Background An estimated 32,000 children develop multidrug-resistant tuberculosis (MDR-TB; Mycobacterium tuberculosis resistant to isoniazid and rifampin) each year. Little is known about the optimal treatment for these children. Methods and findings To inform the pediatric aspects of the revised World Health Organization (WHO) MDR-TB treatment guidelines, we performed a systematic review and individual patient data (IPD) meta-analysis, describing treatment outcomes in children treated for MDR-TB. To identify eligible reports we searched PubMed, LILACS, Embase, The Cochrane Library, PsychINFO, and BioMedCentral databases through 1 October 2014. To identify unpublished data, we reviewed conference abstracts, contacted experts in the field, and requested data through other routes, including at national and international conferences and through organizations working in pediatric MDR-TB. A cohort was eligible for inclusion if it included a minimum of three children (aged <15 years) who were treated for bacteriologically confirmed or clinically diagnosed MDR-TB, and if treatment outcomes were reported. The search yielded 2,772 reports; after review, 33 studies were eligible for inclusion, with IPD provided for 28 of these. All data were from published or unpublished observational cohorts. We analyzed demographic, clinical, and treatment factors as predictors of treatment outcome. In order to obtain adjusted estimates, we used a random-effects multivariable logistic regression (random intercept and random slope, unless specified otherwise) adjusted for the following covariates: age, sex, HIV infection, malnutrition, severe extrapulmonary disease, or the presence of severe disease on chest radiograph. We analyzed data from 975 children from 18 countries; 731 (75%) had bacteriologically confirmed and 244 (25%) had clinically diagnosed MDR-TB. The median age was 7.1 years. Of 910 (93%) children with documented HIV status, 359 (39%) were infected with HIV. When compared to clinically diagnosed patients, children with confirmed MDR-TB were more likely to be older, to be infected with HIV, to be malnourished, and to have severe tuberculosis (TB) on chest radiograph (p < 0.001 for all characteristics). Overall, 764 of 975 (78%) had a successful treatment outcome at the conclusion of therapy: 548/731 (75%) of confirmed and 216/244 (89%) of clinically diagnosed children (absolute difference 14%, 95% confidence interval [CI] 8%–19%, p < 0.001). Treatment was successful in only 56% of children with bacteriologically confirmed TB who were infected with HIV who did not receive any antiretroviral treatment (ART) during MDR-TB therapy, compared to 82% in children infected with HIV who received ART during MDR-TB therapy (absolute difference 26%, 95% CI 5%–48%, p = 0.006). In children with confirmed MDR-TB, the use of second-line injectable agents and high-dose isoniazid (15–20 mg/kg/day) were associated with treatment success (adjusted odds ratio [aOR] 2.9, 95% CI 1.0–8.3, p = 0.041 and aOR 5.9, 95% CI 1.7–20.5, p = 0.007, respectively). These findings for high-dose isoniazid may have been affected by site effect, as the majority of patients came from Cape Town. Limitations of this study include the difficulty of estimating the treatment effects of individual drugs within multidrug regimens, only observational cohort studies were available for inclusion, and treatment decisions were based on the clinician’s perception of illness, with resulting potential for bias. Conclusions This study suggests that children respond favorably to MDR-TB treatment. The low success rate in children infected with HIV who did not receive ART during their MDR-TB treatment highlights the need for ART in these children. Our findings of individual drug effects on treatment outcome should be further evaluated.
