Review

Treatment outcomes among patients with

multidrug-resistant tuberculosis: systematic review
and meta-analysis
Evan W Orenstein, Sanjay Basu, N Sarita Shah, Jason R Andrews, Gerald H Friedland, Anthony P Moll, Neel R Gandhi, Alison P Galvani

Multidrug-resistant (MDR) tuberculosis is a growing clinical and public-health concern. To evaluate existing evidence Lancet Infect Dis 2009;
regarding treatment regimens for MDR tuberculosis, we used a Bayesian random-effects meta-analysis of the available 9: 153–61

therapeutic studies to assess how the reported proportion of patients treated successfully is influenced by differences See Leading Edge page 137
in treatment regimen design, study methodology, and patient population. Successful treatment outcome was defined Tugela Ferry Care and Research
Collaboration, Tugela Ferry,
as cure or treatment completion. 34 clinical reports with a mean of 250 patients per report met the inclusion criteria.
KwaZulu-Natal, South Africa
Our analysis shows that the proportion of patients treated successfully improved when treatment duration was at (E W Orenstein BS, S Basu PhD,
least 18 months, and if patients received directly observed therapy throughout treatment. Studies that combined both N S Shah MD, J R Andrews MD,
factors had significantly higher pooled success proportions (69%, 95% credible interval [CI] 64–73%) than other G H Friedland MD,
A P Moll MBChB, N R Gandhi MD,
studies of treatment outcomes (58%, 95% CI 52–64%). Individualised treatment regimens had higher treatment
A P Galvani PhD); Department
success (64%, 95% CI 59–68%) than standardised regimens (54%, 95% CI 43–68%), although the difference was not of Epidemiology and Public
significant. Treatment approaches and study methodologies were heterogeneous across studies. Many important Health (E W Orenstein, S Basu,
variables, including patients’ HIV status, were inconsistently reported between studies. These results underscore the G H Friedland, A P Galvani), and
AIDS Program (S Basu,
importance of strong patient support and treatment follow-up systems to develop successful MDR tuberculosis
G H Friedland), Yale University
treatment programmes. School of Medicine, New
Haven, CT, USA; Divisions of
Introduction therapy,8 some use DOT only for the intensive phase,9 General Internal Medicine,
Infectious Diseases, and
Mycobacterium tuberculosis strains that are resistant to anti- and others incorporate DOT throughout treatment.10
Epidemiology, Albert Einstein
tuberculosis medications have generated increasing MDR tuberculosis treatment programmes also vary in College of Medicine and
international concern. An estimated 489 000 cases of other characteristics, including the size of drug regimens, Montefiore Medical Center,
multidrug-resistant (MDR) tuberculosis (defined as duration of treatment, definitions of cure, and follow-up Bronx, NY, USA (N S Shah,
N R Gandhi); Department of
resistance to at least isoniazid and rifampin) occurred protocols. The impact of these variations on the Internal Medicine, University
worldwide in 2006.1 Treatment of drug-resistant probability that patients will achieve treatment success is of California San Francisco,
tuberculosis is expensive and complex because it unknown. San Francisco, CA, USA
necessitates the use of second-line tuberculosis drugs, Previous reviews of MDR tuberculosis therapy have (J R Andrews); and Church of
Scotland Hospital and
which are associated with a greater incidence of adverse not identified which factors are the most important Philanjalo, Tugela Ferry
reactions,2,3 and require a longer treatment duration than contributors to treatment success. One review found that (A P Moll)
first-line drugs. However, comprehensive treatment initial drug resistance and treatment composition could Correspondence to:
programmes have shown the efficacy of MDR tuberculosis predict the development of acquired drug resistance and Evan Orenstein, Department of
treatment,4 and mathematical models have suggested that treatment failure for patients receiving first-line drug Epidemiology and Public Health,
Yale University School of
this therapy is cost-effective in resource-poor settings.5 regimens, but did not present outcomes specifically from Medicine, 60 College Street,
Nevertheless, current guidelines for MDR tuberculosis MDR tuberculosis patients or examine the impact of New Haven, CT 06520, USA
management are largely based on expert opinion and case other treatment programme characteristics.11 Other Orenstein.Evan@gmail.com
series, rather than on the results of clinical trials.6 reviews have assessed treatment outcomes in patients
Guidelines in different countries are based on variable with MDR tuberculosis, but included too few studies to
health-system approaches to MDR tuberculosis treatment. statistically determine which factors specifically affect
Some programmes use second-line drug-susceptibility treatment success.12,13
testing to design individualised treatment regimens for In this study, we did a systematic review and
patients with MDR tuberculosis, minimising amplification meta-analysis of available therapeutic studies to
of resistance and sparing patients from otherwise toxic characterise factors associated with improved treatment
drugs. Other programmes use standardised drug regimens outcomes among patients with MDR tuberculosis who
based on population surveys of local drug-susceptibility were treated with second-line drugs. Our analysis
patterns in the context of limited laboratory capacity or assesses the role of individualised versus standardised
pharmaceutical access (such as lack of participation in the treatment regimens, characteristics of patients and
WHO Green Light Committee programme, which programmes, study settings, and outcome definitions on
provides countries with access to quality-assured second- the reported efficacy of MDR tuberculosis treatment.
line drugs at substantially reduced prices).7
Programmes differ in their use of strategies to promote Methods
adherence such as directly observed therapy (DOT). We did our meta-analysis in accordance with QUORUM
Some treatment programmes use only self-administered guidelines.14

www.thelancet.com/infection Vol 9 March 2009 153

 Review

analysis. References from the selected studies were also

565 articles in English identified from PubMed, assessed to ensure that relevant studies were not
Cochrane Database, EMBASE, CINAHL, Web
of Science, and archives of the International omitted. Articles were independently assessed by two
Journal of Tuberculosis and Lung Disease reviewers (EWO and JRA), with disagreements mitigated
(after elimination of duplicates)
by a third author (SB). Studies were required to meet
the following inclusion criteria: (1) confirmation
312 excluded that patients had MDR tuberculosis using
89 reviews
176 with no treatment outcome in human drug-susceptibility testing on cultured M tuberculosis;
beings (2) treatment outcome definitions specified by
32 case reports with fewer than 10 patients
13 with outcomes for disease other
mycobacterial culture endpoints (eg, cure defined as at
than tuberculosis least five consecutive negative cultures during the last
3 commentaries 12 months of treatment); (3) clearly defined treatment
protocols including second-line drugs; and (4) outcomes
253 articles retained for abstract reviews reported according to WHO classifications of success
(including cure or treatment completion), failure,
default (treatment interruption), and death.6 Studies in
19 additional full texts identified 177 abstracts excluded
from references and reviews 75 did not use second-line drugs
which all patients had extensively drug-resistant
81 had outcomes for MDR tuberculosis tuberculosis were excluded.
missing or incomplete
21 case reports with fewer than 10 patients
with MDR tuberculosis Data abstraction
We recorded treatment outcomes according to WHO
classifications of treatment success (cure and treatment
95 articles received full-text review
completion), failure, default, and death.6 patients still on
treatment who were classified as “probable cure” or
62 excluded after review “probable failure” were added to the success and failure
8 not available categories, respectively. patients who remained on treat-
3 duplicate populations
13 with surgery as primary mode of treatment ment but were not assigned an interim outcome were not
27 with outcomes for MDR tuberculosis not included in this analysis. All patients who were classified
disaggregated for patients receiving
second-line drugs
as “transfer out” were added to the default category.
6 with outcomes not reported as success
(cure and completion), failure, default, or died Study characteristics
5 with outcomes not based on culture results
We examined differences in tuberculosis treatment
outcomes for studies using second-line regimens to
33 studies included determine which treatment programme characteristics
28 with individualised treatment regimen
4 with standardised treatment regimen
were associated with a higher proportion of patients
1 with both individualised and achieving treatment success. For each study, we
standardised treatment regimen gathered data on patients’ characteristics, treatment
protocols, and study definitions. Patients’ characteristics
Figure 1: Summary of literature search and study selection
included previous tuberculosis treatment history, HIV
MDR=multidrug resistant.
prevalence, and the mean number of drugs to
Search strategy which patients’ isolates were resistant. Data on treat-
To assess the effect of study characteristics on treatment ment protocols included the mean number of drugs in
success for MDR tuberculosis, we searched for English the regimen, duration of treatment, duration of DOT,
language publications in the MEDLINE database, the and whether the regimen was standardised or
Cochrane Library, EMBASE, CINAHL, and the ISI Web individualised. Studies in which all patients received
of Science with combinations of the keywords the same second-line regimen in addition to any
“tuberculosis”, “multi-drug resistant”, “MDR TB”, and susceptible first-line drugs after MDR tuberculosis
“treatment outcomes”. In addition, the online archives of diagnosis were categorised as having used a standardised
the International Journal of Tuberculosis and Lung Disease treatment approach. Studies that tailored treatment
were reviewed for applicable studies not found in the regimens to each individual patient’s drug-susceptibility
previous search. All database searches were updated in testing results were categorised as using an
December, 2008. We did not exclude any articles on the individualised treatment approach. We hypothesised
basis of publication date. that studies that used an individualised approach would
have a higher proportion of patients achieving treatment
Study selection success than reports of standardised treatment.15 In
The list of publications obtained through this search addition, we recorded length of follow-up and definition
was narrowed to studies considered relevant to our of cure if available.

154 www.thelancet.com/infection Vol 9 March 2009

 Review

Quantitative data synthesis
Pooled percentages for each tuberculosis outcome from
every study category were included in a Bayesian
random-effects meta-analysis.16 We chose a Bayesian
model because of the direct interpretation of credible
intervals of the posterior-effect estimates as belief that
the effect lies in that region. Given the large number of
variables and the small number of studies in our analysis,
we deemed that the correlation to belief would be useful
in the context of subjective clinical judgment in MDR
tuberculosis. The Bayesian model was repeated with five
initialisations using vague prior distributions, and
convergence was assessed using Brooks-Gelman
criteria.17,18 Data reported in each study were used to
calculate 95% credible intervals (CIs). Non-overlapping
CIs indicated significant differences between groups
with a 95% probability of a true difference.
Heterogeneity across studies was estimated by
calculating I2, an index of the proportion of total variation
across studies that is due to heterogeneity rather than to

Study location Years Sample Proportion HIV Length of Drugs in Average Directly observed Definition of cure§
size previously prevalence treatment* regimen†‡ resistance† treatment
treated (%) (%) (months) (n drugs) (n drugs)
Individualised treatment regimen
Goble et al19 USA 1973–1983 167 100% ·· 7¶ 4 ·· During hospitalisation Culture negative for 3 consecutive
months
Burgos et al20 USA 1982–2000 45 65% ·· 24 (11–58) 5·3 3 (2–9) Throughout treatment Culture negative, no clinical
evidence of tuberculosis
Olle-Goig and Bolivia 1983–1993 133 72% ·· 20 ·· 2·8** Self-administered 3 negative cultures, no clinical
Sandy21 evidence of tuberculosis
Geerlings et al22 Netherlands 1985–1998 44 34% 0 11 (9–18) 6 (4–9) 5 Self-administered 2 or more consecutive negative
cultures
Kim et al23 Korea 1988–1996 1011 100% ·· 18|| 5·3 3·7 Self-administered 2 or more negative cultures
Yew et al24 Hong Kong 1990–1997 71 65% 0 14 (9–24) 4·7 (3–6) 3·2** Throughout treatment Culture negative for at least
6 consecutive months
Chiang et al25 Taiwan 1992–1996 299 ·· ·· 18 3·7 (0–7) 3 (2–6) Throughout treatment 2 negative cultures, no clinical
evidence of tuberculosis
Munsiff et al26 USA 1992–1997 574 92% 74% 18 (1–83) 8 (2–15) 5 (2–10) 68% directly observed†† ≥12 months of treatment after the
last negative culture
Tahaoğlu et al9 Turkey 1992–1999 158 100% ·· 18§ 5·5 (3–9) 4·4 (2–9) During hospitalisation WHO guidelines
Narita et al27 USA 1994–1997 80 60% 53% 12§ 4·3 (1–8) 4·7 (2–11) 56% directly observed†† CDC guidelines
Kwon et al28 Korea 1995–2004 149 88% 0 24 (18–30) 6 (5–7) 5 (3–6) During hospitalisation Laserson definition
Mitnick et al4 Peru 1996–1998 75 100% ·· 23 6 (5–9) 6 (2–12) Throughout treatment 12 months of consecutive negative
cultures
Palmero et al29 Argentina 1996–1999 141 64% 0 18 4·2 (3–5) 4·1 (2–7) Self-administered Laserson definition
Kim et al30 South Korea 1996–2005 168 ·· 0 25 (1–136) 6 (3–11) ·· Not reported Laserson definition
Saravia et al15 Peru 1997–2001 52 100% ·· 18–24 5 4·4 Throughout treatment Culture negative for the last
12 months of treatment
Uffredi et al31 France 1998–1999 45 47% 20% 12 3·7 3·9 Self-administered Clinical improvement plus
1 negative culture
Nathanson Estonia, Latvia, 1999–2001 904 87% 1·8% 18–24 5·7 ·· Throughout treatment Laserson definition
et al32 Peru, Philippines
Mitnick et al33 Peru 1999–2002 603 ·· 1·5% 24·8 5·3 5·3 Throughout treatment Laserson definition
Tupasi et al34 Philippines 1999–2002 85 97% ·· 18|| 5 4·2 Throughout treatment Laserson definition
Holtz et al35 Latvia 2000 167 74% ·· 18|| 5·6 (4–8) 5 Throughout treatment Laserson definition
Leimane et al36 Latvia 2000 204 58% ·· 18–38 5·6 (3–8) 4 (2–7) Throughout treatment Culture negative for the last
12 months of treatment
Mirsaeidi et al37 Iran 2000–2002 17 100% 0 18·5 (7–36) 5·2 3·6 During hospitalisation WHO guidelines
Shin et al38 Russia 2000–2002 244 100% 0 18·5 (1–43) 5·8 4·7 (3–9) Throughout treatment Laserson definition
Kim et al39 Korea 2000–2002 1407 72% 1·5% ·· 5 (2–9) 4·2 26% directly observed†† Laserson definition
Keshavjee et al10 Russia 2000–2004 579 97% 0·9% 18|| 5·5 5·3 Throughout treatment Laserson definition
Bartu40 Czech Republic 2001–2004 40 53% 0 ·· 4·5 (0–6) 4·7 .. Culture negative, no clinical
evidence of tuberculosis
Riekstina et al41 Latvia 2002 75 0 ·· <24 ·· ·· Throughout treatment Laserson definition
Cox et al42 Uzbekistan 2003–2005 87 100% ·· 22 (18–30) 7 (5–10) 4·8 Throughout treatment Laserson definition
Eker et al43 Germany 2004–2006 146 53% 4·9% 18 (9–27) ·· 4·7 Throughout treatment Laserson definition
(Continues on next page)

www.thelancet.com/infection Vol 9 March 2009 155

 Review

Study location Years Sample Proportion HIV Length of Drugs in Average Directly observed Definition of cure§
size previously prevalence treatment* regimen†‡ resistance† treatment
treated (%) (%) (months) (n drugs) (n drugs)
(Continued from previous page)
Standardised treatment regimen
Van Deun et al7 Bangladesh 1997–1999 58 ·· ·· 21 7, 5, 2 3·2** During intensive phase Culture negative at end of
treatment and 2 previous occasions
Suárez et al44 Peru 1997–1999 466 100% ·· 18 5, 4 3·3 Self-administered Two negative culture results at the
end of treatment
Saravia et al15 Peru 1997–2001 39 100% ·· 18 5, 4 4·4 Throughout treatment Culture negative for the last
12 months of treatment
Park et al8 South Korea 1998–2000 126 100% ·· 18 5, 4, 3 4·5 (2–10) During hospitalisation Culture negative for 18 months
Masjedi et al45 Iran 2002–2006 43 100% 0 24 5, 4 ·· .. Laserson definition

*The median length of treatment and range is shown if reported. †For studies using standardised treatment regimens, the number of drugs in the regimen is given for the intensive and subsequent phases
separated by commas. ‡The mean number of drugs and range is shown if reported. §CDC guidelines (1994)46 specified cure as two negative cultures followed by 12 months of treatment with at least two
effective drugs; WHO MDR guidelines (1999)47 define “probable cure” as 6 months of negative smears and cultures, and “cure” as 18–24 months of negative cultures; Laserson standard definition48 defines cure as
at least five consecutive negative cultures during the last 12 months of treatment. ¶Only includes hospitalisation time; afterwards, the patient was discharged to the referring physician to continue treatment,
but was not followed in the study. ||After culture conversion. **Only did drug-sensitivity testing for first-line drugs. ††Remaining patients had self-administered treatment.

Table 1: Characteristics of patients, study design, and cure definition in 34 reports of multidrug-resistant (MDR) tuberculosis treatment outcomes, 1973–2006

chance. A p value of less than 0·05 indicated that
heterogeneity among the group of studies being analysed
was significant. The I2 statistic was greater than 50%
(with p<0·05) for each treatment outcome (treatment
success, failure, default, and death). Therefore, a
random-effects analysis, incorporating the impact of both
chance and heterogeneity among study populations and
study design, was chosen over the fixed-effects alternative,
which assumes that differences among study outcomes
are due entirely to chance.
To examine the impact of each study characteristic
individually, we did subgroup analyses by separating
studies based on each characteristic independently. We
thereby ranked the individual factors by the difference in
treatment success between reports that met the cut-off
compared with all other reports. We then examined the
differences in the proportion of patients achieving
treatment success among studies that met the cut-offs
for the two most important factors compared with those
that did not. If those studies did not have a significantly
different proportion of patients achieving treatment
success, we looked at studies that met cut-offs for the
three most important factors. Statistical analyses were
done with WinBUGS and R.
Results
565 publications were obtained through the literature
search. We narrowed these to 253 studies deemed
relevant to this analysis, of which 76 were pursued for
full analysis. After adding studies from references and
reviews, 95 articles were reviewed. Of these, 33 studies
(34 published reports) met the inclusion criteria,
including 28 studies of individualised therapy for MDR
tuberculosis, four studies of standardised therapy, and
one study that had an individualised group and a
standardised group (figure 1). These reports came from
20 countries, including treatment outcomes for a mean
of 250 patients per report (range 17–1407). All of the
studies were retrospective observational cohort studies of
treatment outcomes for MDR tuberculosis patients who
were receiving second-line drugs. These included
29 studies done from 1973 to 2006 that used individualised
treatment regimens and five reports from 1997 to 2006
that used standardised second-line regimens (table 1).
Heterogeneity among studies
Characteristics of the patient population, study design, and
cure definition differed among the included studies, and
reporting procedures varied widely (table 1). The number
of patients previously treated for tuberculosis ranged from
0 to 100% in 30 of the 34 reviewed reports; other studies
omitted this statistic. HIV prevalence ranged from 0 to
74% in the 16 studies in which it was reported. The median
duration of treatment ranged from 11 months to 25 months.
The duration of treatment for individual patients varied
widely, with some patients receiving as long as 136 months
of therapy based on their culture results and clinical
indications.30 21 studies reported the length of follow-up
after treatment, which often varied between patients within
the same study. 16 studies included DOT throughout
treatment, nine maintained DOT during the intensive
phase of therapy or directly observed some but not
all patients, six studies used self-administered therapy, and
three studies did not report the degree to which treatment
was observed. The mean number of drugs used ranged
from 3·7 to 8 (median 5·5). The mean size of drug
regimens was equal between studies of individualised and
standardised therapy, although the range among
standardised studies was smaller (range 5–7). 13 of the
29 reports of individualised regimens and all five reports of
standardised regimens listed which specific drugs or drug
classes were given. Four additional studies reported the

156 www.thelancet.com/infection Vol 9 March 2009

 Review

proportion of patients receiving a fluoroquinolone. The
mean number of drugs to which isolated organisms were
resistant ranged from 2·8 to 6.
The definition of cure varied among studies. 14 studies
followed the current standard definition, which requires
at least five negative cultures during the last year of
treatment.48 Seven additional studies required that the
patient be culture negative for at least 12 months, and
one study mandated 6 months; however, the number of
actual cultures taken was not defined for these five
studies. In 12 studies, three or fewer negative cultures
were necessary to be classified as “cured”.
Treatment outcomes
Across all studies that used second-line drugs in
individualised or standardised protocols, the overall
treatment success estimate, defined as the proportion
of patients who were cured or completed treatment, was
62% (95% CI 58–67%). However, the heterogeneity in
study characteristics led to significant variation in
reported treatment outcomes. Among the 29 reports of
individualised treatment regimens for patients with
MDR tuberculosis (mean 268 patients; range 17–1407),
the mean proportion of patients achieving treatment
success was 64% (95% CI 59–68%; figure 2). The mean
proportion of patients whose outcomes were treatment
failure, default, and death were 6%, 12%, and 11%,
respectively. Among the five studies in which patients
with MDR tuberculosis were treated with a standardised
regimen (mean 146 patients; range 39–466), the mean
proportion achieving treatment success was lower than
that for individualised treatment, but not significantly

Success Failure Default Death

(95% CI) (%) (%) (%)

Individualised treatment
Goble et al19 48 (41–55) 14 13 20
Burgos et al20 67 (55–78) 5 5 22
Olle-Goig and Sandy21 36 (28–44) 1 52 13
Geerlings et al22 72 (60–82) 10 4 13
Kim et al23 48 (44–51) 8 43 1
Yew et al24 68 (58–77) 14 9 8
Chiang et al25 52 (46–57) 10 9 29
Munsiff et al26 40 (36–44) 0 13 47
Tahaoğlu et al9 76 (69–82) 8 11 4
Narita et al27 67 (57–76) 12 3 20
Kwon et al28 68 (61–75) 14 10 7
Mitnick et al4 79 (70–87) 2 9 8
Palmero et al29 53 (45–61) 9 19 19
Kim et al30 65 (58–72) 19 8 8
Saravia et al15 75 (64–84) 4 9 10
Uffredi et al31 59 (47–71) 2 15 24
Nathanson et al32 68 (64–71) 7 11 14
Mitnick et al33 66 (62–70) 2 10 20
Tupasi et al34 67 (58–76) 9 12 11
Holtz et al35 62 (55–69) 16 11 8
Leimane et al36 66 (60–72) 14 13 7
Mirsaeidi et al37 64 (47–79) 23 6 9
Shin et al38 76 (70–81) 7 12 5
Kim et al39 45 (43–48) 5 40 7
Keshavjee et al10 67 (63–70) 8 20 5
Bartu40 68 (55–79) 2 5 25
Riekstina et al41 71 (62–80) 6 15 6
Cox et al42 62 (53–71) 9 14 15
Eker et al43 71 (64–78) 2 17 10

Summary 64 (59–68) 6 (3–9) 12 (8–16) 11 (7–15)

I2=88 I2=94 I2=99 I2=98
(p<0·001) (p<0·001) (p<0·001) (p<0·001)

Standardised treatment
Van Deun et al7 63 (50–76) 8 12 12
Suárez et al44 49 (44–53) 28 12 12
Saravia et al15 51 (43–59) 30 13 11
Park et al8 49 (35–60) 15 32 6
Masjedi et al45 61 (49–75) 15 3 15

Summary 54 (43–68) 18 (7–35) 12 (1–36) 11 (5–19)

I2=52 I2=89 I2=91 I2=62
(p=0·08) (p<0·001) (p<0·001) (p=0·03)
0 25 50 75 100
Treatment success (%)

Figure 2: Treatment success and other treatment outcomes for individualised and standardised treatment of multidrug-resistant tuberculosis
Treatment effects and summaries were calculated using a Bayesian random effects model weighted by study population size. 95% credible intervals (CIs) are shown.
Saravia et al15 had an individualised arm and a standardised arm, which are reported separately.

www.thelancet.com/infection Vol 9 March 2009 157

 Review

length of treatment, level of direct observation of therapy,

Reports Treatment success Heterogeneity
(n)* (95% credible interval) regimen design, number of drugs in the regimen,
percentage of patients receiving fluoroquinolones, start
p I²
year of the study, or cure definition (table 2). Differences
Length of treatment in population characteristics, including prevalence of
>18 months4,7,9,10,15,20,21,23,28,30,32–36,38,41–42,45† 20 66% (61–72) <0·001 79% HIV, mean number of resistant drugs, and proportion
≤18 months or not reported8,19,22,24–27,29,31,37,39–40,43–44 14 56% (49–63) <0·001 83% of patients previously treated for tuberculosis, also did
Directly observed treatment not lead to significantly different outcomes.
Throughout therapy4,10,15,20,24,25,32–36,38,41–43 16 67% (62–72) 0·01 51% Although the proportion of patients achieving
Partial observation‡, self-administered, or not 18 57% (51–64) <0·001 88% treatment success was not significantly different on the
reported7–9,21–23,26–31,37,39,40,44,45
basis of any individual factor, studies that combined the
Regimen design
two factors with the largest effect on success (treatment
Individualised4,9,10,15,19–43 29 64% (59–68) <0·001 88%
length of at least 18 months and use of DOT throughout)
Standardised7,8,15,44,45 5 54% (43–68) 0·08 52% had a pooled success proportion that surpassed all other
HIV prevalence subgroups. In the 12 studies that incorporated both
0%22,24,28–30,37,38,40,45 9 68% (61–74) 0·09 36% factors into their study methods, the pooled success was
Not reported4,7–9,15,19–21,23,25,34–36,41,42,44 17 61% (53–68) <0·001 82% 69% (95% CI 64–73%), which was significantly greater
>0%10,26,27,31–33,39,43 8 59% (49–69) <0·001 93% than the pooled success estimate for the other 22 studies
Drugs in regimen§ that did not meet both criteria (58%; 95% CI 52–64%;
>54,7,9,10,20,22,23,26,28,30,32,33,35–38,42 17 66% (60–71) <0·001 87% figure 3).
≤5 or not reported8,15,19,21,24,25,27,29,31,34,39–41,43-45 17 58% (51–65) <0·001 82%
Average resistance§ Discussion
≥4·4 drugs4,8–10,15,22,26–28,33,35,38,40,42,43 16 66% (59–73) <0·001 83% To determine which patient and programme
<4·4 drugs or not reported7,19–21,23–25,29–32,34,36,37,39,41,44,45 18 58% (52–65) <0·001 85% characteristics facilitate the greatest treatment success,
Start year of study we analysed data from 33 studies in 20 countries that
1998 or later8,10,31–43,45 16 65% (60–70) <0·001 85% included treatment outcomes for a total of 8506 patients
1997 or earlier4,7,9,15,19–30,44 18 60% (52–68) <0·001 87% receiving second-line drug treatment for MDR
Definition of cure tuberculosis. Although the proportion of patients
≥12 months of negative cultures 21 64% (58–69) <0·001 88% achieving treatment success was better in studies that
required4,8,10,15,24,28–30,32–39,41–43,45 used individualised treatment regimens, the difference
<12 months of negative cultures required or not 13 60% (52–68) <0·001 82% was not significant. In fact, no individual patient or
reported7,9,19–23,25–27,31,40,44
programme characteristic was associated with a
Use of fluoroquinolone significantly greater proportion of patients achieving
≥50% of patients4,7–10,15,20,26,29–31,34–38,39,41,42,44,45 21 62% (57–68) <0·001 87% treatment success. Studies that incorporated both
<50% of patients or not reported15,19,21–25,27,28,32,33,40,43 13 62% (53–71) <0·001 85% treatment for longer than 18 months and DOT throughout
Previous tuberculosis treatment the entire treatment period had a significantly greater
<75% or not reported7,20,22,24,25,27,29–31,33,35,36,39–41,43 16 62% (56–68) <0·001 84% proportion of patients achieving treatment success than
≥75%4,8–10,15,19,21,23,26,28,32,34,37,38,42,44,45 18 62% (55–69) <0·001 88% all other studies. When treatment programmes included
*All 34 reports were included and classified by cohort factor. †Reports were classified on the bases of median length of
both of these factors, 69% (95% CI 64–73%) of patients
treatment. All reports in which the duration of treatment was 18 months after culture conversion were classified in the were successfully treated. Only 58% (95% CI 52–64%)
>18 months group. ‡Partial observation includes studies that used directly observed therapy during the intensive or were successfully treated when a maximum of one of the
hospitalisation phase only, or in which some, but not all, patients received directly observed therapy throughout
treatment. §Reports were classified on the basis of the mean number of drugs.
two factors was used. However, substantial heterogeneity
in study designs, patient populations, and reporting limit
Table 2: Pooled treatment success among subgroups of studies using individualised or standardised the scope of this analysis. Prospective trials may be
treatment regimens for multidrug-resistant (MDR) tuberculosis necessary to elucidate specific programme components
that promote MDR tuberculosis treatment success.
different at 54% (95% CI 43–68%; figure 2). The mean All of the 12 studies associated with a higher proportion
proportion whose outcomes were failure, default of patients achieving success that featured extended
(treatment interruption), and death among the duration of treatment and DOT throughout therapy used
standardised regimen studies were 18%, 12%, and 11%, individualised treatment regimens. However,
respectively. standardised second-line therapy programmes may also
We combined studies of individualised and standardised benefit from simultaneously implementing extended
treatment regimens to analyse the effect of each of the treatment duration together with DOT throughout
other study characteristics independently (table 2). Our treatment. Among the five reports of standardised
analysis showed that the proportion of patients treated treatment in this analysis, the two studies with the
successfully did not differ significantly on the basis of highest proportion of patients treated successfully both
any of the following individual study characteristics: treated patients for a duration exceeding 18 months.7,45

158 www.thelancet.com/infection Vol 9 March 2009

 Review

Success Failure Default Death

(95% CI) (%) (%) (%)

Treatment duration >18 months

and DOT throughout
Burgos et al20 69 (59–69) 6 10 21
Mitnik et al4 76 (68–84) 4 11 8
Nathanson et al32 68 (65–71) 7 11 14
Leimane et al36 67 (61–72) 13 13 7
Shin et al38 75 (70–80) 7 12 1
Riekstina et al41 71 (63–79) 6 14 6
Keshavjee et al10 59 (55–63) 8 19 5
Cox et al42 65 (56–72) 9 13 14
Holtz et al35 64 (57–70) 15 11 8
Mitnik et al33 67 (63–70) 3 11 20
Saravia et al15 73 (64–82) 5 11 10
Tupasi et al34 68 (60–76) 9 12 10

Summary 69 (64–73) 7 (4–10) 12 (9–14) 9 (5–14)

I2=36 I2=84 I2=72 I2=94
(p=0·10) (p<0·001) (p<0·001) (p<0·001)

Treatment duraton ≤18 months or

not reported, or incomplete DOT
Olle-Goig and Sandy21 35 (27–43) 1 53 13
Masjedi et al45 65 (52–77) 13 3 18
Van Deun et al7 67 (55–77) 5 12 14
Kwon et al28 67 (60–75) 15 10 7
Kim et al30 64 (57–71) 19 8 8
Kim et al23 48 (45–51) 8 43 1
Tahaoglu et al9 75 (69–81) 8 11 4
Goble et al19 48 (40–55) 14 13 21
Geerlings et al22 70 (58–81) 11 3 13
Yew et al24 67 (57–77) 14 8 8
Chiang et al25 52 (46–57) 10 9 29
Musiff et al26 40 (36–44) 0 13 47
Narita et al27 66 (56–75) 12 2 20
Palmero et al29 52 (45–60) 9 19 19
Saravia et al15 48 (44–53) 28 12 12
Park et al8 51 (43–59) 14 32 4
Kim et al39 45 (43–48) 5 40 7
Bartu40 66 (54–78) 2 4 25
Eker et al43 70 (63–77) 1 18 10
Saravia et al15 48 (34–61) 31 13 11
Uffredi et al31 58 (45–70) 2 15 25
Mirsaeidi et al37 61 (44–77) 26 5 8

Summary 58 (52–64) 7 (3–13) 11 (6–18) 11 (7–17)

I2=86 I2=96 I2=99 I2=98
(p<0·001) (p<0·001) (p<0·001) (p<0·001)
0 25 50 75 100
Treatment success (%)

Figure 3: Treatment success and other treatment outcomes by treatment duration and whether study included directly observed treatment regimens
Treatment effects and summaries were calculated using a Bayesian random effects model weighted by study population size. Individual studies can have different
treatment effects in figure 2 and figure 3 caused by differences in the other studies being analysed simultaneously. 95% credible intervals (CIs) are shown.
Saravia et al15 had an individualised arm and a standardised arm, which are reported separately.

WHO guidelines for the treatment of MDR
tuberculosis encourage a minimum of 18 months of
treatment after culture conversion and DOT throughout
treatment.49 Our meta-analysis underscores the
importance of these particular recommendations and
emphasises their use in combination. Previous reviews
have discussed the use of individualised therapy and the
number and choice of drugs in the regimen,12,13 but have
not focused on the duration of treatment or DOT
throughout therapy. A longer duration of treatment has
been associated with lower likelihood of early relapse
for drug-susceptible tuberculosis.50 Similarly, DOT has
been shown to promote treatment success in non-MDR
tuberculosis,51 although not all studies show clinical
improvement,52 and increased population coverage of
DOT programmes has been associated with countrywide
treatment success.53
Substantial heterogeneity in study characteristics
prevents a more conclusive determination of what
factors have the most effect on the proportion of patients
that achieve treatment success and limits the validity of
this analysis. We found that treatment protocols and
reporting of key patient and study characteristics were
inconsistent across studies. Several studies omitted the
exact protocol used in treatment regimen design, length
of follow-up, and the average number of drugs to which
each patient’s tuberculosis isolate was resistant.
HIV infection, which has been associated with poor

www.thelancet.com/infection Vol 9 March 2009 159

 Review
outcomes among patients receiving treatment for MDR
Search strategy and extensively drug-resistant tuberculosis,54,55 was
and selection assessed in fewer than half of the studies reviewed.
criteria This absence despite the confluence of the HIV and
These are described tuberculosis epidemics exemplifies the heterogeneity
in detail in the in tuberculosis epidemics and associated treatment
Methods section. strategies. Similarly, fluoroquinolone use was only
reported in 13 of 34 studies, despite its association
within studies with treatment success for MDR
tuberculosis.24,25 Thus, the lack of significant findings
associated with certain variables in this analysis may be
due to reporting insufficiency, rather than the absence
of a real association. Furthermore, these variations in
the recording of data necessary for assessing treatment
outcomes underscore the need for standardised data
collection and reporting in programmes and studies of
MDR tuberculosis,48 possibly through the use of an
international registry of treatment outcomes. The
absence of a registry of individual treatment outcomes
or outcome studies precluded the possibility of
estimating publication bias in this meta-analysis.
Standards in outcome reporting are particularly
important in light of the emergence of extensively
drug-resistant tuberculosis.
Many programme characteristics that were reported in
most studies also varied widely. Even within studies,
different patients had highly variable treatment
durations,20,26,30,38 and the size of drug regimens ranged
from very few to very many drugs.9,26,27,30,39,42 In addition,
factors that cannot be measured or were not reported
may have confounded the results of this study. Issues
such as logistics, programme resources, transportation,
food assistance, and social support that are accounted for
in different ways by different programmes but often not
reported in published studies may have affected the
proportion achieving treatment success independently of
factors analysed in this study. Prospective trials comparing
specific aspects of treatment recommendations for MDR
tuberculosis in similar populations would provide greater
insight into improving MDR tuberculosis treatment
protocols. For example, one prospective trial suggests
that addition of high-dose isoniazid as adjuvant therapy
to second-line regimens may improve MDR tuberculosis
treatment.56
The “Stop TB” strategy developed by WHO set the goal
of curing 85% of all detected tuberculosis cases by 2005.57
MDR tuberculosis has presented challenges to achieving
this objective in many areas. Nonetheless, appropriate
second-line treatment for more than 18 months together
with DOT throughout therapy might lead to treatment
success more readily, and should be further investigated
in current trials of MDR tuberculosis therapy. In the
absence of randomised clinical trials, more systematic
documentation of programmatic components and
outcomes of MDR tuberculosis treatment can strengthen
the evidence base for treatment. Comprehensive
treatment of MDR tuberculosis is of vital importance in
160
promoting public health to slow the spread and reduce
the impact of drug-resistant tuberculosis around the
world.
Conflicts of interest
We declare that we have no conflicts of interest.
Acknowledgments
The work of EO and AP was supported by the Notsew Orm Sands
Foundation and the Miriam Burnett foundation. The work of SB was
supported by the US CDC (R36) and NIH (T32). Funding sources played
no role in the design or analysis of this study.
References
1 Wright A, Zignol M. Anti-tuberculosis drug resistance in the world:
fourth global report. Geneva: WHO, 2008.
2 Nathanson E, Gupta R, Huamani P, et al. Adverse events in the
treatment of multidrug-resistant tuberculosis: results from the
DOTS-Plus initiative. Int J Tuberc Lung Dis 2004; 8: 1382–84.
3 Yee D, Valiquette C, Pelletier M, Parisien I, Rocher I, Menzies D.
Incidence of serious side effects from first-line antituberculosis
drugs among patients treated for active tuberculosis.
Am J Respir Crit Care Med 2003; 167: 1472–77.
4 Mitnick C, Bayona J, Palacios E, et al. Community-based therapy for
multidrug-resistant tuberculosis in Lima, Peru. N Engl J Med 2003;
348: 119–28.
5 Resch SC, Salomon JA, Murray M, Weinstein MC. Cost-effectiveness
of treating multidrug-resistant tuberculosis. PLoS Med 2006; 3: e241.
6 WHO. Guidelines for the programmatic management of
drug-resistant tuberculosis. Emergency update 2008. Geneva:
WHO, 2008.
7 Van Deun A, Salim MA, Das AP, Bastian I, Portaels F. Results of
a standardised regimen for multidrug-resistant tuberculosis in
Bangladesh. Int J Tuberc Lung Dis 2004; 8: 560–67.
8 Park SK, Lee WC, Lee DH, Mitnick CD, Han L, Seung KJ.
Self-administered, standardized regimens for multidrug-resistant
tuberculosis in South Korea. Int J Tuberc Lung Dis 2004; 8: 361–68.
9 Tahaoğlu K, Törün T, Sevim T, et al. The treatment of multidrug-
resistant tuberculosis in Turkey. N Engl J Med 2001; 345: 170–74.
10 Keshavjee S, Gelmanova IY, Farmer PE, et al. Treatment of
extensively drug-resistant tuberculosis in Tomsk, Russia:
a retrospective cohort study. Lancet 2008; 372: 1403–09.
11 Lew W, Pai M, Oxlade O, Martin D, Menzies D. Initial drug
resistance and tuberculosis treatment outcomes: systematic review
and meta-analysis. Ann Intern Med 2008; 149: 123–34.
12 Caminero JA. Treatment of multidrug-resistant tuberculosis:
evidence and controversies. Int J Tuberc Lung Dis 2006; 10: 829–37.
13 Mukherjee JS, Rich ML, Socci AR, et al. Programmes and principles
in treatment of multidrug-resistant tuberculosis. Lancet 2004;
363: 474–81.
14 Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D, Stroup DF.
Improving the quality of reports of meta-analyses of randomised
controlled trials: the QUOROM statement. Lancet 1999;
354: 1896–900.
15 Saravia JC, Appleton SC, Rich ML, Sarria M, Bayona J, Becerra MC.
Retreatment management strategies when first-line tuberculosis
therapy fails. Int J Tuberc Lung Dis 2005; 9: 421–29.
16 Smith TC, Spiegelhalter DJ, Thomas A. Bayesian approaches to
random-effects meta-analysis: a comparative study. Stat Med 1995;
14: 2685–99.
17 Gelman A, Carlin JB, Stern HS, Rubin DB. Bayesian data analysis.
Boca Raton, FL: CRC Press, 2003.
18 Brooks S, Gelman A. General methods for monitoring convergence
of iterative simulations. J Comput Graph Stat 1998; 7: 434–55.
19 Goble M, Iseman MD, Madsen LA, Waite D, Ackerson L,
Horsburgh CR Jr. Treatment of 171 patients with pulmonary
tuberculosis resistant to isoniazid and rifampin. N Engl J Med 1993;
328: 527–32.
20 Burgos M, Gonzalez LC, Paz EA, et al. Treatment of
multidrug-resistant tuberculosis in San Francisco: an
outpatient-based approach. Clin Infect Dis 2005; 40: 968–75.
21 Olle-Goig JE, Sandy R. Outcomes of individualised treatment for
multidrug-resistant tuberculosis before DOTS-plus.
Int J Tuberc Lung Dis 2005; 9: 765–70.
www.thelancet.com/infection Vol 9 March 2009

22 Geerligs WA, Van Altena R, De Lange WCM, Van Soolingen D,
Van Der Werf TS. Multidrug-resistant tuberculosis: long-term
treatment outcome in the Netherlands. Int J Tuberc Lung Dis 2000;
4: 758–64.
23 Kim HJ, Hong YP, Kim SJ, Lew WJ, Lee EG. Ambulatory treatment
of multidrug-resistant pulmonary tuberculosis patients at a chest
clinic. Int J Tuberc Lung Dis 2001; 5: 1129–36.
24 Yew WW, Chan CK, Chau CH, et al. Outcomes of patients with
multidrug-resistant pulmonary tuberculosis treated with
ofloxacin/levofloxacin-containing regimens. Chest 2000; 117: 744–51.
25 Chiang CY, Enarson DA, Yu MC, et al. Outcome of pulmonary
multidrug-resistant tuberculosis: a 6-yr follow-up study. Eur Respir J
2006; 28: 980–85.
26 Munsiff SS, Ahuja SD, Li J, Driver CR. Public-private collaboration
for multidrug-resistant tuberculosis control in New York City.
Int J Tuberc Lung Dis 2006; 10: 639–48.
27 Narita M, Alonso P, Lauzardo M, Hollender ES, Pitchenik AE,
Ashkin D. Treatment experience of multidrug-resistant tuberculosis
in Florida, 1994–1997. Chest 2001; 120: 343–48.
28 Kwon YS, Kim YH, Suh GY, et al. Treatment outcomes for
HIV-uninfected patients with multidrug-resistant and extensively
drug-resistant tuberculosis. Clin Infect Dis 2008; 47: 496–502.
29 Palmero DJ, Ambroggi M, Brea A, et al. Treatment and follow-up of
HIV-negative multidrug-resistant tuberculosis patients in an
infectious diseases reference hospital, Buenos Aires, Argentina.
Int J Tuberc Lung Dis 2004; 8: 778–84.
30 Kim HR, Hwang SS, Kim HJ, et al. Impact of extensive drug
resistance on treatment outcomes in non-HIV-infected patients
with multidrug-resistant tuberculosis. Clin Infect Dis 2007;
45: 1290–95.
31 Uffredi ML, Truffot-Pernot C, Dautzenberg B, Renard M, Jarlier V,
Robert J. An intervention programme for the management of
multidrug-resistant tuberculosis in France. Int J Antimicrob Agents
2007; 29: 434–39.
32 Nathanson E, Lambregts-van Weezenbeek C, Rich ML, et al.
Multidrug-resistant tuberculosis management in resource-limited
settings. Emerg Infect Dis 2006; 12: 1389–97.
33 Mitnick CD, Shin SS, Seung KJ, et al. Comprehensive treatment of
extensively drug-resistant tuberculosis. N Engl J Med 2008;
359: 563–74.
34 Tupasi TE, Quelapio MI, Orillaza RB, et al. DOTS-Plus for
multidrug-resistant tuberculosis in the Philippines: global
assistance urgently needed. Tuberculosis (Edinb) 2003; 83: 52–58.
35 Holtz TH, Sternberg M, Kammerer S, et al. Time to sputum culture
conversion in multidrug-resistant tuberculosis: predictors and
relationship to treatment outcome. Ann Intern Med 2006;
144: 650–59.
36 Leimane V, Riekstina V, Holtz TH, et al. Clinical outcome of
individualised treatment of multidrug-resistant tuberculosis in
Latvia: a retrospective cohort study. Lancet 2005; 365: 318–26.
37 Mirsaeidi SM, Tabarsi P, Khoshnood K, et al. Treatment of multiple
drug-resistant tuberculosis (MDR-TB) in Iran. Int J Infect Dis 2005;
9: 317–22.
38 Shin SS, Pasechnikov AD, Gelmanova IY, et al. Adverse reactions
among patients being treated for MDR-TB in Tomsk, Russia.
Int J Tuberc Lung Dis 2007; 11: 1314–20.
39 Kim DH, Kim HJ, Park SK, et al. Treatment outcomes and
long-term survival in patients with extensively drug-resistant
tuberculosis. Am J Respir Crit Care Med 2008; 178: 1075–82.
40 Bartu V. Multidrug-resistant tuberculosis in the Czech Republic:
strategy and therapeutic outcomes. Eur J Clin Microbiol Infect Dis
2007; 26: 603–05.
www.thelancet.com/infection Vol 9 March 2009
Review
41 Riekstina V, Leimane V, Holtz TH, Leimans J, Wells CD. Treatment
outcome cohort analysis in an integrated DOTS and DOTS-Plus TB
program in Latvia. Int J Tuberc Lung Dis 2007; 11: 585–87.
42 Cox HS, Kalon S, Allamuratova S, et al. Multidrug-resistant
tuberculosis treatment outcomes in Karakalpakstan, Uzbekistan:
treatment complexity and XDR-TB among treatment failures.
PLoS ONE 2007; 2: e1126.
43 Eker B, Ortmann J, Migliori GB, et al. Multidrug- and extensively
drug-resistant tuberculosis, Germany. Emerg Infect Dis 2008;
14: 1700–06.
44 Suárez PG, Floyd K, Portocarrero J, et al. Feasibility and
cost-effectiveness of standardised second-line drug treatment for
chronic tuberculosis patients: a national cohort study in Peru.
Lancet 2002; 359: 1980–89.
45 Masjedi MR, Tabarsi P, Chitsaz E, et al. Outcome of treatment of
MDR-TB patients with standardised regimens, Iran, 2002–2006.
Int J Tuberc Lung Dis 2008; 12: 750–55.
46 Centers for Disease Control and Prevention. Treatment of
tuberculosis and tuberculosis infection in adults and children.
Am J Respir Crit Care Med 1994; 149: 1359–74.
47 Farmer P, Kim JY, Mitnick CD, Becerra M. Protocol for the
implementation of individualized treatment regimens for
multidrug-resistant tuberculosis in resource-poor settings. In:
Espinal MA, ed. Multidrug resistant tuberculosis (MDR TB): basis
for the development of an evidence-based case-management
strategy for MDR TB within the World Health Organization’s
DOTS strategy. Geneva: WHO, 1999.
48 Laserson KF, Thorpe LE, Leimane V, et al. Speaking the same
language: treatment outcome definitions for multidrug-resistant
tuberculosis. Int J Tuberc Lung Dis 2005; 9: 640–45.
49 WHO. Guidelines for the programmatic management of
drug-resistant tuberculosis. Geneva: WHO, 2006.
50 Chang KC, Leung CC, Yew WW, Ho SC, Tam CM. A nested
case-control study on treatment-related risk factors for early relapse
of tuberculosis. Am J Respir Crit Care Med 2004; 170: 1124–30.
51 Frieden TR, Sbarbaro JA. Promoting adherence to treatment for
tuberculosis: the importance of direct observation.
Bull World Health Organ 2007; 85: 407–09.
52 Volmink J, Garner P. Directly observed therapy for treating
tuberculosis. Cochrane Database Syst Rev 2007; 4: CD003343.
53 Obermeyer Z, Abbott-Klafter J, Murray CJ. Has the DOTS strategy
improved case finding or treatment success? An empirical
assessment. PLoS ONE 2008; 3: e1721.
54 Wells CD, Cegielski JP, Nelson LJ, et al. HIV infection and
multidrug-resistant tuberculosis: the perfect storm. J Infect Dis
2007; 196 (suppl 1): S86–107.
55 Gandhi NR, Moll A, Sturm AW, et al. Extensively drug-resistant
tuberculosis as a cause of death in patients co-infected with
tuberculosis and HIV in a rural area of South Africa. Lancet 2006;
368: 1575–80.
56 Katiyar SK, Bihari S, Prakash S, Mamtani M, Kulkarni H.
A randomised controlled trial of high-dose isoniazid adjuvant
therapy for multidrug-resistant tuberculosis. Int J Tuberc Lung Dis
2008; 12: 139–45.
57 WHO. The Stop TB strategy: building on and enhancing DOTS to
meet the TB-related Millenium Development Goals. Geneva: WHO,
2006.
161