Design characteristics of sequential multiple assignment randomised trials (SMARTs) for human health: a scoping review of studies between 2009 and 2024.

OBJECTIVE: To characterise the reporting practices of sequential multiple assignment randomised trials (SMARTs) in human health research. DESIGN: Scoping review of protocol and primary analysis papers describing SMARTs published between January 2009 and February 2024. BACKGROUND: SMARTs are innovative trial designs that allow for multiple stages of randomisation to treatment, with randomization potentially based on a patient's response(s) to previous treatment(s). They are uniquely designed to develop sequential adaptive interventions (dynamic treatment regimes (DTRs)) to support personalized clinical decision-making over time. Previous reviews have identified inconsistencies in how the design, implementation and results of SMARTs have been reported in published studies. A comprehensive assessment of SMART reporting practices is lacking and necessary for developing standardised SMART-specific reporting guidelines. METHODS: We systematically searched multiple databases for SMART-related protocol and primary analysis papers published between January 2009 and February 2024. Title, abstract and full-text screenings were performed by pairs of reviewers, with disagreements resolved by consensus. Data extraction included study characteristics, design elements and analytical approaches for embedded or tailored DTRs. Results were synthesised qualitatively and presented descriptively. RESULTS: From 5486 screened studies, 103 (59 protocol papers, 16 primary analysis papers, 14 protocol papers with corresponding primary analysis papers) met the inclusion criteria. Most studies targeted adults (62.7% protocols, 62.5% primary analyses, 42.9% protocol+primary analyses) and were primarily conducted in the USA. Behavioural and mental health constituted the most frequent therapeutic domain. While intervention descriptions and re-randomisation criteria were consistently reported, operational characteristics such as blinding (protocols: 64.4%, primary analyses: 62.5%, protocols+primary analyses: 71.4%) and randomisation details (protocols: 55.9%, primary analyses: 37.5%, protocols+primary analyses: 50.0%) were inconsistently documented. Only 46.7% of primary analyses evaluated embedded DTRs, and none explored deeply tailored DTRs. CONCLUSIONS: Despite the increased adoption of SMART designs, substantial reporting variability persists. Most primary analyses underuse the capability of SMARTs to generate data for developing DTRs. SMART-specific standardised reporting guidelines can help accelerate the scientific and clinical impact of SMARTs.

Duke Scholars

Author Nikki Freeman Biostatistics & Bioinformatics, Division of Biostatistics

Author Kevin J. Anstrom Biostatistics & Bioinformatics, Division of Biostatistics