Four Months of Rifampin or Nine Months of Isoniazid for Latent Tuberculosis in Adults

Trial Design and Objectives

We conducted an open-label, parallel-group, randomized, controlled trial comparing a 4-month regimen of rifampin with a 9-month regimen of isoniazid for the treatment of latent tuberculosis infection in adults. The primary objective was to compare the rates of confirmed active tuberculosis in the two groups among all eligible patients during 28 months after randomization (modified intention-to-treat analysis, defined as the study population remaining after valid and prespecified exclusions after randomization) (see the protocol, available with the full text of this article at NEJM.org). The secondary objectives were to compare the following in the two groups: the rate of confirmed active tuberculosis plus clinically diagnosed active tuberculosis per 100 person-years; the rate of confirmed or clinically diagnosed tuberculosis per 100 person-years among patients who completed the trial therapy per the protocol; the cumulative incidence of adverse events of grades 3 to 5, overall and those considered by the adjudication panel to be drug-related and occurring throughout the course of therapy or within the maximum time allowed for the completion of the rifampin regimen (120% of 4 months, or 146 days); the percentage of patients in each trial group who completed the trial therapy, which was defined as receipt of at least 80% of the doses; and the rate of drug-resistant active tuberculosis per 100 person-years.

Interventions

The control regimen was oral isoniazid at a dose of 5 mg per kilogram of body weight (maximum dose, 300 mg) taken daily for 9 months (270 doses), with vitamin B₆ (pyridoxine) added for adults who were at risk for neuropathy.^6-8 The experimental regimen was oral rifampin at a dose of 10 mg per kilogram (maximum dose, 600 mg) taken daily for 4 months (120 doses). The trial drugs were purchased at full cost, unless they were provided free of charge by the local tuberculosis program; no sponsor paid for drugs, in part or in whole.

Population Eligibility and Randomization

The selected trial sites in Australia, Benin, Brazil, Canada, Ghana, Guinea, Indonesia, Saudi Arabia, and South Korea had extensive experience in previous clinical trials. Trial staff at all the sites received initial training in Good Clinical Practice and trial procedures, and monitoring visits were conducted twice per year (see the Supplementary Appendix, available at NEJM.org).

Adults (18 years of age or older) were enrolled if they had a documented positive tuberculin skin test or interferon-γ–release assay, if they met the criteria for an increased risk of reactivation to active tuberculosis (see the Supplementary Appendix),^6,8 and if their provider recommended treatment with isoniazid. Before randomization, adults underwent medical evaluation, including radiography of the chest, to rule out active tuberculosis. Testing for the human immunodeficiency virus (HIV) was offered to participants who had risk factors for HIV infection. The exclusion criteria were exposure to a patient with active tuberculosis whose isolates were resistant to either trial drug, current or planned pregnancy, the use of medications with potentially serious interactions with either trial drug, history of allergy to either trial drug, or current active tuberculosis. All the eligible patents provided written informed consent.

Randomization was generated centrally, by computer, in blocks of varying length (2 to 8) and stratified according to center with an assignment ratio of 1:1. All the contacts within the same household were assigned to the same trial group if they were all identified within the same week.

Ascertainment of Active Tuberculosis

Follow-up during treatment occurred monthly for the first 2 months and at a minimum of every 8 weeks thereafter. Follow-up after treatment began after the participant completed or discontinued treatment and was conducted by telephone, visits at the health facility, or home visits every 3 months until 28 months after randomization. Hence, the minimum number of total trial-related visits was 11 for participants who had been randomly assigned to the 4-month rifampin regimen and 12 for those who had been randomly assigned to the 9-month isoniazid regimen. Suspected active tuberculosis was investigated according to a standard protocol. The names of all the trial participants who did not complete 28 months of follow-up or who had confirmed or clinically diagnosed active tuberculosis, plus the names of a random 10% sample of participants who completed follow-up, were sent to the local tuberculosis program to ascertain whether any of these participants had been reported as having received a diagnosis of active tuberculosis.

Confirmed tuberculosis was defined as a positive culture for Mycobacterium tuberculosis or a finding of caseating granulomas in a biopsy specimen obtained from any site. If there were no culture results, a positive acid-fast smear or a positive nucleic acid amplification test for M. tuberculosis complex was considered to be confirmatory. All clinical, radiologic, and microbiologic information of the participants with suspected active tuberculosis, including the response to treatment for active tuberculosis, was reviewed by a panel of three physicians who had experience regarding tuberculosis. The panel members were unaware of the trial-group assignments and the evaluations of the care providers, and each physician independently judged whether active tuberculosis was probable or unlikely. Clinically diagnosed tuberculosis was defined as a judgment of probable active tuberculosis by at least two of the three physicians.

Measurement of Treatment Completion

Patients brought all the remaining doses of the trial drug to every visit for pill counts. Treatment completion was defined as receipt of at least 80% of the doses within 12 months for rifampin or within 18 months for isoniazid. This threshold was selected because of evidence that efficacy was greatest if at least 80% of the doses of isoniazid were taken¹⁹ and evidence that the total number of doses taken is the key determinant in the prevention of tuberculosis.²⁰ Treatment completion per protocol was defined as receipt of at least 80% of the recommended doses within the allowed time (146 days for the rifampin regimen and 324 days for the isoniazid regimen).

Measurement of Safety during Treatment

At each follow-up visit, participants were questioned about and examined for adverse events. Suspected adverse events were investigated, managed, and reported according to standardized protocols (see Part 2 in the Supplementary Appendix). Information about suspected adverse events was reviewed by a different three-member panel who had expertise in clinical–epidemiologic tuberculosis research; these panel members were also unaware of the trial-group assignments and the judgments of the care providers. Panel members independently judged the severity of adverse events according to published recommendations (for hepatic events²¹ and for all other events²²) and categorized the events as follows: an adverse event that was not related to a trial drug; an adverse event of grade 1 or 2 that was related to a trial drug (not serious); an adverse event of grade 3 or 4 that was related to a trial drug (generally considered to lead to trial-drug discontinuation if related to a trial drug); or a grade 5 event (death) that was related to a trial drug. If opinions differed, the majority opinion was used; if all three opinions differed, the panel members rereviewed the blinded data.

Oversight

This trial was approved by the Biomedical Clinical Research Ethics Board of the McGill University Health Centre Research Institute and by the responsible ethics review committee at each participating site. All the authors vouch for the accuracy and completeness of the data and analyses presented and for the fidelity of the trial to the protocol.

Statistical Analysis

We originally planned that a sample of 3283 participants in each group would provide the trial with at least 80% power (at an alpha level of 0.05) to detect significantly superior effectiveness of the rifampin regimen as compared with the isoniazid regimen in preventing tuberculosis. For the assessment of tuberculosis prevention, we planned to include the 847 participants from the earlier phase 2 trial because they had been enrolled, undergone randomization, and had been treated and followed according to the same methods. (For details, see the articles by Menzies et al.¹⁷ and Aspler et al.¹⁸) To account for a potential 15% loss to follow-up (on the basis of the results in the phase 2 trial) and clustering within households, we increased enrollment to 6800 participants.

We also prespecified that this number of participants would provide the trial with more than 90% power to declare noninferior efficacy of the 4-month rifampin regimen, with a maximum tolerated difference in cumulative incidence between the two regimens of 0.75 percentage points. This value was based on an expected cumulative incidence of 3% over a 28-month period among untreated close contacts or other high-risk persons,^7,23-25 a 90% protective efficacy of the 9-month isoniazid regimen,⁹ and a minimum acceptable efficacy of the 4-month rifampin regimen of 65% (which had been achieved with the 6-month isoniazid regimen in previous trials¹⁹ and was widely considered to be sufficiently acceptable for use^6,8).

Statistical analyses were directed by the principal investigator and co-investigators, who remained unaware of the trial-group assignments until the analyses had been completed. Differences between groups were tested for significance by Student’s t-test (if normally distributed) or a Wilcoxon two-sample test (if not normally distributed) or by chi-square analysis for categorical variables. Rates of active tuberculosis were based on person-time of follow-up; patients who were lost to follow-up contributed to person-time until the date of the last contact.

Rate differences and their 95% confidence intervals were estimated with the use of generalized estimating equations to account for clustering in families, on the basis of Poisson distribution, with a log link.²⁶ The differences in the treatment-completion rates and the risk differences for adverse events, with 95% confidence intervals, were calculated with a binomial distribution with an identity link, with the use of generalized estimating equations. If there were no events, risk differences were estimated with the use of the method of Newcombe.²⁷ When superiority was not found, noninferiority with regard to rates of active tuberculosis was assessed by comparing the upper limit of the confidence interval of the rate difference to the prespecified margin.^28,29

Interim analyses for safety were performed after 25%, 50%, and 75% of the participants had been enrolled, and the analyses were reviewed by the data and safety monitoring board. Full details of the trial design and analyses are provided in the protocol and the statistical analysis plan. All the analyses were performed with the use of SAS software, version 9.4 (SAS Institute).

Trial Participants

Figure 1. Figure 1. Adult Participants in Phase 2 and 3 Trials for Primary Outcome of Active Tuberculosis.

The number of patients from the phase 2 trial who completed the rifampin regimen per protocol was smaller than in the previous report¹⁷ because of the shorter allowed time in the current analysis. Among the patients in the isoniazid group who did not complete therapy per protocol, four died during treatment and are not included in this subgroup in the figure. In the two trial groups, the subgroups of patients who completed follow-up exclude patients in whom active tuberculosis developed. The deaths that are listed in the bottom row of boxes were deemed not to be related to tuberculosis.

The treatment-phase results in the 847 participants who were included in the earlier safety (phase 2) trial have been reported previously¹⁷ and are not repeated here. In those participants, only the post-treatment outcomes of active tuberculosis are reported here. In the phase 3 trial, we assessed 16,907 potential participants from October 2009 through December 2014, of whom 6063 underwent randomization (Figure 1). Of these, 37 participants were excluded after randomization because they were close contacts of patients with active tuberculosis that was caused by isolates resistant to isoniazid or rifampin and 14 withdrew consent, which left 6012 adults in the modified intention-to-treat analysis. Of these, 5744 participants (95.5%) completed 28 months of follow-up after treatment. Of the 847 participants who had been in the safety trial, 614 (72.5%) completed follow-up after treatment.

Table 1. Table 1. Characteristics of the Participants in the Phase 3 Trial (Modified Intention-to-Treat Population).

The demographic and clinical characteristics of the participants in the two groups of the phase 3 trial are shown in Table 1. More than 70% of the participants were close contacts of someone with tuberculosis, and 854 of the total 6012 participants (14.2%) lived with at least one other trial participant. Table S1 in the Supplementary Appendix presents the characteristics of the participants, according to trial group, in the phase 2 and 3 trials combined.

Table 2. Table 2. Completion of Treatment in the Phase 3 Trial (Modified Intention-to-Treat Population).*

The rate of treatment completion in the phase 3 trial was significantly higher with the 4-month rifampin regimen than with the 9-month isoniazid regimen (difference, 15.1 percentage points; 95% confidence interval [CI], 12.7 to 17.4) (Table 2). Table S2 in the Supplementary Appendix shows the results regarding treatment completion for the phase 2 and 3 trials combined.

The most common reason for noncompletion of the treatment regimen was the participant’s decision to stop taking the trial drug. There were significant differences in demographic and clinical characteristics between participants who completed therapy and those who did not, and there were also significant differences between participants who completed the full 28 months of follow-up and those who were lost to follow-up. Among the participants who did not complete therapy or follow-up, there were no significant differences in characteristics between the two trial groups. Details are provided in Tables S5 through S8 in the Supplementary Appendix.

Efficacy

Table 3. Table 3. Primary End Point of Occurrence of Active Tuberculosis among All Participants.

There were eight cases of confirmed active tuberculosis and nine cases of clinically diagnosed active tuberculosis during active follow-up in the phase 2 and 3 trials combined (Table 3). Seven of the eight patients with confirmed tuberculosis and five of the nine patients with clinically diagnosed tuberculosis had also been reported to the local tuberculosis authorities as having received a diagnosis of active tuberculosis, but no additional cases were detected by passive case-finding procedures.

Table 4. Table 4. Occurrence of Active Tuberculosis in the Phase 3 Trial.

The differences in rates between the rifampin group and the isoniazid group were as follows: for confirmed tuberculosis in the modified intention-to-treat population, the difference was less than 0.01 cases per 100 person-years (95% CI, в€’0.14 to 0.16), which is equivalent to a difference in the cumulative incidence over the 28-month period of less than 0.02 (95% CI, в€’0.33 to 0.37). The difference for confirmed or clinically diagnosed tuberculosis in the modified intention-to-treat population was less than 0.01 cases per 100 person-years (95% CI, в€’0.23 to 0.22, which is equivalent to в€’0.54 to 0.51 over the 28-month period); and for confirmed or clinically diagnosed tuberculosis in the per-protocol analysis, the difference was в€’0.02 cases per 100 person-years (95% CI, в€’0.30 to 0.26, which is equivalent to в€’0.70 to 0.61 over the 28-month period). Among the phase 3 trial participants who completed therapy per the protocol, the rate difference between the rifampin group and the isoniazid group for confirmed plus clinically diagnosed tuberculosis was в€’0.02 cases per 100 person-years (95% CI, в€’0.33 to 0.29, which is equivalent to в€’0.77 to 0.68 over the 28-month period) (Table 4).

In all these analyses, the upper boundary of the 95% confidence interval for the difference in the rates of confirmed active tuberculosis or of confirmed or clinically diagnosed tuberculosis was less than the prespecified margin for noninferiority. However, the rifampin regimen was not superior to the isoniazid regimen.

Drug Resistance

Among the eight participants with confirmed active tuberculosis, drug-susceptibility test results were not available for four (three cases were diagnosed on the basis of histologic evidence of necrotizing granulomas, and cultures were contaminated in the fourth), two had susceptibility to all drugs tested, and isolates obtained from two participants showed drug resistance. One participant received a diagnosis of isoniazid-resistant active tuberculosis 8 weeks after starting isoniazid, and the other received a diagnosis of active tuberculosis less than 2 months after the completion of the 4-month rifampin regimen. The isolate that was obtained from this patient showed resistance to rifampin, according to Xpert MTB/RIF testing, although the isolate was pansusceptible on traditional phenotypic testing. These two participants were both household contacts of participants with new cases of active tuberculosis who had been enrolled at sites where the diagnosis of index cases was based on smear microscopy alone, so drug-susceptibility results were not available for either index case. All the participants with active tuberculosis, including the two with drug-resistant isolates, were treated successfully and remained free from disease according to clinical and imaging assessments for at least 1 year after the end of treatment.

Safety

Table 5. Table 5. Adverse Events in the Phase 3 Trial (Safety Population).

To account for the potential problem of differential ascertainment, owing to the fact that the duration of the isoniazid regimen was longer than the duration of the rifampin regimen, we estimated rate differences for adverse events that occurred during the first 146 days after randomization. The rifampin group had significantly lower rates of adverse events of grades 3 to 5 than the isoniazid group in analyses that included all such adverse events (rate difference, в€’1.1 percentage points; 95% CI, в€’1.9 to в€’0.4) and in analyses that included only adverse events that were considered by the independent panel to be related to the trial drug (в€’1.0 percentage point; 95% CI, в€’1.6 to в€’0.4) (Table 5).

Drug-induced hepatitis was the most common adverse event of grade 3 or 4 overall and was significantly less frequent in the rifampin group than in the isoniazid group in analyses that included all such events, that included only events that were adjudicated as being possibly or probably related to the trial drug, and that included only events occurring in the first 146 days. Table S3 in the Supplementary Appendix shows the results regarding total adverse events in the phase 2 and 3 trials combined, and Table S4 in the Supplementary Appendix shows the results for other types of adverse events.

In this trial involving more than 6800 adults in nine countries, a 4-month regimen of rifampin was not inferior to a 9-month regimen of isoniazid in preventing active tuberculosis; however, the rifampin regimen was also not superior to the isoniazid regimen. The rate of treatment completion was higher in the rifampin group than in the isoniazid group. In the rifampin group, there was a significantly lower incidence of adverse events of grades 3 to 5, particularly hepatotoxic adverse events, than in the isoniazid group in analyses that included all such adverse events, that included only events that were adjudicated as being drug-related, and that included only events that occurred in the first 146 days after randomization.

This trial had a number of strengths. Selection bias should have been minimized by randomization and by the absence of differences in the demographic and clinical characteristics between the two trial groups among the participants who did not complete therapy and among those did not complete follow-up. Of the 6859 participants who were included in the modified intention-to-treat analysis from the phase 2 and 3 trials, only 7.9% did not complete 28 months of follow-up, although the rate of loss to follow-up was higher in the phase 2 trial than in the phase 3 trial. This result should have enhanced our likelihood of detecting active tuberculosis, which was further enhanced by the passive case-finding strategy to detect active tuberculosis among participants who had been lost to follow-up. The large sample size provided adequate power to detect clinically meaningful differences between the two regimens. The trial sites were in settings that had widely varying levels of resources, which may enhance the generalizability of the results. Active tuberculosis and adverse events were assessed according to detailed protocols for investigation and management and were adjudicated in a blinded fashion by independent panels. The trial drugs were administered by the participants themselves daily, so treatment completion and the detection of adverse events are presumed to be less affected by the mode of administration than would be the case in trials that compared participant-administered isoniazid with administration of isoniazid–rifapentine that was directly observed by health personnel.^30-33

This trial has a number of important limitations. The open-label design may have introduced bias in the ascertainment of treatment completion or adverse events; however, this approach has been used previously in trials of rifamycin-based regimens^30-34 to take advantage of their shorter duration. To safeguard against bias, all the final diagnoses of adverse events and active tuberculosis were made by independent review panels whose members were unaware of the trial-group assignments. Treatment completion was based on the assessments of the patients and providers in addition to pill counts, which are not highly reliable methods.^35,36 In total, only 255 participants with HIV infection were enrolled in either phase 2 or phase 3 of the trial, which reduces the applicability of our findings to this population.

A further limitation was the low event rate of active tuberculosis in each group in the modified intention-to-treat analysis, as has been seen in other trials.^30,31 This makes the conclusions less robust. However, given the observed rate of confirmed or clinically diagnosed tuberculosis of 0.11 cases per 100 person-years among participants who completed the 9-month isoniazid regimen, and assuming 90% efficacy of this regimen,⁹ we would expect a rate of 1.2 cases per 100 person-years in the untreated population, which, if extrapolated over the 28-month duration of the trial, would be equivalent to a cumulative incidence of 2.8% — close to the anticipated cumulative incidence of 3%. Rates of disease were lower than expected among participants who did not complete treatment, in part because the completion rate was higher than expected in the two trial groups and also because the participants who did not complete treatment received the therapy for a median of 3 months in the isoniazid group and this duration of isoniazid therapy is known to have modest efficacy.¹⁹ Another factor may have been the very small number of HIV-infected persons — a population that in other trials involving participants with latent tuberculosis infection had much higher rates of active disease.^32,34,37,38

This trial adds to the mounting evidence of benefits of rifamycin-containing regimens of 3 or 4 months’ duration. Numerous other observational and experimental studies have shown significantly higher rates of treatment completion with the shorter rifamycin-based regimens than with 9 months of isoniazid therapy^{12-14,16,17,31} and efficacy that is at least equivalent.^{8,31,32,39,40} The findings of this trial also corroborate evidence from other trials^17,31 and observational studies^12,13 that rifamycin-based regimens are associated with substantially lower rates of hepatotoxic effects — a major limitation of isoniazid²¹ — and with lower overall rates of adverse events of grade 3 or 4.^12,13,17 However, all these earlier studies may have overestimated the relative safety of 4 months of rifampin owing to ascertainment bias.

In conclusion, a 4-month regimen of rifampin was not inferior to a 9-month regimen of isoniazid for the prevention of active tuberculosis. The rifampin group had a significantly higher rate of treatment completion and fewer drug-related adverse events of grades 3 to 5.