‘An introduction to covariate adjustment in trials’ – PSI covariate adjustment event

Later today I’m delighted to be participating in an online event organised by the Statisticians in the Pharmaceutical Industry (PSI) on the topic of covariate adjustment in clinical trials. The slides of my introduction talk can be viewed and downloaded below. In the talk I’ll very briefly be covering

  • conditional and marginal effects/estimands
  • reasons to adjust analyses of randomised trials for baseline covariates
  • robustness properties of linear regression models
  • interpretation of effect estimates from regression models
  • the standardisation / G-formula estimator of marginal effects, as mentioned in the FDA’s covariate adjustment guidance
  • some properties of treatment effect estimators when stratified randomisation is used

One aspect that I unfortunately will only have a little time to touch on is the ongoing debate about marginal vs. conditional estimands. For a bit more discussion on this point, see this previous blog post.

PhD on causal inference for competing risks data

Applications are invited for a 3-year PhD studentship from the ESRC UBEL DTP (UCL, Bloomsbury and East London Doctoral Training Partnership)

We are seeking applicants who would like to pursue PhD research on the project described below. This project is offered as part of the Longitudinal Analysis and Design topic under the Quantitative Social Science Pathway of the ESRC UBEL DTP. Successful applicants will based in the Department of Medical Statistics at LSHTM. Further information on the funding scheme can be found at https://ubel-dtp.ac.uk/esrc-studentships/

Read more

Mixed models repeated measures (mmrm) package for R

I was recently made aware of the release of the mmrm package in R. It has been developed by a group of programmers and statisticians at a number of pharmaceutical companies, led by Daniel Sabanes Bove at Roche, as part of the ASA Biopharmaceutical Section Software Engineering Working Group. I’ve written previously about fitting mixed models for repeated measures (MMRM) using R, Stata and SAS. In R, this can be done using the gls function in the nlme package, but there are a number of limitations with this approach. For example, it is difficult (or impossible) to fit models where you allow the covariance parameters to be distinct between treatment groups. In this post, I’ll take a very quick look at the new mmrm package in R.

Read more