PhD studentship – stipend (tax free) of £21,843 pa for 3 years + home tuition fees (overseas fees cannot be covered)

Supervisors: Professor Victoria Cornelius, Dr Hadith Rastad, Dr Tim Rawson

Background and rationale

Sequential Multiple Assignment Randomised Trials (SMARTs) are increasingly used to evaluate multistage treatment strategies that adapt to patient response.1,2 They are now common in mental health, substance use, chronic disease management and other areas. A recent review by Freeman et al. (2009–2024) provides the most comprehensive overview of SMARTs to date and highlights several important common problems.1 First, sample size calculations rarely reflect the multistage structure and the intended treatment‑sequence comparisons. Second, reporting of initial and subsequent randomisation procedures is often incomplete. Third, primary aims and estimands are frequently unclear. Finally, embedded‑regime and tailored‑regime analyses are under‑used:  Only a minority of trials analyse prespecified treatment sequences, and none in the review implemented deeply tailored Dynamic Treatment Regimens methods in primary reports.1

There are several complex aspects to SMARTs and these require specialised attention. 3-7 Specific supportive guidance is lacking and there is a lack of consideration regarding reporting, interpretability, reproducibility and clinical impact of SMARTs. Existing resources tend to present methods at a high level and are not tailored to the practical needs of applied trial statisticians. This aim of this PhD will be to understand and develop a practical, method‑by‑method guideline that support statisticians in the design and analysis choices of SMART trial. This includes:

  • Approaches to sample size calculation and design considerations 4
  • Selection of suitable analysis models for SMART designs, aligned with chosen estimands and assumptions. Comparison of model performance (under violated assumptions, with respect to bias and loss of efficiency)
  • Assessment of ease of implementation and interpretation in standard statistical software.
  • Identification of minimum reporting items needed to make SMART methods transparent and reproducible.

Importantly, the project will be anchored in a clinical case study: the PATH sepsis trial, a stratified, sequentially randomised trial of precision antimicrobial prescribing with a nested feasibility assessment of second‑stage randomisation. PATH provides a rich, realistic setting in which to:

  • Apply and compare candidate SMART analysis strategies to actual two‑stage treatment sequences;
  • Explore sample‑size and design issues; and
  • Translate methodological recommendations into concrete reporting guidance for multi‑stage trials.

Overall aim

To develop a practical, method focused guideline for the design, analysis and reporting of SMART trials.

Planned work

The student will:

  • Undertake a structured review of current practice, including design, analysis, sample‑size methods and reporting of SMARTs. 
  • Carry out an analytical review of existing sample‑size approaches for SMARTs and evaluate their performance through targeted simulation studies. 
  • Conduct methodological and simulation work comparing key analysis strategies for treatment sequences (e.g. weighting/replication, g‑methods,), focusing on bias, efficiency and robustness. 
  • Apply candidate methods to PATH trial data to illustrate their implementation, assess feasibility in a real setting and generate design inputs for a future SMART.
  • Synthesise findings into explicit recommendations on method choice, assumptions, complexity, potential biases and minimum reporting requirements, then refine these via feedback from methodologists and applied trial statisticians (e.g. survey or workshop).

Methods overview

  • Evidence synthesis: extended scoping/systematic review of SMARTs, structured data extraction and descriptive analyses of current practice.
  • Methodological work: focused simulations using realistic but generic SMART scenarios; comparison of methods on bias, efficiency and coverage.
  • Guideline development: synthesis of review and simulation results into decision tables, worked examples and reporting items, with iterative refinement based on feedback from biostatisticians experienced in SMARTs.

Candidate profile

This PhD is suitable for a UK resident with:

  • Strong analytical and programming skills (e.g. R, Stata or similar)
  • A Degree or Master’s with a substantial statistics component
  • Evidence of interest or experience in clinical trials methodology and adaptive designs

The candidate will join Imperial Clinical Trials Unit and will be working closely with a study team in Department of Infectious Disease at Imperial. The location is at Stadium House, Wood Lane, London, W12 7RH.  The PhD is funded by a new exciting partnership between and the Fleming Initiative, led by Imperial College London and Imperial College Healthcare NHS Trust, with the aim of tackling anti-microbial resistance (AMR).

Application process

To apply, please submit the following documents by email to ictu-phd@imperial.ac.uk, quoting ‘Developing Practical Method Based Guidance for Trial Statisticians for the Design and Analysis of SMART trials PhD Studentship Application’ in the subject line:

Closing date: 23:59 GMT Sunday 22nd March 2026. Interviews will take place on 17th April 2026.  

References 
1.    Freeman NLB, Browder SE, Rowland B, et al. Design characteristics of sequential multiple assignment randomised trials (SMARTs) for human health: a scoping review of studies between 2009 and 2024. BMJ Open 2025;15:e105506. doi:10.1136/bmjopen 2025 105506.
2.    Almirall D, Nahum Shani I, Sherwood NE, Murphy SA. Introduction to SMART designs for the development of adaptive interventions: with application to weight loss research. Transl Behav Med 2014;4(3):260–74. doi:10.1007/s13142 014 0265 0.
3.    Kidwell KM, Almirall D. Sequential, multiple assignment, randomized trial designs. JAMA 2023;329(4):336–7. doi:10.1001/jama.2022.23582.
4.    Kidwell KM, Yoo H, Lions RW, et al. SMART sample size calculators for planning sequential multiple assignment randomized trials. BMC Med Res Methodol 2018;18:121. doi:10.1186/s12874 018 0570 x.
5.    Orellana L, Rotnitzky A, Robins JM. Dynamic regime marginal structural mean models for estimation of optimal dynamic treatment regimes, part I: main content. Int J Biostat 2010;6(2):Article 8. doi:10.2202/1557 4679.1180.
6.    Nahum Shani I, Qian M, Almirall D, et al. Q learning: a data analysis method for constructing adaptive interventions. Psychol Methods 2012;17(4):478–94. doi:10.1037/a0029373.
7.    Laber EB, Lizotte DJ, Qian M, Pelham WE, Murphy SA. Dynamic treatment regimes: technical challenges and applications. Electron J Stat 2014;8(1):1225–75. doi:10.1214/14 EJS923.
8.    Murphy SA. Optimal dynamic treatment regimes. J R Stat Soc B 2003;65(2):331–66. doi:10.1111/1467 9868.00389