Call for PhD Applicants
Students interested in undertaking a PhD in trials methodology are encouraged to contact
Expressions of interest should include a copy of your:
i. CV
ii. Academic Transcript
We currently have the following projects available:
Dealing with missing outcome data in randomised trials with extended follow-up
Supervisors: Professor Katherine Lee (Murdoch Children’s Research Institute, Melbourne) and Dr Thomas Sullivan (South Australian Health and Medical Research Institute, Adelaide).
Location: Melbourne or Adelaide.
Background: Missing outcome data can pose a serious problem in randomised trials, requiring careful consideration during analysis to avoid biasing conclusions. The challenges of missing outcome data are especially prominent in “extended follow-up” studies, where an additional follow-up period is commenced after the protocol defined completion of a trial to learn about longer-term impacts of the intervention. Extended follow-up studies often involve additional eligibility restrictions and consent processes, which can further limit the amount of data available for analysis. While the method of multiple imputation has been widely recommended for addressing missing outcome data in randomised trials, its application in extended follow-up studies is not well understood.
Specific objectives:
- To review the literature on methods of multiple imputation for randomised trials and assess their suitability and limitations for trials with extended follow-up.
- To explore how the inclusion of randomised participants ruled ineligible or not consenting to extended follow-up in an imputation model affects the validity of treatment effect estimates.
- To evaluate the ability of multiple imputation to address large blocks of missing outcome data, as occurs when participants fail to provide any data during extended follow-up, compared with alternative approaches such as complete case analysis or using multiple imputation in combination with inverse probability weighting.
- To develop approaches and provide guidance on conducting sensitivity analyses in extended follow-up studies when the probability of missing data is suspected to depend on unobserved data.
Optimising dynamic treatment regimens using sequential multiple assignment randomised trial data with missing data
Supervisors: Dr Robert Mahar (Murdoch Children’s Research Institute/University of Melbourne), Prof Katherine Lee (Murdoch Children’s Research Institute), Prof Julie Simpson (University of Melbourne).
Location: Melbourne
Background: Clinicians often face multi-stage and dynamic decisions when treating patients with either chronic or progressive medical conditions. Dynamic treatment regimens formalise such sequential decision problems. Sequential multiple assignment randomised trials (SMARTs) are clinical studies that randomise patients to different treatments over time, which provides data that can be used to optimise dynamic treatment regimens. But repeatedly randomising patients to treatments can lead to complicated patterns of missing data, a problem with few existing SMART-specific solutions. This PhD project will develop and evaluate statistical methods to optimise dynamic treatment regimens using data from hypothetical SMART designs with problematic missing data.
Specific objectives:
- To review the literature on methods to impute missing data for randomised trials and assess their suitability and limitations for SMARTs.
- To formalise patterns of SMART missing data using directed acyclic graphs.
- To evaluate the ability of different imputation methods to deal with different patterns of missingness compared with alternative approaches such as complete case analysis.
- To develop approaches and provide guidance on conducting sensitivity analyses in SMARTs when the probability of missing data is suspected to depend on unobserved data.
Handling missing data in repeatedly measured outcomes when assessing change over time
Supervisors: Prof Katherine Lee (Murdoch Children’s Research Institute / The University of Melbourne), Dr Rheanna Mainzer (Murdoch Children’s Research Institute)
Location: Clinical Epidemiology and Biostatistics Unit, Murdoch Children’s Research Institute
Background: Missing data are common in medical research, particularly in studies with multiple waves of follow-up. One particular challenge is how to handle missing data in analyses that involve the change in a variable over time where data are missing at one or more time points for some individuals. Analysis methods specifically designed for longitudinal data, such as mixed effects models and generalised estimation equations, allow all individuals with the outcome measured at any time point to be included in the analysis, but would exclude participants who have missing data at all time points. Multiple imputation (MI) is a popular method for handling missing data, including in longitudinal studies, which would enable all participants to be included in the analysis. This PhD project will develop and evaluate statistical methods for analysing change over time in the presence of missing data.
Specific objectives:
• To review the literature on methods for analysing change over time in the presence of missing data.
• To evaluate different methods for handling missing data in studies where the change over time is of interest.
Further Projects
We welcome EOIs for the following project:
Accelerated failure time models in health studies
Supervisor: Professor Rory Wolfe (Monash University, School of Public Health and Preventive Medicine)
Location: Melbourne
