Joint SSA Biostatistics/Victorian Branch Event
The SSA Biostatistics and Victorian branch are excited to announce that they are hosting a joint event in August. Benefitting from the fact that Australia will be hosting the International Society for Clinical Biostatistics conference for the first time, we had the opportunity to invite three fantastic early to mid-career biostatisticians from overseas – Valentijn de Jong (Netherlands), Jennifer Thompson (UK) and Simon White (UK) – to come and present their research in Melbourne. Come and join us to hear more about their work. Free drinks and nibbles will be provided before the talks, and you are welcome to join us for dinner after the event (meal at own expense).
Registration is welcome but not essential.
Individual patient data meta-analysis of intervention studies with time-to-event outcomes: A review of the methodology and a case study
Many randomized trials involve time-to-event outcomes. Evidence from these trials can be synthesized by combining individual patient data (IPD) and performing a so called IPD meta-analysis (IPD-MA). We performed a narrative literature review to provide an overview of methods for conducting an IPD-MA of intervention studies with a time-to-event outcome. Hereby, we focused on assessing good methodological practice for modeling frailty of study patients across studies, for modeling heterogeneity of treatment effects, for choosing appropriate association measures, for dealing with (differences in) censoring and follow-up times, and for addressing time-varying treatment effects, and effect modification. We discuss parametric and semi-parametric methods for modeling time-to-event data in an IPD-MA, and describe how to implement these in a one-stage or two-stage meta-analysis framework. We will illustrate several key methods in an empirical IPD-MA where clinical trial data from 1225 patients from 5 studies were combined to compare the effects of Carbamazepine and Valproate on the incidence of seizures in epilepsy patients.
Advice for using generalised estimating equations in a stepped wedge trial
Generalised estimating equations (GEE) have been suggested as a robust approach to analysis of stepped-wedge cluster randomised trials (SW-CRTs) because of their property of consistent estimation of effects even with a misspecified correlation structure. However, their performance for SW-CRTs has not been fully explored. In this talk I will discuss the results of a simulation study comparing the performance of GEE with robust standard errors and different working correlation matrices versus a generalised linear mixed model with a random effect for cluster to analyse SWTs with a range of within-cluster correlation structures. In brief, all models gave unbiased intervention effect estimates regardless of the true correlation structure or the working correlation matrix. With 60 clusters, GEE gave appropriate confidence interval coverage in most scenarios, even when the correlation structure was misspecified. With 48 or fewer clusters, GEE confidence intervals had low coverage that was corrected by the Fay and Graubard small-sample adjustment. These findings lead us to recommend that trialists use GEE with the Fay and Graubard small-sample correction when analysing SWTs to obtain robust intervention-effect estimates, confidence intervals, and p-values.
Simon White, Medical Research Council Biostatistics Unit, University of Cambridge
Modelling heterogeneity in neuroimaging: people and the brain
Neuroimaging allows us to gain insight in vivo on the structure and activity of the brain across multiple modalities (MRI, fMRI, EEG, MEG, etc). Clearly there is significant spatial structure, the left and right hemispheres for example, and any analyses must respect and account for the spatial structure.
Although there is interest in the function of the brain itself, ultimately we are interested in linking the brain with other outcomes; linking people and the brain. This talk will present work on linking neuroimaging and epidemiological data, in particular considering the issue of heterogeneity within our analysis
|Time:||5:45 am - 7:30 pm|
|Cost:||free, but meal at own expense|
|Location:||University College The University of Melbourne,
40 College Crescent,