BACKGROUND: Globally it is estimated that 480 000 people developed multidrug-resistant tuberculosis (MDR-TB) in 2014 and 190 000 people died from the disease. Successful treatment outcomes are achieved in only 50 % of patients with MDR-TB, compared to 86 % for drug susceptible disease. It is widely held that delay in time to initiation of treatment for MDR-TB is an important predictor of treatment outcome. The objective of this review was to assess the existing evidence on the outcomes of multidrug- and extensively drug-resistant tuberculosis patients treated early (≤4 weeks) versus late (>4 weeks) after diagnosis of drug resistance. METHODS: Eight sources providing access to 17 globally representative electronic health care databases, indexes, sources of evidence-based reviews and grey literature were searched using terms incorporating time to treatment and MDR-TB. Two-stage sifting in duplicate was employed to assess studies against pre-specified inclusion and exclusion criteria. Only those articles reporting WHO-defined treatment outcomes were considered for inclusion. Articles reporting on fewer than 10 patients, published before 1990, or without a comparison of outcomes in patient groups experiencing different delays to treatment initiation were excluded. RESULTS: The initial search yielded 1978 references, of which 1475 unique references remained after removal of duplicates and 28 articles published pre-1990. After title and abstract sifting, 64 papers underwent full text review. None of these articles fulfilled the criteria for inclusion in the review. CONCLUSIONS: Whilst there is an inherent logic in the theory that treatment delay will lead to poorer treatment outcomes, no published evidence was identified in this systematic review to support this hypothesis. Reports of programmatic changes leading to reductions in treatment delay exist in the literature, but attribution of differences in outcomes specifically to treatment delay is confounded by other contemporaneous changes. Further primary research on this question is not considered a high priority use of limited resources, though where data are available, improved reporting of outcomes by time to treatment should be encouraged.