Address: |
MRC Biostatistics Unit,
Institute of Public Health,
University Forvie Site,
Robinson Way,
Cambridge. UK.
CB2 0SR |
Predictions from hierarchical random-effects models
When biological or physiological variables change over time interest often lies in making predictions of future measurements or the time taken to reach a certain threshold value. My interest lies in how such predictions can be made using random-effects growth models, both classically and in a Bayesian framework. Applications include modelling the growth of abdominal aortic aneurysms (AAA) and modelling the antibody response to HIV infection as measured through an incidence assay.
Clinical trials
I am currently interested in implementing efficient designs for phase II cancer trials. I am also involved in a large multi-centre abdominal aortic aneurysm surgical trial (IMPROVE), and a number of psychiatric and health intervention trials in collaboration with the MRC General Practice Research Framework.
Multi-parameter evidence synthesis
Knowledge about a certain epidemiological quantity, such as prevalence of an infectious disease, is often obtained from a number of studies and data sources of varying quality. I am interested in the use of complex multi-parameter evidence syntheses to coherently synthesise the available data accounting for potential biases. Such models can be used to highlight areas of research where knowledge gaps exist. One example of this is in the estimation of hepatitis C prevalence, where a lack of knowledge about the number of ex-injecting drug users in the population results in uncertain estimates of hepatitis C prevalence. Another application is a bias-adjusted meta-analysis of the effect of routine anti-D prophylaxis in antenatal care of Rhesus negative women from studies of varying relevance and rigour.
Multi-state modelling
I am interested in the use of multi-state models to describe disease progression using data where the times of transition between disease states are generally interval-censored. One particular interest is the situation where examination times are informative with respect to the outcome of interest. An application of this situation comes from a cohort of hepatitis C infected individuals where disease state is measured infrequently and irregularly. I have also used a multi-state framework within backcalculation models to explicitly model disease and diagnosis processes in order to estimate incidence and prevalence of an infectious disease. Applications involve using Bayesian backcalculation methods to estimate HIV and HCV incidence from routine diagnosis and surveillance data. |
Sweeting MJ, De Angelis D, Hickman M, Ades AE. Estimating hepatitis C prevalence in England and Wales by synthesizing evidence from multiple data sources. Assessing data conflict and model fit. Biostatistics. 2008;9(4):715-734
Sweeting MJ, De Angelis D, Brant LJ, Harris HE, Mann AG, Ramsay ME. The burden of hepatitis C in England. Journal of Viral Hepatitis. 2007;14(8):570-576.
Sweeting MJ, De Angelis D, Neal KR, Ramsay ME, Wright M, Brant L, Harris HE. Estimating progression to cirrhosis in three UK hepatitis C cohorts: the effect of recruitment bias. Journal of Clinical Epidemiology. 2006;59(2): 144-52.
Sweeting MJ, De Angelis D, Aalen OO. Bayesian back-calculation using a multi-state model with application to HIV. Statistics in Medicine. 2005;24(24):3991-4007. |
