Beyond dichotomisation: Efficient estimation of response rates using continuous outcomes
Presenting modelling approaches that can retain continuous clinical outcome, while still estimating clinically meaningful response rates
Dichotomisation of continuous outcomes into 'responder' and 'non-responder' categories remains widespread in clinical research, particularly where a threshold carries clinical meaning (e.g. blood pressure control), or is established convention (e.g. RECIST criteria for tumour response). While easy to interpret, this practice is well known to discard information, reducing statistical efficiency and inflating required sample sizes if a study is powered for the responder endpoint.
In this talk, Michael will discuss alternative modelling approaches that retain the continuous outcome, while still targeting the same clinically meaningful response rate estimand. By leveraging the full information contained in the continuous outcome, such methods can achieve substantial gains in efficiency compared with dichotomised analyses. However, these gains come with a potential cost: if the assumed parametric model is misspecified, the resulting estimates may be biased and type I error rate can be inflated.
Michael will outline the process to estimate response rates and standard errors from fitted regression-based parametric models and present a suite of potential models from simple linear regression to a locally targeted estimation approach. He will consider the robustness of these methods to model misspecification and present results from an extensive simulation study under a range of data-generating mechanisms. He will then discuss diagnostic tools and a threshold sensitivity ('tipping-point') analysis for assessing the validity of these approaches in practice.
The talk on efficient yet robust estimation of responder estimands will be of interest to statisticians and researchers working in the design and analysis of clinical trials and observational studies.
Speaker
- Dr Michael Sweeting - Director of Statistical Innovation, GSK
Event notices
- Please note that you can join this event in person or you can join the session remotely
- Please note that the recording link will be listed on this page when available
Admission
Contact
