New analytical and simulation tools in clinical oncology

New analytical and simulation tools in clinical oncology

PhD studentship in Mathematical Sciences

School of Mathematical Sciences, University of Nottingham
and BAST Inc. Limited, Loughborough

***Enhanced stipend of £16,000 per annum plus additional support for computing equipment and travel***

Supervised by:
• Dr Gilles Stupfler and Dr Chris Brignell (Nottingham)
• Dr Blesson Chacko (BAST Inc. Ltd.)
Background: Cancer drug developers seek to show that satisfactory drug exposure increases the rate of beneficial patient responses, but without accounting for the risk of adverse effects. This paradigm is the so-called cause-specific (CS) approach to hazard modelling, in which even a marginal improvement in average patient outcomes is seen as a positive result, even if treatment induces strong adverse reactions (typically showing as dropouts in clinical trials). It relies on the Cox proportional hazards model, which can be rather restrictive.
By contrast, sub-distribution (SD) hazard modelling clearly sets apart dropout and right-censoring. As such, it classifies dropout as a competing risk. In this paradigm, a treatment will only show positive results when the effect of treatment on the response (e.g. lifetime) is substantially stronger that the effect on the dropout rate. This is what payers, rather than drug developers, are interested in.
SD hazard models are, however, more difficult to develop. An additional difficulty in oncology trials is that the median age of patients is typically high (around 70 years). Death from one of several causes during trial is therefore a distinct possibility; death is a special event in survival analysis, since a patient who leaves the study due to their death will not carry any risk forward, contrary to censored or dropping-out patients.
Aims: This PhD project will be set in the context of SD hazard modelling. The focus will be the development of fully parametric SD hazard models that:
1. account for competing events (dropout, death, …),
2. can handle covariate information such as body weight so as to produce precise analyses tailored to individual patients,
3. and are more flexible than the standard proportional hazards model.
The rationale behind the development of such models is the gradual recognition among the industry that a patient’s individual risk of wasting limited life expectancy during treatment should easily be quantified so that drug exposure can be optimised. The project will in particular involve designing a parametric method to compute the so-called cumulative incidence functions (CIFs), which enables the analyst to evaluate the effect of a covariate on the response to treatment; the use of a parametric method will allow for the development of prediction techniques with a view on improving the transition from Phase II to Phase III in oncology trials. Power and sensitivity calculations will also be carried out.
This project will be led jointly by the School of Mathematical Sciences and the private company BAST Inc. Ltd. through their scientific director Dr Joachim Grevel. The student will be primarily based at the University of Nottingham during the first year of the studentship, and then at BAST’s premises in the second, third and fourth years. Weekly supervisory meetings will be organised with BAST and the academic supervisory team for the whole duration of the studentship.