation of currently available fixed-dose combination (FDC) tablets, which are designed for standard dosing. This represents a significant programme challenge, especially in resource-limited settings where FDCs are integral to treatment adherence and supply chain efficiency. Moreover, since current standard regimens already achieve high cure rates in most populations, future implementation of HDR regimens should be considered selectively—for example, in patients with delayed response, high bacillary burden or in special populations. Although these operational implications are beyond the primary scope of our meta-analysis, they highlight the need for further clinical trials and health policy assessments before widescale adoption. Ultimately, this study contributes updated comparative evidence on the efficacy and safety of rifamycin dosing strategies and supports future efforts to optimise treatment outcomes while preserving feasibility and accessibility.

Conclusions

Our study suggested that when all available rifamycin dosing regimens were considered, P20 was the most likely promising regimen for the strongest anti-TB activity. Compared with standard-dose rifampin, all HDR regimens showed similar safety profiles and did not result in a significant increase in the incidence of SAEs. This comparative information provided valuable insights for clinicians and researchers, while they were evaluating the relative efficacy and safety of different HDR regimens.

Data availability statement

Data are available on reasonable request.