| Background. Multi-drug resistant tuberculosis (MDR-TB) is a major threat to global tuberculosis control. While observational studies have reported outcomes of MDR-TB treatment, there have been no randomized controlled trials for MDR-TB treatment outcomes. We did a systematic review and meta-analysis to examine individual and study-level factors associated with treatment outcomes for MDR-TB in the observational studies.;Results. After screening 2187 titles and abstracts, 265 articles were identified for retrieval and full-text review, and of these, 72 articles met the inclusion criteria and were included in the meta-analysis. Data analysis was performed using the 64 unique cohorts reported by the 72 articles. The cohorts were quite heterogeneous in characteristics and outcomes. The mean size of the cohorts was 124 patients (range 25 to 1011). The mean age of participants in the cohorts was 39 years with females accounting for about one third. The median length of treatment was 18 months, and the average number of drugs in the regimen was five. The overall pooled rates of cumulative success (successful patients who did not relapse) was 50%, of cumulative failure (failure plus relapse) was 17%, of death was 13% and of default was 18%. These pooled outcome rates, however, must be interpreted with caution because of heterogeneity across studies. Subgroup and meta-regression analyses helped identify several factors associated with improved outcomes. Factors significantly associated with increased treatment success are treatment duration longer than 20 months, use of more than three sensitive drugs, individualized regimen, use of fluoroquinolones, or use of second-line agents in general. Factors that were significantly associated with high treatment mortality were high prevalence of HIV co-infection and use of three or fewer drugs. Low default rate was most strongly associated with shorter treatments and directly observed therapy. Use of second-line drugs was significantly associated with higher default rate. Considerable heterogeneity remained even within subgroups.;Conclusion. Outcomes of MDR-TB appear to vary considerably across studies and populations. The heterogeneity among studies poses a challenge in interpreting the results of this meta-analysis for clinical care, underscoring the need for future research to clarify optimal treatment of MDR-TB.;Method. We searched MEDLINE, EMBASE, BIOSIS, Web of Science from 1970 to July 2008, for publications in any language that described at least one treatment outcome among at least 25 patients with microbiologically proven MDR-TB. Data were extracted and where missing, principle investigators were contacted for more information. Rates of treatment outcomes were pooled using random effects. Subgroup analyses and meta-regression models were used to explore sources of heterogeneity. |
