Fully funded PhD opportunity in the Department of Statistical Science, UCL.
Applicants are invited for a funded 3-year PhD studentship, which will be placed in the Department of Statistical Science, University College London. The title of the PhD project is "Extending the multi-state model for the natural history of prostate cancer". The project is part of the research in the Prostate Working Group in the Cancer Intervention and Surveillance Modeling Network (CISNET). The aim of the PhD project is to extend the multi-state modelling that is currently used at UCL, and to compare the modelling with other approaches in the Prostate Working Group. Details are given and the end of this email
Stipend:
The stipend for 2020/2021 is £17,285, which is increased yearly following inflation. There is additional funding for equipment and travel.
Applying:
All candidates should first apply for admission to the Research Degree: Statistical Science (RRDSTASING01). Details and instructions can be found on
www.ucl.ac.uk/statistics/prospective-postgraduates/phd
which includes a link to
www.ucl.ac.uk/prospective-students/graduate/research-degrees/statistical-science-mphil-phd
where you can apply online.
For this project, you do *not* need to prepare an outline proposal. You are encouraged to contact Ardo van den Hout for more information or any questions that you may have.
Deadline for application: 24 May 2021
Contact:
Ardo van den Hout
Department of Statistical Science
University College London
email address: Este enderezo de correo está a ser protexido dos robots de correo lixo. Precisa activar o JavaScript para velo.
--------------------------------------
Details:
Continuous-time multi-state models can be used to describe disease progression. Death as a competing risk can be accounted for by including an absorbing dead state. In this manner, we have defined a four-state model for the progression of prostate cancer. Individuals are in state 1 if they are cancer free. State 2 is defined by having the disease in a detectable state. Individuals are in state 3 if they have cancer that is clinically diagnosed. State 4 is the dead state.
The continuous-time model consists of parametric hazard submodels for each of the transitions. A hazard model can be defined by linking the hazard (risk) of a transition to covariates such as age or biomarkers. Assumptions in our current model for prostate cancer: the disease is progressive (no transitions possible from state 2 to state 1), transitions can happen at any time (continuous-time process), and only the current state and current values of covariates determine the risk of moving to the next state (partial Markov assumption). The model includes measurement error (misclassification of state).
Longitudinal data are used to estimate the model parameters. We use data from two groups: a screened group, and a control group. In the screened group, individuals are screened regularly. For this group, we do not use follow-up data for individuals once they have been observed in state 2. In the control group we have no observations of state 2 since the controls are not screened. The left-truncation in this data is taken into account in the statistical inference.
The model can help to describe cancer progression in more than one way. Of specific interest is inference on time spent in the detectable state, and how this time depends on covariates. In addition, the model can help to decide on screening strategies, treatment timing and approaches.
The current four-state model can be extended in several ways. Examples are the inclusion of more covariates (such as genetic information), extending the number of states (such as splitting state 2 into screen-detectable early to screen-detectable late), taking into account the history of the process (semi-Markov model), or adding transitions (such as the transition from state 3 to 4 in our model). It is also possible to account for unobserved heterogeneity by defining frailty models.
The PhD project will start with studying the current methodological framework (multi-state survival models, stochastic processes, statistical computing, and micro simulation), and will then develop extensions of the current multi-state model. Throughout the PhD project, the modelling developed at UCL will be compared with modelling proposed by other teams in the Prostate Working Group of CISNET.
The primary supervisor of the PhD project will be Dr. Ardo van den Hout. The research will be in collaboration with Professor Nora Pashayan, who is in the UCL Department of Applied Health Research.
Candidates should have a strong background in mathematical statistics (for instance, by having a BSc or MSc in statistics or in a similar field), and have an interest in research in stochastic processes. The research in the project is computationally intensive, and experience or affinity with high-level programming is a requirement.
-------------------
