Clinical Infectious Diseases

MAJOR ARTICLE

Effect of Intermittency on Treatment Outcomes in

Pulmonary Tuberculosis: An Updated Systematic Review
and Metaanalysis
James C. Johnston,1 Jonathon R. Campbell,2 and Dick Menzies3
1
Division of Respiratory Medicine, Department of Medicine, University of British Columbia; TB Services, BC Centre for Disease Control, and 2Faculty of Pharmaceutical Sciences, University of

Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019

British Columbia, Vancouver,; and 3Division of Respiratory Medicine, Department of Medicine, McGill University; McGill International TB Centre, Montreal, Quebec, Canada

Background.  Intermittent regimens offer operational advantages in tuberculosis treatment, but their efficacy has been ques-
tioned. We updated a systematic review and metaanalysis to examine the efficacy of different intermittent dosing schedules in first-
line pulmonary tuberculosis therapy.
Methods.  An updated search included randomized control trials (RCTs) that reported on first-line pulmonary tuberculosis
therapy between June 2008 and March 2016. We pooled proportions of failure, relapse, and acquired drug resistance (ADR) for 4
dosing schedules: daily throughout, thrice weekly throughout, daily (intensive phase) then thrice weekly (continuation phase), and
daily (intensive phase) then twice weekly (continuation phase). Metaregression was performed using a negative binomial regression
model.
Results.  After screening 5874 citations, 7 RCTs with 10 arms were added for a total of 56 RCTs with 110 arms. The pooled pro-
portion of relapse was significantly higher in arms with thrice weekly therapy throughout (6.8; 95% confidence interval [CI], 3.8–9.9)
and twice weekly therapy in the continuation phase (7.3; 95% CI, 3.5–11.1) when compared with daily therapy (2.5; 95% CI, 1.8–3.2;
P < .01). Metaregression revealed higher rates of relapse (2.2; 95% CI, 1.2–4.0), failure (3.7; 95% CI, 1.1–12.6), and ADR (10.0; 95%
CI, 2.1–46.7) in arms with thrice weekly throughout and higher rates of failure (3.0; 95% CI, 1.0–8.8) with twice weekly in the con-
tinuation phase when compared with daily therapy.
Conclusion.  Thrice weekly dosing throughout therapy, and twice weekly dosing in the continuation phase appear to have worse
microbiological treatment outcomes when compared with daily therapy.
Keywords.  tuberculosis; intermittent therapy; treatment outcomes; systematic review; metaanalysis.

Over the past 3 decades, intermittent therapy has been widely
used in first-line tuberculosis treatment. This dosing strategy
reduces medication and healthcare worker costs for tubercu-
losis programs and has facilitated global scale-up of directly
observed therapy [1–3]. Various intermittent dosing regimens
for tuberculosis treatment have been recommended in more
than 130 countries [4]. Unfortunately, high rates of relapse and
acquired drug resistance have been noted in trials and tubercu-
losis programs that use intermittent therapy [2, 3, 5]. Since 2008
the World Health Organization (WHO) has recommended a
daily regimen as the preferred dosing schedule for all tubercu-
losis therapy [6]. However, intermittent regimens continue to
be an option in several national tuberculosis guidelines [7, 8]
Received 7 October 2016; editorial decision 12 January 2017; accepted 6 February 2017;
published online February 14, 2017.
Correspondence: J. Johnston, TB Services, BC Centre for Disease Control, 655 West 12th
Avenue, Vancouver, BC, V5V4R4 (james.johnston@bccdc.ca).
Clinical Infectious Diseases®  2017;64(9):1211–20
© The Author 2017. Published by Oxford University Press for the Infectious Diseases Society
of America. All rights reserved. For permissions, e-mail: journals.permissions@oup.com.
DOI: 10.1093/cid/cix121
and, until recently, was the recommended treatment schedule
in some high-incidence regions [4, 9].
Previous systematic reviews have examined outcomes of dif-
ferent dosing schedules [2, 3, 5, 10–13], including a 2009 sys-
tematic review by Menzies et  al that compared daily therapy
to 3 intermittent dosing schedules [5]. In this analysis, thrice
weekly therapy throughout treatment and twice weekly therapy
in the intensive phase were both associated with a nonsignifi-
cant increase in failure, relapse, and acquired drug resistance.
Thrice weekly dosing throughout therapy was also associated
with a higher rate of acquired drug resistance in metaregres-
sion. These results, though not statistically significant, pointed
toward daily therapy as the preferred regimen for first line
tuberculosis therapy.
Since 2009, several high-quality randomized, control tri-
als (RCTs) have been published that used the first-line WHO
standard treatment regimen [14–20]. These data have offered an
opportunity to update our metaanalysis of the effect of intermit-
tency on first line tuberculosis treatment outcomes.
We performed an update from the 2009 systematic review
performed by Menzies et  al [5] to address the following

Intermittent Therapy Systematic Review  •  CID 2017:64 (1 May) • 1211

question: What are the rates of treatment failure, relapse, and
acquired drug resistance with the different dose administration
schedules?
METHODS
In this review, we followed methods described in the PRISMA
statement for reporting of systematic reviews and metaanalyses
[21] (see PRISMA checklist in Supplementary Material).
Search Strategy
In the original review, a search of PubMed, Embase, and the
Cochrane Library was performed for the period 1965 to June
2008 and reported in detail elsewhere [5]. We performed an
updated search in the same 3 electronic databases from 1 June
2008 to 15 March 2016. Our search was restricted to RCTs
that reported treatment outcomes in first-line pulmonary
tuberculosis therapy. We included publications in English,
French, and Spanish. We examined the references of all full
text results, recent treatment guidelines, and prior system-
atic reviews for relevant studies not captured by the previous
search strategy. Two authors (J. C. J., J. R. C.) reviewed titles,
abstracts, and full text articles for inclusion.
Study Selection
We included RCTs that reported tuberculosis treatment out-
comes of failure, relapse, and/or acquired drug resistance.
Only data on new patients (ie, not previously treated) were
included. Consistent with current WHO recommendations,
Figure 1.  Study selection flow diagram. Abbreviations: ADR, acquired drug resistance; PLOS, public library of science; RCT, randomized control trial.
1212 • CID 2017:64 (1 May) •  Johnston et al
we included only studies with treatment regimens that used
rifampin for 6  months or longer (this criterion was added
since the last review). Arms with drug-susceptible strains or
unknown drug susceptibility (ie, smear only) were included
(this criterion was added since last review). Trial arms that
included rifapentine, rifabutin, second-line therapies, or non-
drug therapy were excluded. We also excluded trial arms that
involved once-weekly or monodrug therapy at any point in
the study.
Data Extraction and Quality Assessment
Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019
Consistent with the original review, we restricted studies to
high-quality RCTs. To be considered high quality, studies had
the following features: bacteriologic confirmation of initial
diagnosis (through microscopy or culture), bacteriologic con-
firmation of failure, and/or relapse; fewer than 10% of patients
refusing therapy, dropped out, moved away, or were otherwise
unaccounted for during therapy; and use of concealed alloca-
tion during randomization.
Outcome Definitions
Treatment failure was defined as sputum smears and/
or cultures consistently positive at the end of therapy or
≥5 months of therapy [22]. Relapse was defined as positive
smears and/or cultures that required therapy after treatment
success (completion or cure). In the presence of genotyp-
ing data, individuals with genotypic evidence of reinfection
were not counted as relapse. Acquired drug resistance was
Table 1.  Characteristics of Included Studies and Arms Outcomes were obtained from per protocol analysis when
available. Participants who did not complete therapy because
By Study By Arm
of drug reaction, treatment noncompletion, and death were
Characteristic n (%) n (%) not included in analysis.
Total 56 (100) 110 (100)
Publication language Statistical Analyses
 English 54 (96) 103 (94) We performed 2 types of analysis for this study. In the first
 French 2 (4) 7 (6)
metaanalysis, we pooled the risk differences for treatment
Year study began
 1969–1979 22 (39) 42 (38)
failure, relapse, and acquired drug resistance in RCTs with
 1980–1989 11 (20) 34 (31) head-to-head comparisons of different intermittency sched-
 1990–1999 14 (25) 21 (19) ules. Only 1 study that reported head-to head compari-

Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019

 2000–2009 2 (4) 3 (3) sons was identified, so this planned metaanalysis was not
 2010–Present 7 (12) 10 (9)
performed.
Sponsorship
In the second analysis, we combined data from all studies
 Public 55 (98) 108 (98)
 Corporate 1 (2) 2 (2) using a random effects model (procedure NLMIXED in SAS
Measured outcomes version 4.3) with the exact binomial likelihood approach to
 Failure 55 (98) 108 (98) calculate the pooled proportion, 95% confidence interval (CI),
 Relapse 53 (95) 105 (95) and test for significant differences in outcomes of treatment
  Acquired drug resistance 37 (66) 83 (75)
failure, relapse, and acquired drug resistance by rifampin
Adequate randomization 43 (77) 75 (68)
intermittency schedule [23]. Effectively, each RCT arm was
Blinding
  Single or double 4 (7) 6 (5) treated as an individual cohort, and the proportion of each
 None 52 (93) 104 (95) outcome was pooled to estimate an overall pooled effect esti-
Supervision of rifampin mate of proportion of each outcome by 1 of 4 predefined
 None/partial 27 (48) 47 (43) intermittency schedules: daily throughout (defined as dose
  All doses 29 (52) 63 (57)
administration ≥5 days per week), thrice weekly throughout
Drugs used in regimen
 Isoniazid 56 (100) 109 (99)
(defined as dose administration 3 times per week throughout
 Pyrazinamide 44 (79) 81 (26) therapy); daily then twice weekly (defined as daily [≥5 days
 Streptomycin 19 (34) 41 (37) per week] in the intensive phase, then 2 times per week in the
  Second-line drugs 1 (2) 2 (2) continuation phase); and daily then thrice weekly (defined as
Within study characteristics daily [≥5 days per week] in the intensive phase, then 3 times
Number of drugs to which strain is susceptible
per week in the continuation phase). We did not analyze stud-
Intensive phasea
 1 — 0
ies with twice weekly throughout therapy given the limited
 2 — 5 (6) data on this regimen.
 3 — 43 (50)
  4 or more 38 (44)
Multivariate Analysis
Continuation phasea
We performed metaregression to examine outcomes of fail-
  0 or 1 — 1 (1)
 2 — 60 (70) ure, relapse, and acquired drug resistance after adjusting for
  3 or more — 25 (29) relevant predefined covariates. Two models were developed.
Drug sensitivity The first model included the covariates of dosing schedule,
  Unknown (mix) — 24 (22) duration of pyrazinamide, duration of streptomycin, num-
  Pan sensitive — 86 (78)
ber of drugs to which the organism was susceptible in the
Rifampin use and duration
  6–7 mo — 93 (85)
intensive and continuation phase, supervision of therapy,
  8+ mo — 17 (15) proportion with default, and length of follow-up (for relapse,
Schedule of administration acquired drug resistance, and combination relapse–failure
  Daily throughout — 64 (58) outcome). Due to the high correlation between initial drug
  Daily then thrice weekly — 18 (17) resistance and the number of drugs to which the organism
  Daily then twice weekly — 9 (8)
was susceptible to in the intensive and continuation phases,
  Thrice weekly throughout — 19 (17)
a
the second model did not include the number of susceptible
Where Drug Susceptibility Testing was performed (86 arms).
drugs used but instead included the initial drug resistance
profile as a covariate. Metaregression was performed using
defined as new or additional resistance to 1 or more tuber- negative binomial regression in Stata (version 12.0) [23]. In
culosis drugs received among those who failed or relapsed. this metaregression, the arms of each study were the unit of

Intermittent Therapy Systematic Review  •  CID 2017:64 (1 May) • 1213

 Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019

Figure 2.  Forest plot of combined failure and relapse by administration schedule. Abbreviation: CI, confidence interval.

1214 • CID 2017:64 (1 May) •  Johnston et al

Table 2.  Stratified Estimates of Treatment Failures in Randomized Control Trials in New Cases

Pooled Event Rate

Factor Studies (N) Events/ Participants (N) (Across All Trials) 95% CI I2 (95% CI)

Duration of rifampin
  Rifampin 6–7 moa 92 163/11 940 0.4 0.1–0.6 0.15 (0.0–0.35)
  Rifampin 8+ mo 16 17/1461 0.2 0–0.6 0 (0–0.51)
Use of intermittent therapy
  Daily throughouta 62 112/8223 0.2 0.1–0.4 0.27 (0–0.47)
  Daily then thrice weekly 18 19/2075 0.4 0–1.1 0 (0–0.49)
  Daily then twice weekly 9 21/793 1.3b 0–2.9 0 (0–0.62)
  Thrice weekly throughoutc 19 28/2310 0.6 0–1.4 0 (0–0.48)

Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019

Initial drug resistance
  DST not done/reported 24 103/4392 2.0d 0.6–3.4 0.26 (0–0.55)
a
  Sensitive to all tuberculosis drugs 84 77/9009 0.2 0.1–0.3 0 (0–0.26)
Duration of pyrazinamide
  No pyrazinamidea 29 55/3295 0.1 0–0.3 0.46 (0.16–0.65)
  1–3 mo 60 117/8002 0.6b 0.2–1.1 0.09 (0–0.34)
  4+ mo 19 8/2104 0.2 0–0.4 0 (0–0.48)
Duration of streptomycin
  No streptomycina 67 176/9092 0.7 0.3–1.1 0.32 (0.08–0.50)
  1–3 mo 26 3/2475 0.1d 0–0.14 0 (0–0.43)
  4+ mo 15 1/1834 0.03d 0–0.1 0 (0–0.52)
Number of drugs to which strains susceptiblee
  Initial phase
  2 drugs 5 52/839 0.3b 0–1.5 0.92 (0.83–0.96)
  3 drugs 41 13/3944 0.03 0–0.09 0 (0–0.35)
  4 drugsa 38 12/4226 0.02 0–0.06 0 (0–0.37)
  Continuation phase
  0–1 drugs 1 0/84 0 — 0 (-,-)
  2 drugs 58 66/6284 0.04 0–0.1 0.04 (0–0.29)
  3 drugsa 25 11/2641 0.02 0–0.6 0 (0–0.43)
Supervision of therapy
  All doses fully superviseda 63 108/7055 0.2 0.0–0.5 0.10 (0.0–0.34)
  None or partial DOT 45 72/6346 0.3 0.0–0.5 0.01 (0.0–0.35)

Abbreviation: CI, confidence interval; DST, drug susceptibility testing; DOT, directly observed therapy.
a
Reference category.
b
Significantly different from reference at P < .05.
c
In all but 1 trial, if therapy was intermittent initially, the same schedule was continued throughout therapy.
d
Significantly different from reference at P < .01.
e
In a few trials, the number of drugs was the same throughout—these were classified according to the starting regimen.

analysis; the cumulative proportion of tuberculosis treat-
ment outcomes was the dependent variable, and tuberculo-
sis treatment characteristics were the independent variables.
An offset was used to account for study size. Effect estimates
were interpreted as adjusted incidence rate ratios, and the
significance of each factor in the models was evaluated using
the log likelihood ratio test [24].
Assessment of Heterogeneity and Bias
We assessed heterogeneity of risk differences for each com-
parison by estimating the I2 statistic and associated 95% CIs.
For this calculation, studies in which both treatment arms had
no events were corrected by 0.5. We assessed bias by generat-
ing funnel plots for the proportion of each outcome (failure,
relapse, acquired drug resistance) in the primary analysis.
Primary and Sensitivity Analysis
For our primary outcome, we performed metaanalysis and metar-
egression on all trial arms with rifampin duration ≥6  months.
We included individuals with smear or culture positivity with
a pan-sensitive strain or an unknown susceptibility profile to
reflect care in many low-resource settings. We included studies
with rifampin duration ≥6 months to reflect the current WHO
standard. We performed similar analyses in several subsets of
the study arms; trial arms with confirmed drug-susceptible dis-
ease (with and without streptomycin resistance); trial arms with-
out any patients with known human immunodeficiency virus
(HIV) infection; trial arms with drug-resistant isolates; and trial
arms that used 2 months of isoniazid (H), rifampin (R), pyrazi-
namide (Z), +/- ethambutol (E), followed by 4  months of HR
+/- E (abbreviated as 2HRZ(E)/4HR(E)).

Intermittent Therapy Systematic Review  •  CID 2017:64 (1 May) • 1215

Table 3.  Stratified Estimates of Relapse in Randomized Control Trials in New Cases

Pooled Event Rate

Factor Studies (N) Events/Participants (N) (Across All Trials) 95% CI I2 (95% CI)

Duration of rifampin
  Rifampin 6–7 moa 89 481/10 934 3.8 2.8–4.7 0.50 (0.36–0.61)
  Rifampin 8+ mo 16 22/1250 1.4b 0.4–2.4 0 (0–0.51)
Use of intermittent therapy
  Daily throughouta 59 254/7475 2.5 1.8–3.2 0.29 (0.02–0.49)
  Daily then thrice weekly 18 72/2007 3.0 1.0–5.1 0 (0–0.49)
  Daily then twice weekly 9 49/572 7.3b 3.5–11.1 0.64 (0.27–0.83)
  Thrice weekly throughoutc 19 128/2130 6.8b 3.8–9.9 0.71 (0.55–0.82)

Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019

Initial drug resistance
  DST not done/reported 22 201/3913 5.4d 3.1–7.6 0.70 (0.54–0.81)
  Sensitive to all tuberculosis drugsa 83 302/8271 2.9 2.1–3.7 0.26 (0.03–0.44)
Duration of pyrazinamide
  No pyrazinamidea 28 112/2578 3.2 1.7–4.7 0.40 (0.05–0.62)
  1–3 mo 59 303/7733 3.2 2.2–4.2 0.39 (0.17–0.56)
  4+ mo 18 88/1873 3.8 1.6–5.9 0.58 (0.29–0.75)
Duration of streptomycin
  No streptomycina 65 358/8289 3.4 2.5–4.4 0.49 (0.32–0.62)
  1–3 mo 25 68/2167 2.3 1.1–3.5 0 (0–0.43)
  4+ mo 15 77/1728 4.2 1.9–6.6 0.59 (0.28–0.77)
Number of drugs to which strains susceptiblee
  Initial phase
  2 drugs 5 34/609 4.6d 0–12.2 0.40 (0–0.78)
  3 drugs 41 124/3842 1.7 0.5–2.9 0.04 (0–0.32)
  4 drugsa 37 144/3820 2.5 0.8–4.2 0.33 (0–0.56)
  Continuation phase
  0–1 drugs 1 2/80 2.0 0–9.3 0 (-,-)
  2 drugs 58 187/5898 1.8 0.8–2.9 0.05 (0–0.31)
  3 drugsa 24 113/2293 3.5 0.6–6.5 0.54 (0.28–0.71)
Supervision of therapy
  All doses fully superviseda 59 313/6345 4.6 3.4–5.8 0.51 (0.34–0.64)
  None or partial DOT 46 190/5839 2.3b 1.6–3.0 0.28 (0–0.50)

Abbreviation: CI, confidence interval; DST, drug susceptibility testing; DOT, directly observed therapy.
a
Reference category.
b
Significantly different from reference at P < .01.
c
In all but 1 trial, if therapy was intermittent initially, the same schedule was continued throughout therapy.
d
Significantly different from reference at P < .05.
e
In a few trials, the number of drugs was the same throughout—these were classified according to the starting regimen.

RESULTS
Study Selection and Assessment
After removing duplicates, 5874 citations were identified
from the updated search. Of these, 546 studies were retained
for abstract review and 105 for full text review. An additional
3 full text studies were identified from the reference search,
resulting in 108 publications for full text review. After full text
review, 7 RCTs were included for analysis (Figure 1, Table 1),
adding 3338 participants from 10 trial arms. This included
5 arms that reported results of daily therapy, 2 arms that
reported daily followed by intermittent therapy, and 3 arms
that reported intermittent therapy throughout treatment.
All updated studies involved adults and reported at least 1
arm with the standard WHO first-line treatment regimen of
2HRZE/4HR(E).
After combining the results of the original search strategy and
limiting inclusion to only studies that used ≥6 months of rifampin
in a strain with an unknown susceptibility profile or that was pan
susceptible, the updated systematic review included 56 RCTs with
110 arms; 8 RCTs from the original analysis were excluded from
analysis (Figure 2). The outcomes of failure were examined in 108
arms with 13 401 participants; relapse was available for 105 arms
with 12 184 participants; and acquired drug resistance was avail-
able for 83 arms with 7443 participants. The updated literature
search yielded no studies that examined intermittency schedules in
head-to-head comparison and only 1 study with 223 patients that
reported on daily vs twice-weekly therapy from the prior search.
Pooled Results Across Studies
We pooled results from the 110 arms that reported treatment
outcomes from regimens with ≥6 months of rifampin in arms

1216 • CID 2017:64 (1 May) •  Johnston et al

Table 4.  Stratified Estimates of Acquired Drug Resistance in Randomized Control Trials in New Cases

Pooled Event Rate

Factor Studies (N) Events/Participants (N) (Across All Trials) 95% CI I2 (95% CI)

Duration of rifampin
  Rifampin 6–7 moa 58 26/6210 0.1 0–0.3 0 (0–0.31)
  Rifampin 8+ mo 14 2/1233 0.1 0–0.2 0 (0–0.54)
Use of intermittent therapy
  Daily throughouta 43 11/4700 0.09 0–0.2 0 (0–0.35)
  Daily then thrice weekly 9 1/588 0.08 0–0.3 0 (0–0.62)
  Daily then twice weekly 5 2/377 0.2 0–0.6 0 (0–0.74)
  Thrice weekly throughoutb 15 16/1778 0.3 0–0.8 0 (0–0.52)

Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019

Duration of pyrazinamide
  No pyrazinamidea 23 4/2349 0.05 0–0.1 0 (0–0.45)
  1–3 mo 32 10/3111 0.1 0–0.3 0 (0–0.39)
  4+ mo 17 16/1983 0.1 0–0.4 0 (0–0.50)
Duration of streptomycin
  No streptomycina 37 21/3504 0.1 0–0.3 0 (0–0.37)
  1–3 mo 20 3/2105 0.04 0–0.1 0 (0–0.47)
  4+ mo 15 6/1834 0.1 0–0.3 0 (0–0.52)
Number of drugs to which strains susceptiblec
  Initial phase
  2 drugs 2 2/209 0.4 0–1.9 0 (-,-)
  3 drugs 35 14/3312 0.05 0–0.1 0 (0–0.38)
  4 drugsa 35 14/3922 0.1 0–0.2 0 (0–0.38)
  Continuation phase
  0–1 drugs 1 0/84 0 — 0 (-,-)
  2 drugs 50 17/4986 0.1 0–0.2 0 (0–0.33)
  3 drugsa 21 13/2373 0.1 0–0.3 0 (0–0.46)
Supervision of therapy
  All doses fully superviseda 43 27/4418 0.2 0–0.5 0 (0–0.35)
  None or partial DOT 29 3/3025 0.03d 0–0.1 0 (0–0.41)

Abbreviation: CI, confidence interval; DOT, directly observed therapy.

a
Reference category.
b
In all but 1 trial, if therapy was intermittent initially, the same schedule was continued throughout therapy.
c
In a few trials, the number of drugs was the same throughout—these were classified according to the starting regimen.
d
Significantly different from reference at P < .05.

with strains that were pan susceptible or did not undergo DST.
Four administration schedules were analyzed, with outcomes
from primary analysis reported on 64 arms with 8839 partic-
ipants in daily therapy; 19 arms/2310 participants with thrice
weekly throughout; 18 arms/2075 participants with daily then
thrice weekly; and 9 arms/793 participants with daily then twice
weekly (Figure 2). The pooled proportions of participants with
failure, relapse, and acquired drug resistance by drug intermit-
tency schedule are listed in Tables 2–4. There was a significant
increase in the pooled proportion of participants with relapse
in those receiving thrice weekly (6.8; 95% CI, 3.8–9.9) and daily
then twice weekly (7.3; 95% CI, 3.5–11.1) compared with daily
therapy throughout (2.5; 95% CI, 1.8–3.2; P < .01). Participants
who received daily then twice weekly therapy also had an
increase in pooled proportion of failure (1.3; 95% CI, 0–2.9)
compared with those who received daily throughout (0.2; 95%
CI, 0.1–0.4; P  <  .05). There was no significant difference in
the proportion of participants with acquired drug resistance
regardless of administration schedule.
Metaregression
After adjusting for the potentially confounding treatment fac-
tors described above, participants who received thrice weekly
therapy throughout had significantly higher incidence rate
ratios (IRRs) for failure (3.7; 95% CI, 1.1–12.6), relapse (2.2;
95% CI, 1.2–4.0), and acquired drug resistance (10.0; 95% CI,
2.1–46.7) when compared with participants with daily therapy
(Table 5). Participants with daily then twice weekly therapy had
a significantly higher IRR for failure (3.0; 95% CI, 1.0–8.8) and
a nonstatistically significant increase in relapse (1.8; 95% CI,
1.0–3.3) when compared with those with daily therapy. Similar
to the results of univariate analysis, daily then thrice weekly
therapy had similar rates of failure, relapse, and acquired drug
resistance when compared with daily therapy.

Intermittent Therapy Systematic Review  •  CID 2017:64 (1 May) • 1217

Table 5.  Adjusted Incidence Rate Ratios of Failure, Relapse, Acquired Drug Resistance—From Negative Binomial Regression

Factor Failure IRR (95% CI) Relapse IRR (95% CI) Acquired Drug Resistancea IRR (95% CI)
b
Duration of rifampin
  Rifampin 6–7 mo 1.0 (reference) 1.0 (reference) 1.0 (reference)
  Rifampin 8+ mo 1.1 (0.4–3.6) 0.4 (0.2–0.7) 1.8 (0.4–8.4)
  Overall significance (P value)c (0.841) (0.002) (0.508)
Use of intermittent therapyb
  Daily throughout 1.0 (reference) 1.0 (reference) 1.0 (reference)
  Daily then thrice weekly 1.5 (0.4–5.4) 1.2 (0.6–2.3) 0.6 (0.1–5.7)
  Daily then twice weekly 3.0 (1.0–8.8) 1.8 (1.0–3.3) 1.0 (0.2–5.0)
  Thrice weekly throughout 3.7 (1.1–12.6) 2.2 (1.2–4.0) 10.0 (2.1–46.7)

Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019

  Overall significance (P value)c (0.0860) (0.027) (0.020)
Initial drug resistanced
  Sensitive to all tuberculosis drugs 1.0 (reference) 1.0 (reference) 1.0 (reference)
  DST not done/reported 6.2 (2.6–14.8) 2.5 (1.7–3.68) —
  Overall significance (P value) c (<0.001) (<0.001) —
Duration of pyrazinamideb
  No pyrazinamide 1.0 (reference) 1.0 (reference) 1.0 (reference)
  Any pyrazinamide 3.6 (0.7 to 18.1) 0.4 (0.2–0.7) 1.7 (0.2–12.2)
  Overall significance (P value)c (0.106) (0.003) (0.614)
Duration of streptomycinb
  No streptomycin 1.0 (reference) 1.0 (reference) 1.0 (reference)
  Any streptomycin 0.1 (0.0–0.2) 0.6 (0.4–0.8) 0.2 (0.1–0.6)
Overall significance (P value)c (<0.001) (0.004) (0.006)
Number of drugs to which strains susceptibleb
  Initial phase
  2 drugs 12.7 (1.6–99.8) 0.43 (0.16–1.12) 1.8 (0.1–28.4)
  3 drugs 0.4 (0.2–0.9) 0.5 (0.3–0.8) 0.8 (0.2–2.9)
  4 drugs 1.0 (reference) 1.0 (reference) 1.0 (reference)
   Overall significance (P value)c (<0.001) (0.027) (0.727)
  Continuation phase
  0–1 drugs 0 (-,-) 2.04 (0.31–13.33) 0 (-,-)
  2 drugs 1.6 (0.6–4.4) 1.09 (0.7–1.7) 1.2 (0.4–3.9)
  3 drugs 1.0 (reference) 1.0 (reference) 1.0 (reference)
   Overall significance (P value)c (0.603) (0.7385) (0.874)

Abbreviations: CI, confidence interval; DST, drug susceptibility testing; IRR, incidence rate ratios.
a
Acquired drug resistance in both failure and relapse cases combined.
b
Adjusted IRR from multivariate binomial regression. Model includes all variables listed plus length of follow-up after end of treatment (for all except Failure), therapy supervision, and
treatment completion. Statistically significant estimates in bold.
c
Overall significance of each factor in multivariate models, from log likelihood test.
d
Adjusted IRR from multivariate binomial regression. Model includes duration of rifampin, intermittency, duration of pyrazinamide, duration of streptomycin, therapy supervision, treatment
completion, and length of follow-up after end of treatment (for all except Failure). Statistically significant estimates in bold.

Sensitivity Analysis and Assessment of Bias
Results of several analyses are presented in Supplementary
Material. Given the concern over the effect of HIV on inter-
mittency outcomes, we aimed to pool data on intermittency
by HIV status. Unfortunately, data on HIV-infected individu-
als were incompletely reported; 9 arms examined outcomes in
only HIV-infected populations, while 15 arms reported >10%
significantly higher, while there was a nonsignificant increase in
the failure rate (see Supplementary Material). After accounting
for the low proportions for each outcome, visual inspection of
the funnel plots revealed no imbalances consistent with publi-
cation bias (see Supplementary Material).
DISCUSSION

HIV prevalence in study participants. Overall, there were too
few studies with HIV-infected participants to examine pooled
proportion of treatment outcomes or perform metaregression
by HIV proportion. We performed an analysis that excluded
arms with any HIV-infected participants to assess the effect
on pooled outcomes. In metaregression of this population, the
pooled proportion of relapse and acquired drug resistance were
In this updated review in which the effect of intermittency on
first-line pulmonary tuberculosis treatment outcomes was
examined, we found a significant increase in microbiologically
defined adverse outcomes in participants who received thrice
weekly therapy throughout treatment. This finding was main-
tained after we accounted for potential confounding variables in
metaregression and on multiple sensitivity analyses that varied

1218 • CID 2017:64 (1 May) •  Johnston et al

by susceptibility profile, treatment regimens, and HIV status. The
evidence against intermittent therapy in the continuation phase
was less clear. Daily therapy in the intensive phase followed by
twice weekly therapy appeared to have worse outcomes than daily
therapy throughout, with higher rates of failure and relapse in
pooled analysis and a significant increase in failure in metaregres-
sion. Meanwhile, thrice weekly therapy in the continuation phase
showed a nonsignificant increase in relapse in pooled analysis but
without a clear signal toward worse outcomes in metaregression.
Our findings build on evidence from recent systematic reviews
and modeling studies to argue against intermittent dosing
throughout first-line pulmonary tuberculosis therapy [2, 3, 5, 11].
A previous systematic review performed by members of our group
demonstrated an increase in the pooled proportion of relapse with-
out a statistically significant increase in failure or acquired drug
resistance [5]. Similarly, in their systematic review of RCTs and
cohort studies, Kasozi et al reported a significant increase in relapse
compared with daily therapy but did not evaluate other microbio-
logic treatment outcomes [11]. Lastly, Chang et al demonstrated an
increase in relapse in a systematic review and modeling study in
which they examined different therapy schedules [3].
Our review has several strengths. We examined data from 56
RCTs spanning several decades and resource settings. The RCTs
we examined were considered high quality with comprehensive
reporting and robust, microbiologically confirmed treatment
outcomes. We found similar outcomes in multivariate analysis
and numerous sensitivity analyses that varied by patient popu-
lation, susceptibility profiles, and treatment regimens.
Our study also has several limitations. First, similar to the
2009 systematic review, we found only 1 trial that examined
head-to-head comparisons of rifampin intermittency [5]. The
absence of head-to-head comparisons increases the potential
for confounding by patient, strain, and treatment characteris-
tics, as the randomization performed in each RCT did not apply
to dose schedule itself. The absence of comparisons between
dosing schedules also precluded network metaanalysis, which
requires multiple pairwise comparisons between treatment reg-
imens for robust analysis [25]. In our case, the limited number
of comparisons between dosing schedules and the limited sam-
ple size within each comparison arm would have limited this
analytic technique. Instead, we performed metaregression to
control for potential confounding covariates.
In addition, many of the studies were performed before 1980,
when treatment regimens, HIV prevalence, and programmatic
characteristics were substantially different from today. We
attempted to control for this deficiency by examining stud-
ies with WHO-recommended treatment regimens in order to
reflect more modern tuberculosis treatment and in order to
compare outcomes with and without streptomycin therapy. We
are reassured by the results of these sensitivity analyses, which
were largely similar to the outcomes of our primary analysis
(Supplementary Material).
Finally, we did not examine treatment noncompletion and
adverse events as treatment outcomes. Instead, we focused on
more strictly defined, microbiologically confirmed treatment
outcomes in an attempt to minimize between-study variability.
Intermittent regimens were developed to overcome patient and
programmatic barriers to therapy; these regimens offer less pill
burden to patients, less potential for daily adverse events, and
fewer barriers to programmatic implementation of directly
observed therapy [1]. Consistent with the programmatic ben-
efits of intermittent therapy, a systematic review by Kasozi et
al demonstrated decreased default with intermittent regimens
Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019
compared with daily regimens throughout [11]. Their review
included several cohorts, which are presumably more reflec-
tive of programmatic settings. In contrast, our analysis focused
on data from high-quality RCTs with strictly defined, micro-
biologically confirmed treatment outcomes. Presumably, data
obtained under RCT conditions reflect more idealized condi-
tions, which may in turn result in improved adherence rates
and may underestimate the benefits of intermittent therapy. We
recognize that intermittent therapy may result in lower treat-
ment noncompletion rates but also recognize that low non-
completion rates can be achieved in tuberculosis programs in a
variety of resource settings. We do not believe that implemen-
tation of a less efficacious regimen to overcome programmatic
deficiencies is necessarily a desirable approach.
CONCLUSIONS
This review provides evidence against the use of intermittent
tuberculosis therapy throughout treatment, demonstrating that
this type of intermittent regimen is inferior in all 3 outcomes
measured on pooled analysis and metaregression. Likewise,
twice weekly therapy in the intensive phase appears to be asso-
ciated with worse outcomes, while thrice weekly therapy in the
continuation phase does not appear to be clearly inferior but
may have an increased risk relapse.
Despite the wealth of trial and operational data on the WHO
standard 6-month treatment regimen, we still have much to
learn about dosing of tuberculosis therapy. Adequately powered
clinical trials that examine dosing schedules in first-line treat-
ment of drug-susceptible tuberculosis are still needed to fully
understand the impact of intermittency on tuberculosis treat-
ment outcomes in first-line therapy.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases
online. Consisting of data provided by the authors to benefit the reader,
the posted materials are not copyedited and are the sole responsibility of
the authors, so questions or comments should be addressed to the corre-
sponding author.
Notes
Financial support.  This work was supported by the Michael Smith
Foundation for Health Research (J. C. J.).
Intermittent Therapy Systematic Review  •  CID 2017:64 (1 May) • 1219
 Potential conflicts of interest.  All authors: No potential conflicts of
interest. All authors have submitted the ICMJE Form for Disclosure of
Potential Conflicts of Interest. Conflicts that the editors consider relevant to
the content of the manuscript have been disclosed.
References
1. Iseman MD. Tuberculosis therapy: past, present and future. Eur Respir J Suppl
2002; 36:87s–94s.
2. Ahmad Khan F, Minion J, Al-Motairi A, Benedetti A, Harries AD, Menzies D. An
updated systematic review and meta-analysis on the treatment of active tubercu-
losis in patients with HIV infection. Clin Infect Dis 2012; 55:1154–63.
3. Chang KC, Leung CC, Yew WW, Chan SL, Tam CM. Dosing schedules of
6-month regimens and relapse for pulmonary tuberculosis. Am J Respir Crit Care
Med 2006; 174:1153–8.
4. Jain Y. India should introduce daily drug treatment for tuberculosis. BMJ 2013;
347:f6769.
5. Menzies D, Benedetti A, Paydar A, et al. Standardized treatment of active tuber-
culosis in patients with previous treatment and/or with mono-resistance to isoni-
azid: a systematic review and meta-analysis. PLoS Med 2009; 6:e1000150.
6. World Health Organization. Treatment of tuberculosis: guidelines. Geneva,
Switzerland: WHO, 2010.
7. National Institute for Health and Care Excellence. Tuberculosis Guidelines.
United Kingdom: NICD, 2016.
8. Nahid P, Dorman SE, Alipanah N, et  al. Official American Thoracic Society/
Centers for Disease Control and Prevention/Infectious Diseases Society of
America clinical practice guidelines: treatment of drug-susceptible tuberculosis.
Clin Infect Dis 2016; 63:e147–95.
9. World Health Organization. Standards for TB Care in India. India: WHO, 2014.
10. Khan FA, Minion J, Pai M, et al. Treatment of active tuberculosis in HIV-coinfected
patients: a systematic review and meta-analysis. Clin Infect Dis 2010; 50:1288–99.
11. Kasozi S, Clark J, Doi SA. Intermittent versus daily pulmonary tuberculosis treat-
ment regimens: a meta-analysis. Clin Med Res 2015; 13:117–38.
12. Bose A, Kalita S, Rose W, Tharyan P. Intermittent versus daily therapy for treating
tuberculosis in children. Cochrane Database Syst Rev 2014; Cd007953.
13. Mwandumba HC, Squire SB. Fully intermittent dosing with drugs for treating
tuberculosis in adults. Cochrane Database Syst Rev 2001; Cd000970.
1220 • CID 2017:64 (1 May) •  Johnston et al
14. Blanc FX, Sok T, Laureillard D, et al; CAMELIA Study Team. Earlier versus later
start of antiretroviral therapy in HIV-infected adults with tuberculosis. N Engl J
Med 2011; 365:1471–81.
15. Gillespie SH, Crook AM, McHugh TD, et al; REMoxTB Consortium. Four-month
moxifloxacin-based regimens for drug-sensitive tuberculosis. N Engl J Med 2014;
371:1577–87.
16. Johnson JL, Hadad DJ, Dietze R, et al. Shortening treatment in adults with non-
cavitary tuberculosis and 2-month culture conversion. Am J Respir Crit Care Med
2009; 180:558–63.
17. Lienhardt C, Cook SV, Burgos M, et al; Study C Trial Group. Efficacy and safety
of a 4-drug fixed-dose combination regimen compared with separate drugs for
treatment of pulmonary tuberculosis: the study C randomized controlled trial.
JAMA 2011; 305:1415–23.
18. Merle CS, Fielding K, Sow OB, et  al; OFLOTUB/Gatifloxacin for Tuberculosis
Project. A four-month gatifloxacin-containing regimen for treating tuberculosis.
Downloaded from https://academic.oup.com/cid/article-abstract/64/9/1211/2993841 by Airlangga University user on 20 November 2019
N Engl J Med 2014; 371:1588–98.
19. Mfinanga SG, Kirenga BJ, Chanda DM, et  al. Early versus delayed initia-
tion of highly active antiretroviral therapy for HIV-positive adults with
newly diagnosed pulmonary tuberculosis (TB-HAART): a prospective,
international, randomised, placebo-controlled trial. Lancet Infect Dis 2014;
14:563–71.
20. Swaminathan S, Narendran G, Venkatesan P, et al. Efficacy of a 6-month ver-
sus 9-month intermittent treatment regimen in HIV-infected patients with
tuberculosis: a randomized clinical trial. Am J Respir Crit Care Med 2010;
181:743–51.
21. Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred report-
ing items for systematic reviews and meta-analyses: the PRISMA statement. BMJ
2009; 339:b2535.
22. World Health Organization. Definitions and reporting framework for tuberculo-
sis—2013 revision. Switzerland: WHO, 2013.
23. Hamza TH, van Houwelingen HC, Stijnen T. The binomial distribution of
meta-analysis was preferred to model within-study variability. J Clin Epidemiol
2008; 61:41–51.
24. Glasziou PP, Sanders SL. Investigating causes of heterogeneity in systematic
reviews. Stat Med 2002; 21:1503–11.
25. Mills EJ, Thorlund K, Ioannidis JP. Demystifying trial networks and network
meta-analysis. BMJ 2013; 346:f2914.