Annals of Internal Medicine
Treatment of Latent Tuberculosis Infection
An Updated Network Meta-analysis
Dominik Zenner, MD; Netta Beer, PhD; Ross J. Harris, MSc; Marc C. Lipman, MD; Helen R. Stagg, PhD*; and
Marieke J. van der Werf, MD, PhD*
Background: Treatment of latent tuberculosis infection (LTBI) is
an important component of tuberculosis (TB) control, and this
study updates a previous network meta-analysis of the best LTBI
treatment options to inform public health action and program-
matic management of LTBI.
Purpose: To evaluate the comparative efcacy and harms of
LTBI treatment regimens aimed at preventing active TB among
adults and children.
Data Sources: PubMed, Embase, and Web of Science from in-
dexing to 8 May 2017; clinical trial registries; and conference
abstracts. No language restrictions were applied.
Study Selection: Randomized controlled trials that evaluated
human LTBI treatments and recorded at least 1 of 2 prespecied
end points (hepatotoxicity and prevention of active TB).
Data Extraction: 2 investigators independently extracted data
from eligible studies and assessed study quality according to a
standard protocol.
Data Synthesis: The network meta-analysis of 8 new and 53
previously included studies showed that isoniazid regimens of 6
months (odds ratio [OR], 0.65 [95% credible interval {CrI}, 0.50
to 0.83]) or 12 to 72 months (OR, 0.50 [CrI, 0.41 to 0.62]),
T uberculosis (TB) is a global priority infectious dis-
ease that caused an estimated 1.4 million deaths in
2015 (1). Many strategies are required to reach the
United Nations' Sustainable Development Goal of end-
ing the global TB epidemic by 2030 (2); tackling latent
TB infection (LTBI), including providing preventive
treatment to persons at high risk for TB, is a key action
in achieving both the Sustainable Development Goal
and the targets of the World Health Organization's End
TB Strategy (3).
Many treatment regimens for LTBI are available
globally; 5 are recommended by the World Health Or-
ganization (4), and 4 of these by the Centers for Dis-
ease Control and Prevention (5). Evidence on the ef-
cacy of shorter regimens with a reduced pill burden is
evolving, but more information on effectiveness is still
urgently needed (6).
We previously published a network meta-analysis
(NMA) of randomized controlled trials (RCTs) that
identied the most effective and least harmful preven-
tive treatment regimens (7). The results of this meta-
analysis and Bayesian network analysis served as the
evidence base for the World Health Organization's
2014 LTBI guidelines (4, 8). As part of the European
Centre for Disease Prevention and Control's decision to
provide new guidance on programmatic LTBI control in
the European Union/European Economic Area and
candidate countries, we identied the need to update
Annals.org
Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017
REVIEW
rifampicin-only regimens (OR, 0.41 [CrI, 0.19 to 0.85]), rifampicin-
isoniazid regimens of 3 to 4 months (OR, 0.53 [CrI, 0.36 to 0.78]),
rifampicin-isoniazid-pyrazinamide regimens (OR, 0.35 [CrI, 0.19
to 0.61]), and rifampicin-pyrazinamide regimens (OR, 0.53 [CrI,
0.33 to 0.84]) were efcacious compared with placebo. Evidence
existed for efcacy of weekly rifapentine-isoniazid regimens
compared with no treatment (OR, 0.36 [CrI, 0.18 to 0.73]). No
conclusive evidence showed that HIV status altered treatment
efcacy.
Limitation: Evidence was sparse for many comparisons and
hepatotoxicity outcomes, and risk of bias was high or unknown
for many studies.
Conclusion: Evidence exists for the efcacy and safety of
6-month isoniazid monotherapy, rifampicin monotherapy, and
combination therapies with 3 to 4 months of isoniazid and
rifampicin.
Primary Funding Source: U.K. National Institute for Health Re-
search. (PROSPERO: CRD42016037871)
Ann Intern Med. doi:10.7326/M17-0609 Annals.org
For author afliations, see end of text.
This article was published at Annals.org on 1 August 2017.
* Drs. Stagg and van der Werf share joint last authorship.
our 2014 review (9). The results are presented in this
report.
METHODS
To ensure consistency, we used the same method-
ology as in our previous study (7). We summarize this
approach below. We registered this study in the
PROSPERO database (CRD42016037871).
Data Sources and Searches
PubMed, Embase, Web of Science, and gray litera-
ture were searched until 8 May 2017 with no language
restrictions. We reviewed and included our previous
search and added all relevant articles, focusing on the
time after the previous search (Supplement, available
at Annals.org).
Study Selection and Extraction
We used the same selection criteria as before and
included RCTs that analyzed LTBI treatments. We
excluded nonrandomized, observational, and animal
See also:
Web-Only
Supplement
CME/MOC activity
Annals of Internal Medicine 1
 REVIEW Network Meta-analysis of LTBI Treatments

studies and those that had insufcient information on ow charts, for which we used PowerPoint for Mac
one or both of our main end points (hepatotoxicity and (Microsoft).
development of active TB). As in our previous study, 2
Role of the Funding Source
of 5 potential reviewers (D.Z., N.B., H.R.S., M.C.L., and
The U.K. National Institute for Health Research had
M.v.d.W.) independently screened titles, abstracts, and
no inuence or role in this review.
full-text articles and resolved discrepancies by consen-
sus or, if needed, by consultation with a third reviewer.
Data Extraction and Quality Assessment RESULTS
Two reviewers (D.Z. and N.B.) independently ex- We identied 1576 articles in the database search,
tracted data to a standardized template, which in- and of these, 1434 were left after deduplication (Sup-
cluded details of the study, population, follow-up, treat- plement Figure 1; all supplement tables and gures are
ment regimens, and outcomes of interest. Study quality available at Annals.org). Eight articles were selected
was assessed by the same 2 reviewers using Higgins and used to update or add to the 53 studies included in
and colleagues' quality assessment tool (10). Disagree- the original review (Supplement Table 1). Two of the
ments were resolved in consultation with a third re- added reports provided data from studies that had
viewer (M.v.d.W.). been included in the original review (12, 13). Saman-
dari and colleagues' 2015 article (12) overlapped with
Data Synthesis and Analysis
their 2011 article (14); the former was used to provide
We compared both of the main outcomes, active
an update for the outcome of development to active TB
TB prevention and hepatotoxicity, using conventional
but did not report any additional hepatotoxicity results.
random-effects meta-analysis and an NMA approach.
Sterling and associates' 2016 article (13) provided an
Analyses included grouping and stratifying treatment
update solely for the persons living with HIV included in
regimens and estimating differences in treatment ef-
their original 2011 article (15). In the interest of analytic
fects between groups using ratios of odds ratios (RORs)
power, we thus retained the earlier study for the overall
according to study-level variables, as described by
analysis but used the later study for the HIV-stratied
Stagg and colleagues (Supplement) (7). Predened
analysis. Danel and colleagues (16) included 4 groups,
stratied analyses required study populations to be
2 comparing patients with deferred antiretroviral ther-
grouped by HIV status (HIV-positive, nonHIV-positive
apy (ART) and 2 comparing patients with early ART ini-
[including low-proportion HIV-positive, dened as
tiation. We extracted and analyzed data separately for
5%], or not stated), age (adults vs. children), immuno-
both groups. Two additional trials were initially consid-
suppression status (as a result of HIV infection or other
ered but then excluded from our analysis. Hosseinipour
conditions), and TB incidence when and where the RCT
and associates (17) compared preventive with active TB
was done (high vs. low). The robustness of the HIV strat-
treatments in an HIV-positive population (not compara-
ication was also tested by sensitivity analysis using a
ble to our eligible studies), and Torre-Cisneros and col-
stricter denition of nonHIV-positive status (only stud-
leagues (18) compared levooxacin with an isoniazid
ies where being HIV-positive was an exclusion criterion
(INH) preventive treatment regimen in transplant recip-
and those from before 1990, when endemic levels
ients. The trial was stopped early for toxicity events and
were low) (11).
did not yield interpretable efcacy or toxicity data due
Study and review quality were also analyzed as pre-
to heterogeneity among the groups, among other
viously. We assessed the effect of study quality do-
reasons.
mains (such as blinding, allocation concealment, ran-
Four of the new articles contained extractable data
domization, and outcome assessment) on estimated
on hepatotoxicity (12, 16, 19, 20), and all 8 on develop-
treatment efcacy by estimating RORs between studies
ment of active TB (12, 13, 16, 19 23). Of the 8 reports,
with adequate domains and those with inadequate or
6 were done in countries with high incidence of TB (12,
unclear domains. Publication bias was similarly as-
16, 19, 2123) and 6 were done solely in persons living
sessed to determine whether average SEs of log odds
with HIV (12, 13, 16, 19 21). All new studies included
ratios for each study were associated with a difference
INH monotherapy in at least 1 group; in addition, 1
in estimated treatment efcacy. We examined network
included INH-rifapentine (RPT) (13), 1 included INH-
consistency by informally comparing estimated treat-
ethambutol (EMB) (23), and 1 included both INH-
ment effects from NMA with those from standard pair-
rifampicin (RMP) and INH-pyrazinamide (PZA) (20)
wise meta-analysis and by comparing deviance infor-
(Supplement Table 2). Sixteen regimens were thus in-
mation criterion statistics for the NMA using an
cluded in the evidence network, 1 more (INH-EMB for
inconsistency model. In this model, the difference be-
12 months) than in the previous review (Table 1 and
tween comparisons is unconstrained, rather than the
Figures 1 and 2).
usual NMA assumption of consistent treatment effects
for different comparisons. Study Quality
We used STATA, version 13 (StataCorp), for data Many newly included studies had high or unknown
processing and classical analysis and WinBUGS, version risk of bias (Supplement Table 3). However, as previ-
1.4.3 (Medical Research Council Biostatistics Unit of ously, all indicators for high risk of bias were associated
the University of Cambridge), for the Bayesian NMA. with only weak evidence for modication of treatment
Most gures were also produced with STATA, except efcacy; for instance, inadequate or unclear allocation
2 Annals of Internal Medicine Annals.org

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017

 Network Meta-analysis of LTBI Treatments
concealment had an ROR of 0.74 (95% credible interval
[CrI], 0.31 to 1.42). Similar results were seen for high or
unclear risk of bias due to inadequate randomization
and blinding, which is consistent with other studies
(24). The ROR for incomplete or unclear versus ade-
quate outcome reporting was 0.91 (CrI, 0.42 to 1.43),
and for selective reporting, 1.03 (CrI, 0.55 to 1.75).
Results for hepatotoxicity were highly uncertain for
all domains due to the sparsity of data, and no rm
evidence enabled us to conrm or reject any modica-
tion of treatment effects according to study quality
domains.
Prevention of Active TB
An assessment of 16 regimens found that INH reg-
imens of 6 months or 12 to 72 months; RMP-only regi-
mens; RMP-INH regimens of 3 to 4 months; and RMP-
INH-PZA, RMP-PZA, and INH-EMB regimens of 12
months were efcacious (P < 0.05). Table 1 shows esti-
mated odds ratios (ORs) for each regimen versus pla-
cebo and no treatment under the NMA model. The
INH-EMB 12-month regimen showed the greatest ef-
fect but with substantial uncertainty (OR vs. placebo,
0.20 [CrI, 0.04 to 0.82]), and the shorter INH-EMB reg-
imen showed no evidence of efcacy. Two INH-
rifabutin (RFB) regimens, which used different doses of
RFB, showed ORs (vs. placebo) of 0.30 but with sub-
stantial uncertainty. The RMP-INH-PZA regimen showed
a two-thirds reduction in active TB and was strongly
statistically signicant (OR vs. placebo, 0.35 [CrI, 0.19 to
0.61]). In addition, the use of RMP or RMP-INH for 3 or
4 months seemed to be efcacious, with reductions in
active TB of half or more, although these results are
based on limited data. The ORs for all lengths of INH
regimens had overlapping CrIs and varying levels of
uncertainty, with the most robust evidence for 6 or
more than 12 months and the greatest efcacy for 12
months or more. Treatment rankings, although not pro-
viding quantied differences in treatment efcacy, are
also useful in understanding uncertainty in the evi-
dence base. For instance, the INH-EMB 12-month reg-
imen had the highest median rank, but its CrI (rank 1 to
Table 1. ORs and Treatment Rankings for the Prevention of Active TB, Derived From the Network Meta-analysis
Regimen OR vs. Placebo (95% CrI)
No treatment 1.62 (1.062.47)
Placebo 1.00 (reference)
INH 34 mo 0.93 (0.551.50)
INH 6 mo 0.65 (0.500.83)
INH 9 mo 0.75 (0.351.62)
INH 1272 mo 0.50 (0.410.62)
RFB-INH 0.30 (0.051.50)
RFB-INH (high) 0.30 (0.051.52)
RPT-INH 0.58 (0.301.12)
RMP 0.41 (0.190.85)
RMP-INH 1 mo 1.05 (0.372.77)
RMP-INH 34 mo 0.53 (0.360.78)
RMP-INH-PZA 0.35 (0.190.61)
RMP-PZA 0.53 (0.330.84)
INH-EMB 0.87 (0.322.36)
INH-EMB 12 mo 0.20 (0.040.82)
CrI = credible interval; EMB = ethambutol; INH = isoniazid; OR = odds ratio; PZA = pyrazinamide; RFB = rifabutin; RMP = rifampicin; RPT =
rifapentine; TB = tuberculosis.
Annals.org
Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017
REVIEW
11) also included some of the lowest ranks. The INH
regimen of 12 months or more had more certainty but
was ruled out of being one of the best treatments, and
several other regimens (RFB-INH, RMP, and RMP-INH-
PZA) had CrIs that included rank 1. Treatment rankings
are shown in Table 1, and histograms of rankings are in
Supplement Figure 2 and Supplement Table 4.
Stratifying the results on the basis of HIV status
showed no statistically signicant differences in effect
estimates for each regimen or the overall pattern of
results when studies in exclusively HIV-positive popula-
tions (24 studies) were compared with those in non
HIV-positive or unspecied populations (28 studies)
(Supplement Table 4). The estimated ROR for the over-
all difference in treatment efcacy (1.45 [CrI, 0.89 to
2.31]) indicates an inconclusive but potentially weaker
efcacy for treatment versus placebo or no treatment in
HIV-positive patients. The SD of the random effect for
treatment modication was 0.164 (CrI, 0.007 to 0.957),
indicating reasonable consistency across regimens for
differences in treatment efcacy, albeit with wide CrIs.
Inclusion of high versus low TB incidence in the
country of study as a covariate reduced between-study
variability; treatment was generally less efcacious in
high-incidence populations (ROR, 1.58 [CrI, 1.01 to
2.48]). Subgroup analysis by year and age also resulted
in little change in treatment rankings. Using covariate
models, we saw lower treatment effects in more recent
years, although the evidence for this was fairly weak
(ROR for 1992 to 2004 vs. pre-1992, 1.58 [CrI, 0.82 to
2.82], and for 2005 onward vs. pre-1992, 1.34 [CrI, 0.59
to 2.38]). No evidence indicated a relationship between
adherence and efcacy (ROR per 10% decrease, 0.94
[CrI, 0.54 to 1.30]). Restricting analysis to culture-
conrmed TB cases did not alter our conclusions.
Comparing results derived from a random-effects
pairwise meta-analysis with corresponding estimates
from the NMA model revealed some differences in OR
estimates between the models (Figure 2; Supplement
Table 5 and Supplement Figure 3). Many treatment
comparisons showed a stronger benecial effect in a
OR vs. No Treatment (95% CrI) Rank (95% CrI)
1.00 (reference) 16 (1416)
0.62 (0.410.94) 13 (1115)
0.57 (0.311.02) 13 (815)
0.40 (0.260.60) 10 (712)
0.46 (0.220.95) 11 (415)
0.31 (0.210.47) 6 (410)
0.18 (0.030.95) 3 (115)
0.19 (0.030.98) 3 (115)
0.36 (0.180.73) 8 (314)
0.25 (0.110.57) 5 (112)
0.65 (0.231.71) 14 (416)
0.33 (0.200.54) 7 (411)
0.21 (0.110.41) 3 (18)
0.33 (0.180.58) 7 (312)
0.54 (0.191.56) 12 (416)
0.12 (0.020.54) 2 (111)
Annals of Internal Medicine 3
 REVIEW Network Meta-analysis of LTBI Treatments

Figure 1. Treatment networks. standard pairwise meta-analysis than in an NMA; how-

ever, the differences had P values greater than 0.05
and were found predominantly where the effect esti-
INH- mates from pairwise meta-analysis were imprecise.
INH- EMB
4 INH EMB 12 mo RFB-INH
1272 6 mo Hepatotoxicity
INH 9 mo
mo
Because of the limited data on hepatotoxicity, we
INH 6
RFB-INH present results from the direct comparisons. Hepato-
(high)
INH
mo
2 2
toxicity results from the NMA are shown in Table 2.
6
34 Estimates were largely consistent with the direct com-
mo
RPT-INH parisons (Supplement Table 3, 4, and 6). Twenty pair-
3 3 wise comparisons were available. Results from the
3

2
standard meta-analysis suggest that the RMP-only and
9
7
RMP
34 mo
RPT-INH regimens had lower rates of hepatotoxicity
3
17
than an INH-only regimen of 6, 9, or 12 to 72 months
No 8
treatment RMP-
(Supplement Table 7 and Supplement Figure 3). RMP-
INH INH regimens also had lower hepatotoxicity than the
2 1 mo
Placebo
INH-only regimen, although evidence for this was good
2
2
RMP- only when compared with INH regimens of 12 to 72
2 4m
RMP- RMP- 2
INH months. We found good evidence that regimens con-
PZA INH- 34 mo
PZA taining PZA had higher hepatotoxicity compared with 6
months of INH or 12 weeks of RPT-INH. No data were
INH- available on the hepatotoxicity of the RFB-INH and INH-
INH-
3 EMB 6
EMB 12 RFB-INH EMB regimens. Stratifying the results on the basis of
INH mo
1272 mo immunosuppression, HIV status, and TB incidence did
INH 9 mo
mo not affect our conclusions.
RFB-INH
INH 6 (high)
mo 2 Inconsistency
2
INH 34
5 Comparing inconsistency models with consistency
mo
RPT-INH models found no evidence that the additional complex-
3
2 ity of the former was required (Supplement Figure 4).
However, because this study used a full random-effects
2 RMP
34 mo
model and the data exhibited a moderate level of
9 7
2 between-study heterogeneity, the power to detect in-
17 5
No consistency was low and this result cannot be inter-
treatment RMP-
INH preted as conclusive evidence of network consistency.
2 2
1 mo Results from 4 studies, including that of Ma and col-
Placebo
2
leagues (23), were extreme in comparison with other
RMP-
RMP- RMP-
2 INH studies, but these data points tted poorly in both the
2
PZA INH- 34 mo NMA and the inconsistency model, indicating extreme
PZA
heterogeneity rather than nontransitivity.
Publication Bias
2 INH- INH- RFB-INH
INH EMB 6 EMB 12 RFB-INH (high) In a pairwise meta-analysis, no evidence suggested
1272 mo mo
INH 9 mo
that treatment effects had been exaggerated in smaller
mo
studies (a proxy for publication bias) in any of the com-
INH 6 2
mo parisons (Harbord test P 0.300). In NMA, the covari-
INH 34 RPT-INH ate effect for study SE indicated that estimated treat-
mo
ment effects tended to show a greater benet in
3
3 smaller, less precise studies, although this result was
RMP
2 2
34 mo
not statistically signicant (ROR per 0.2-change in SE,
6 0.83 [CrI, 0.66 to 1.04]) (Supplement Figure 5).
2
4 RMP-
No 5
treatment INH
1 mo

Placebo RMP-
2 INH Unlabeled connections indicate 1 study reporting that treatment pair;
34 mo for 2 or more studies, lines are proportionally thicker and labeled with
RMP- RMP- the number of studies. Color coding indicates classes of treatment.
PZA INH-
PZA
Dotted boxes indicate a lack of data. RFB-INH: RFB, 300 mg, plus INH,
750 mg, twice weekly for 3 mo; RFB-INH high: RFB, 600 mg, plus INH,
750 mg, twice weekly for 3 mo. EMB = ethambutol; INH = isoniazid;
PZA = pyrazinamide; RFB = rifabutin; RMP = rifampicin; RPT = rifapen-
tine. Top. For all studies. Middle. For those with active tuberculosis
data. Bottom. For those with hepatotoxicity data.

4 Annals of Internal Medicine Annals.org

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017

 Network Meta-analysis of LTBI Treatments REVIEW

DISCUSSION study on the efcacy of INH-EMB combination therapy,

We present the results of an updated review and
meta-analysis investigating the efcacy and toxicity of
treatment regimens for LTBI (7). This review conrms
that all currently recommended regimens (4) are safe
and efcacious and provides more robust evidence to
demonstrate the efcacy of rifamycin-containing regi-
mens, including 3- to 4-month RMP monotherapy, to
prevent active TB. Although we also included a new
no TB cases were seen in the groups receiving INH or
INH-EMB, so this small study did not add statistically
signicant evidence to the network. We did not nd
statistically signicant differences in treatment efcacy
between regimens in HIV-positive and HIV-negative pa-
tients, although efcacy may have been weaker in HIV-
positive populations. Studies in HIV-positive popula-
tions have been done only recently and are therefore

Figure 2. Comparison of odds ratios for active tuberculosis obtained from random-effects pairwise meta-analysis with
corresponding estimates from NMA.

Regimen Odds Ratio

INH 12 mo
vs. INH-EMB
vs. RMP-INH 3 mo
vs. RMP-PZA
vs. RPT-INH
INH 4 mo
vs. INH 12 mo
vs. INH 6 mo
vs. RMP-INH 1 mo
vs. RMP-INH 3 mo
vs. RMP-INH-PZA
INH 6 mo
vs. INH 12 mo
vs. RFB-INH
vs. RFB-INH (high)
vs. RMP
vs. RMP-INH 3 mo
vs. RMP-INH-PZA
vs. RMP-PZA
vs. RPT-INH
INH 9 mo
vs. RMP-INH 3 mo
vs. RPT-INH
No treatment
vs. INH 12 mo
vs. INH 4 mo
vs. INH 6 mo
vs. INH 9 mo
vs. RMP-INH 1 mo
vs. RMP-INH 3 mo
vs. RMP-INH-PZA
vs. RMP-PZA
Placebo
vs. INH 12 mo
vs. INH 4 mo
vs. INH 6 mo
vs. INH-EMB 12 mo
vs. RMP
vs. RMP-INH 3 mo
vs. RMP-INH-PZA
vs. RMP-PZA
RMP
vs. RMP-INH 3 mo
vs. RMP-PZA
RMP-INH 1 mo
vs. RMP-INH 3 mo
vs. RMP-INH-PZA
RMP-INH 3 mo
vs. RMP-INH-PZA
vs. RMP-PZA
RPT-INH
vs. RMP-INH 3 mo
vs. RMP-PZA

0.02 0.05 0.1 0.2 0.5 1 2 5 10 20 50

Direct NMA

Where data on direct comparisons of treatment pairs are available for active tuberculosis, these may be pooled by means of standard meta-analysis
for each pair of treatments in turn. The resulting estimates are then compared with those obtained from the NMA analysis, which incorporates
indirect evidence and the overall network structure in addition to the direct evidence. EMB = ethambutol; INH = isoniazid; NMA = network
meta-analysis; PZA = pyrazinamide; RFB = rifabutin; RMP = rifampicin; RPT = rifapentine.

Annals.org Annals of Internal Medicine 5

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017

 REVIEW
Table 2. ORs and Treatment Rankings for Hepatotoxicity, Derived From the Network Meta-analysis
Regimen OR vs. Placebo (95% CrI)
No treatment 0.24 (0.060.75)
Placebo 1.00 (reference)
INH 6 mo 0.27 (0.100.60)
INH 9 mo 0.41 (0.081.62)
INH 1272 mo 0.66 (0.261.32)
RPT-INH 0.13 (0.030.42)
RMP 0.03 (<0.020.16)
RMP-INH 34 mo 0.17 (0.050.46)
RMP-INH-PZA 0.58 (0.073.72)
RMP-PZA 0.80 (0.252.17)
CrI = credible interval; INH = isoniazid; OR = odds ratio; PZA = pyrazinamide; RMP = rifampicin; RPT = rifapentine.
highly correlated with year of publication; more recent
studies were also associated with a modest decrease in
treatment efcacy, so some confounding may exist.
In addition to a regimen's efcacy, toxicity, and risk
for adverse events, 3 other key factors should be con-
sidered when making recommendations in clinical
practice: cost, length of treatment, and pill burden. A
recent systematic review showed that shorter regimens
were associated with higher treatment completion
rates (25). The relatively brief 3-month RPT-based regi-
men, with only 12 doses, has obvious advantages in
shortened treatment length. Publications focusing on a
general population (15), HIV-infected persons (13), and
children and adolescents (26) have demonstrated
good efcacy and safety. Unfortunately, because these
publications derive from the same core study, they add
limited evidence to our network.
Few other systematic reviews on this topic have
been published recently. Our search revealed 2 articles
in which LTBI treatment efcacy and toxicity were part
of a wider review (27, 28) and 3 that focused on specic
populations, 2 on children (29, 30) and 1 on HIV-
infected adults (31). The study collection in these re-
views was more restricted than ours, but the results of
all studies are compatible with ours.
Our search also revealed 2 ongoing registered tri-
als. One of these (NCT02651259), which aims to evalu-
ate the pharmacokinetics, acceptability, and safety of
the RPT-INH regimen in pregnant and postpartum
women, is expected to complete recruitment in De-
cember 2018 (32). The other (NCT02980016) is a prag-
matic trial that investigates incidence and treatment
completion rates in an HIV-positive population, com-
paring a single RPT-INH course with a 6-month INH reg-
imen and with a periodic RPT-INH regimen (33). This
trial is scheduled to complete recruitment in June
2019. Both trials look at important aspects of RPT-
based regimens and support the much-needed devel-
opment of shorter LTBI treatment regimens with less
frequent doses.
Our study presents 2 parallel forms of meta-
analysis. Effect estimates can sometimes show greater
benet in pairwise comparisons than in NMA, with the
latter constraining treatment effects to be consistent
across comparisons. Although the difference was not
statistically signicant, it is possible that treatment ef-
6 Annals of Internal Medicine
Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017
Network Meta-analysis of LTBI Treatments
OR vs. No Treatment (95% CrI) Rank (95% CrI)
1.00 (reference) 4 (27)
4.12 (1.3315.88) 9 (710)
1.10 (0.403.17) 5 (37)
1.70 (0.358.05) 6 (310)
2.72 (0.967.44) 8 (610)
0.52 (0.132.15) 2 (15)
0.14 (0.020.81) 1 (12)
0.72 (0.212.37) 3 (26)
2.41 (0.2520.02) 7 (210)
3.32 (0.9911.23) 9 (610)
fects reported in our pairwise results could be slightly
overestimated.
The main limitation of our work is the underlying
studies, many of which have a high risk of bias, are
small, or use nonstandard end point denitions. Many
publications reported on overlapping RCT populations;
we carefully avoided double counting by applying stan-
dard inclusion rules to these studies. In general, this
meant including the study with the largest cohort and
longest follow-up, which was often the most recent
work. Because LTBI reactivation risk is usually highest
shortly after infection, prolonging cohort follow-up time
may lead to decreased effect estimates, particularly for
long-term regimens. Our replacement of efcacy esti-
mates from Samandari and colleagues' 2011 article
(14) with those from their 2015 article (12) demon-
strated this; the efcacy of 36 months' INH treatment
waned over time. It is possible, therefore, that our anal-
ysis provides conservative efcacy estimates in some
instances, particularly for longer regimens and in set-
tings with a high TB burden where the potential for
reinfection is greater. We tested for publication bias;
however, our ability to do so was limited for some com-
parisons owing to sparse data. Lastly, although we pro-
vide stratied analyses for many covariates, including
HIV co-infection, little evidence exists on how concom-
itantly administered treatments, such as HIV ART, may
affect LTBI treatment efcacy. Although some recent
insights (34) have shown that beginning ART early
affects TB incidenceindependent of LTBI treatment
(16) data points were too limited to include ART as a
cofactor in our analysis.
In conclusion, despite limitations that include qual-
ity and reporting standards of the underlying studies,
the evidence for safety and efcacy of most standard
treatment regimens is robust (although evidence re-
mains sparse for the 9-month INH regimen). The evi-
dence for rifamycin-containing regimens, including
RPT, is improving. More evidence is needed, particu-
larly for RPT-based regimens; for the INH-RMP combi-
nation; for alternative treatments with a shorter dura-
tion and lower pill burden; and to assess the effect of
covariates, such as ART. It is reassuring nonetheless to
reafrm the strengthening evidence for shorter rifamy-
cin regimens.
Annals.org
 Network Meta-analysis of LTBI Treatments
From Institute for Global Health, University College London;
Public Health England; and Royal Free London National
Health Service Foundation Trust, London, United Kingdom,
and European Centre for Disease Prevention and Control,
Stockholm, Sweden.
Disclaimer: The views expressed in this manuscript are those
of the authors and not necessarily those of the U.K. National
Health Service, National Institute for Health Research, or De-
partment of Health.
Acknowledgment: Drs. Lipman and Stagg thank the U.K. Na-
tional Institute for Health Research for funding. The authors
thank Siyue Zhang and Dr. Hongxin Zhao for assistance with
the Chinese translation.
Financial Support: This report is independent research sup-
ported by postdoctoral fellowship PDF-2014-07-008 from the
U.K. National Institute for Health Research (Dr. Stagg).
Disclosures: Dr. Stagg reports travel and subsistence from the
World Health Organization, grants from the U.K. Department
of Health, and nonnancial support from Sano during the
conduct of the study and personal fees and other support
from Otsuka Pharmaceutical outside the submitted work. Au-
thors not named here have disclosed no conicts of interest.
Disclosures can also be viewed at www.acponline.org/authors
/icmje/ConictOfInterestForms.do?msNum=M17-0609.
Reproducible Research Statement: Study protocol:
PROSPERO CRD42016037871. Statistical code: See the Sup-
plement (available at Annals.org). Data set: Available from Dr.
Zenner (e-mail, dominik.zenner@ucl.ac.uk).
Requests for Single Reprints: Dominik Zenner, MD, Tubercu-
losis Section, Respiratory Diseases Department, Public Health
England, 61 Colindale Avenue, London NW9 5EQ, United
Kingdom; e-mail, dominik.zenner@ucl.ac.uk.
Current author addresses and author contributions are avail-
able at Annals.org.
References
1. World Health Organization. Global Tuberculosis Report 2016. Ge-
neva: WHO Pr; 2016. Accessed at www.who.int/tb/publications
/global_report/en on 21 May 2017.
2. United Nations. Sustainable Development Goals: 17 Goals
to Transform Our World. Accessed at www.un.org/sustainable
development/sustainable-development-goals on 21 May 2016.
3. World Health Organization. Global strategy and targets for tuber-
culosis prevention, care and control after 2015. Accessed at www
.who.int/tb/post2015_strategy/en on 21 May 2017.
4. World Health Organization. Guidelines on the Management of
Latent Tuberculosis Infection. Geneva: WHO Pr; 2015. Accessed
at www.who.int/tb/publications/ltbi_document_page/en on 21 May
2017.
5. Centers for Disease Control and Prevention. Treatment of latent
TB infection. Accessed at www.cdc.gov/tb/publications/ltbi
/treatment.htm on 21 May 2017.
6. Fox GJ, Dobler CC, Marais BJ, Denholm JT. Preventive therapy for
latent tuberculosis infectionthe promise and the challenges. Int J
Infect Dis. 2017;56:68-76. [PMID: 27872018] doi:10.1016/j.ijid.2016
.11.006
Annals.org
Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017
REVIEW
7. Stagg HR, Zenner D, Harris RJ, Munoz L, Lipman MC, Abubakar I.
Treatment of latent tuberculosis infection: a network meta-analysis.
Ann Intern Med. 2014;161:419-28. [PMID: 25111745] doi:10.7326
/M14-1019
8. Getahun H, Matteelli A, Abubakar I, Aziz MA, Baddeley A, Barreira
D, et al. Management of latent Mycobacterium tuberculosis infection:
WHO guidelines for low tuberculosis burden countries. Eur Respir J.
2015;46:1563-76. [PMID: 26405286] doi:10.1183/13993003.01245-
2015
9. Sandgren A, Vonk Noordegraaf-Schouten JM, Oordt-Speets AM,
van Kessel GB, de Vlas SJ, van der Werf MJ. Identifying components
for programmatic latent tuberculosis infection control in the Euro-
pean Union. Euro Surveill. 2016;21. [PMID: 27589214] doi:10.2807
/1560-7917.ES.2016.21.34.30325
10. Higgins JP, Altman DG, Gtzsche PC, Juni P, Moher D, Oxman
AD, et al; Cochrane Bias Methods Group. The Cochrane Collabora-
tion's tool for assessing risk of bias in randomised trials. BMJ. 2011;
343:d5928. [PMID: 22008217] doi:10.1136/bmj.d5928
11. Cooper NJ, Sutton AJ, Morris D, Ades AE, Welton NJ. Address-
ing between-study heterogeneity and inconsistency in mixed treat-
ment comparisons: application to stroke prevention treatments in
individuals with non-rheumatic atrial brillation. Stat Med. 2009;28:
1861-81. [PMID: 19399825] doi:10.1002/sim.3594
12. Samandari T, Agizew TB, Nyirenda S, Tedla Z, Sibanda T, Mosi-
maneotsile B, et al. Tuberculosis incidence after 36 months' isoniazid
prophylaxis in HIV-infected adults in Botswana: a posttrial observa-
tional analysis. AIDS. 2015;29:351-9. [PMID: 25686683] doi:10.1097
/QAD.0000000000000535
13. Sterling TR, Scott NA, Miro JM, Calvet G, La Rosa A, Infante R,
et al; Tuberculosis Trials Consortium, the AIDS Clinical Trials Group
for the PREVENT TB Trial. Three months of weekly rifapentine and
isoniazid for treatment of Mycobacterium tuberculosis infection in
HIV-coinfected persons. AIDS. 2016;30:1607-15. [PMID: 27243774]
doi:10.1097/QAD.0000000000001098
14. Samandari T, Agizew TB, Nyirenda S, Tedla Z, Sibanda T, Shang
N, et al. 6-month versus 36-month isoniazid preventive treatment for
tuberculosis in adults with HIV infection in Botswana: a randomised,
double-blind, placebo-controlled trial. Lancet. 2011;377:1588-98.
[PMID: 21492926] doi:10.1016/S0140-6736(11)60204-3
15. Sterling TR, Villarino ME, Borisov AS, Shang N, Gordin F, Bliven-
Sizemore E, et al; TB Trials Consortium PREVENT TB Study Team.
Three months of rifapentine and isoniazid for latent tuberculosis in-
fection. N Engl J Med. 2011;365:2155-66. [PMID: 22150035] doi:10
.1056/NEJMoa1104875
16. Danel C, Moh R, Gabillard D, Badje A, Le Carrou J, Ouassa T,
et al; TEMPRANO ANRS 12136 Study Group. A trial of early antiret-
rovirals and isoniazid preventive therapy in Africa. N Engl J Med.
2015;373:808-22. [PMID: 26193126] doi:10.1056/NEJMoa1507198
17. Hosseinipour MC, Bisson GP, Miyahara S, Sun X, Moses A, Riv-
iere C, et al; Adult AIDS Clinical Trials Group A5274 (REMEMBER)
Study Team. Empirical tuberculosis therapy versus isoniazid in adult
outpatients with advanced HIV initiating antiretroviral therapy
(REMEMBER): a multicountry open-label randomised controlled trial.
Lancet. 2016;387:1198-209. [PMID: 27025337] doi:10.1016/S0140
-6736(16)00546-8
18. Torre-Cisneros J, San-Juan R, Rosso-Fernandez CM, Silva JT,
Munoz-Sanz A, Munoz P, et al. Tuberculosis prophylaxis with levo-
oxacin in liver transplant patients is associated with a high incidence
of tenosynovitis: safety analysis of a multicenter randomized trial.
Clin Infect Dis. 2015;60:1642-9. [PMID: 25722196] doi:10.1093
/cid/civ156
19. Rangaka MX, Wilkinson RJ, Boulle A, Glynn JR, Fielding K, van
Cutsem G, et al. Isoniazid plus antiretroviral therapy to prevent tu-
berculosis: a randomised double-blind, placebo-controlled trial. Lan-
cet. 2014;384:682-90. [PMID: 24835842] doi:10.1016/S0140-6736
(14)60162-8
20. Rivero A, Lopez-Cortes L, Castillo R, Lozano F, Garca MA, Dez F,
et al; Grupo Andaluz para el estudio de las Enfermedades Infeccio-
sas (GAEI). [Randomized trial of three regimens to prevent tubercu-
Annals of Internal Medicine 7
 REVIEW
losis in HIV-infected patients with anergy]. Enferm Infecc Microbiol
Clin. 2003;21:287-92. [PMID: 12809582]
21. Gray DM, Workman LJ, Lombard CJ, Jennings T, Innes S, Grob-
belaar CJ, et al. Isoniazid preventive therapy in HIV-infected children
on antiretroviral therapy: a pilot study. Int J Tuberc Lung Dis. 2014;
18:322-7. [PMID: 24670570] doi:10.5588/ijtld.13.0354
22. Kim SH, Lee SO, Park IA, Kim SM, Park SJ, Yun SC, et al. Isoniazid
treatment to prevent TB in kidney and pancreas transplant recipients
based on an interferon--releasing assay: an exploratory randomized
controlled trial. J Antimicrob Chemother. 2015;70:1567-72. [PMID:
25608587] doi:10.1093/jac/dku562
23. Ma L, Lin B, Wang L, Wang D, Li G, Wang G. [Preventive therapy
for iatrogenic active tuberculosis in systemic lupus erythematosus
patients]. Zhonghua Yi Xue Za Zhi. 2014;94:3579-82. [PMID:
25622838]
24. Savovic J, Jones HE, Altman DG, Harris RJ, Juni P, Pildal J, et al.
Inuence of reported study design characteristics on intervention ef-
fect estimates from randomized, controlled trials. Ann Intern Med.
2012;157:429-38. [PMID: 22945832]
25. Stuurman AL, Vonk Noordegraaf-Schouten M, van Kessel F,
Oordt-Speets AM, Sandgren A, van der Werf MJ. Interventions for
improving adherence to treatment for latent tuberculosis infection: a
systematic review. BMC Infect Dis. 2016;16:257. [PMID: 27268103]
doi:10.1186/s12879-016-1549-4
26. Villarino ME, Scott NA, Weis SE, Weiner M, Conde MB, Jones B,
et al; International Maternal Pediatric and Adolescents AIDS Clinical
Trials Group. Treatment for preventing tuberculosis in children and
adolescents: a randomized clinical trial of a 3-month, 12-dose regi-
men of a combination of rifapentine and isoniazid. JAMA Pediatr.
2015;169:247-55. [PMID: 25580725] doi:10.1001/jamapediatrics
.2014.3158
27. Pease C, Hutton B, Yazdi F, Wolfe D, Hamel C, Quach P, et al.
Efcacy and completion rates of rifapentine and isoniazid (3HP) com-
pared to other treatment regimens for latent tuberculosis infection: a
8 Annals of Internal Medicine
Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017
Network Meta-analysis of LTBI Treatments
systematic review with network meta-analyses. BMC Infect Dis. 2017;
17:265. [PMID: 28399802] doi:10.1186/s12879-017-2377-x
28. Kahwati LC, Feltner C, Halpern M, Woodell CL, Boland E, Amick
HR, et al. Primary care screening and treatment for latent tuberculo-
sis infection in adults: evidence report and systematic review for the
US Preventive Services Task Force. JAMA. 2016;316:970-83. [PMID:
27599332] doi:10.1001/jama.2016.10357
29. Gwee A, Coghlan B, Curtis N. Question 1: what are the options
for treating latent TB infection in children? Arch Dis Child. 2013;98:
468-74. [PMID: 23661667] doi:10.1136/archdischild-2013-303876
30. Ayieko J, Abuogi L, Simchowitz B, Bukusi EA, Smith AH, Rein-
gold A. Efcacy of isoniazid prophylactic therapy in prevention of
tuberculosis in children: a meta-analysis. BMC Infect Dis. 2014;14:91.
[PMID: 24555539] doi:10.1186/1471-2334-14-91
31. Ayele HT, Mourik MS, Debray TP, Bonten MJ. Isoniazid prophy-
lactic therapy for the prevention of tuberculosis in HIV infected
adults: a systematic review and meta-analysis of randomized trials.
PLoS One. 2015;10:e0142290. [PMID: 26551023] doi:10.1371
/journal.pone.0142290
32. Evaluating the Pharmacokinetics, Tolerability, and Safety of
Once-Weekly Rifapentine and Isoniazid in HIV-1-Infected and
HIV-1-Uninfected Pregnant and Postpartum Women With Latent Tu-
berculosis Infection. ClinicalTrials.gov: NCT02651259. Accessed at
https://clinicaltrials.gov/ct2/show/study/NCT02651259 on 18 May
2017.
33. Evaluation of the Effect of 3HP vs Periodic 3HP vs 6H in HIV-
Positive Individuals (WHIP3TB). ClinicalTrials.gov: NCT02980016.
Accessed at https://clinicaltrials.gov/ct2/show/record/NCT02980016
on 18 May 2017.
34. Gupta RK, Rice B, Brown AE, Thomas HL, Zenner D, Anderson L,
et al. Does antiretroviral therapy reduce HIV-associated tuberculosis
incidence to background rates? A national observational cohort
study from England, Wales, and Northern Ireland. Lancet HIV.
2015;2:e243-51. [PMID: 26423197] doi:10.1016/S2352-3018(15)
00063-6
Annals.org
 Current Author Addresses: Dr. Zenner: Tuberculosis Section,
Respiratory Diseases Department, Public Health England, 61
Colindale Avenue, London NW9 5EQ, United Kingdom.
Drs. Beer and van der Werf: European Centre for Disease
Prevention and Control, Granitsvag 8, 17165 Solna, Sweden.
Mr. Harris: Statistics, Modelling and Bioinformatics Depart-
ment, Public Health England, 61 Colindale Avenue, London
NW9 5EQ, United Kingdom.
Dr. Stagg: Institute for Global Health, University College Lon-
don, 3rd Floor, 30 Guilford Street, London WC1N 1EH, United
Kingdom.
Dr. Lipman: Centre for Respiratory Medicine and Department
of Medicine, Royal Free London National Health Service Foun-
dation Trust, London NW3 2QG, United Kingdom.
Annals.org
Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/0/ on 08/02/2017
Author Contributions: Conception and design: D. Zenner,
M.J. van der Werf.
Analysis and interpretation of the data: D. Zenner, R.J. Harris,
M.C. Lipman, H.R. Stagg.
Drafting of the article: D. Zenner, N. Beer, R.J. Harris, H.R.
Stagg, M.J. van der Werf.
Critical revision of the article for important intellectual con-
tent: D. Zenner, R.J. Harris, M.C. Lipman, H.R. Stagg, M.J. van
der Werf.
Final approval of the article: D. Zenner, N. Beer, R.J. Harris,
M.C. Lipman, H.R. Stagg, M.J. van der Werf.
Statistical expertise: R.J. Harris.
Administrative, technical, or logistic support: D. Zenner, N.
Beer.
Collection and assembly of data: D. Zenner, N. Beer, H.R.
Stagg, M.J. van der Werf.
Annals of Internal Medicine