This Prospero registered study (ID: CRD42018095600) followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.
Data sources
We searched Medline (EbscoHost), Web of Science Core collections, CINAHL (EbscoHost) and Cochrane library for articles published between January 2004 (2004 was when roll out of ART across SSA started) and May 2018. The search strategy was modified from those used by Samuels et al. [19] and Isaakidis et al. [7]. The following Medical Subject Headings and free text terms were used; Tuberculosis, Multidrug-Resistant Tuberculosis, MDR-TB, HIV, Human Immunodeficiency Virus, People living with HIV, PLWH, HIV/TB co-infection, drug resistant tuberculosis/HIV coinfection, Multidrug resistant tuberculosis HIV coinfection, Sub Saharan Africa, Africa South of Sahara, Antiretroviral therapy, Antiretroviral medication, Antiretroviral regimen, Antiretroviral administration, Highly active antiretroviral, HAART, ART, ARV, Treatment, Treatment outcome, Survival, Success, Failure, default, cure, died, loss to follow up, treatment completed (appendix 7).
To identify studies only in grey literature, abstracts submitted to the annual conference of the International Journal of Tuberculosis and Lung disease (IJTLD) from 2005-2017 were manually searched. Authors whose abstracts matched inclusion criteria were contacted via e-mail for additional data. Annual reports and publications by Medecins Sans Frontieres, WHO, United Nations Joint Commission on HIV/AIDS (UNAIDS) and US Presidents Emergency Plan for AIDS Relief (PEPFAR) were also searched.
Inclusion criteria
Included studies fulfilled the following characteristics: 1) Included culture or drug susceptibility testing confirmed MDR-TB patients coinfected with HIV, 2) clearly reported the use of either individualized, standardized or mixed MDR-TB treatment regimen, 3) reported the use of ART during the treatment of MDR-TB and HIV coinfection 4) reported at least one of the six outcomes recommended by Learson et al and WHO [10,11], 5) documented a minimal age of 15 years for study participants, 6) conducted between January 2004 when ART roll out started to May 2018 (with the exception of studies that compared MDR-TB, and MDR-TB and HIV coinfected patients that began with collection of MDR-TB data earlier than 2004 and subsequently incorporated MDR-TB and HIV coinfection data from 2004), and 7) conducted in SSA, and published in English.
Study selection
After exhaustive database, bibliographic and manual searching, retrieved studies were screened with duplicates removed. Publication titles and abstracts were initially screened, and non-relevant ones excluded. The full text of retained studies were read and those that did not match inclusion criteria were excluded with justification. Study quality and risk of bias in reporting findings was assessed. Two authors (EDC) and (VH) independently conducted article and abstract screening, while the third author (MCVH) validated these. Disagreements on articles to include/exclude were resolved through consensus.
Data extraction
We extracted data using a data extraction form, designed in Microsoft Excel based on those used in other reviews and piloted on five studies. Data was extracted by one author and checked by another. In cases of disagreement, a consensus was reached among all authors. Key data extracted for each study include: year of publication, study design, study country, study description, sample size, participant age, number with confirmed MDR-TB and HIV coinfection, number with MDR-TB, number on ART & time of initiation on ART, type of MDR-TB regimen, MDR-TB treatment duration, treatment setting (centralised hospital-based treatment or decentralised community, clinic and hospital- based treatment), treatment outcomes, and predictors of MDR-TB treatment failure and mortality (appendix 8).
Quality assessment
Cohort and case control studies were assessed for quality using the Newcastle-Ottawa quality assessment scale [20] and the randomized control trial using the National Institute of Health (NIH), quality assessment tool for controlled intervention studies [29]. Indicators of good study quality captured by both tools were; use of standardized method to confirm MDR-TB and HIV status; large sample size with a cut off value of at least 40 participants; multicentre study; integration of home, clinic and hospital based care; use of appropriate statistical tests to classify and report outcomes; taking confounders (demographics, socioeconomic status, previous treatment history) into account during data analysis; clear method of participant selection; and representativeness of study participants to the population of MDR-TB and HIV coinfected population. In addition, use of a valid source for retrieving outcomes and participant information; adequate treatment duration; reporting less than 1/3 missing data at final analysis compared to original population recruited, and proof of ethical review of the study were also considered.
Treatment outcomes were recorded following the proposal by Learson et al [10] and WHO [11] definitions (Appendix 1).
Data analysis and synthesis
STATA 13.1 (Statacorp) and Review manager version 5.3 were used for meta-analysis. Heterogeneity among studies was evaluated at 95% confidence level using binary effect analysis. To facilitate analysis and enable computation of dichotomous effect measures, outcomes were grouped into two categories; treatment success (cure and competed treatment) and unsuccessful treatment (death, defaulted, lost to follow up, failure). Studies were included in the analysis based on outcomes reported. A sub-group analysis for treatment success and unsuccessful treatment in those with MDR-TB according to HIV status was conducted using the Mantel-Haensel random effects method. Heterogeneity of binary covariates was estimated using I2 and P values, at 95% confidence level.