Overview of Research at the MRC Biostatistics Unit
Contents
Mission of the Unit
Progress in biomedical science is increasingly dominated by quantitative inference in the presence of uncertainty, and this is the focus of biostatistical endeavour. Innovative statistical methods are required for issues that arise with:
- new sources of data: bioinformatics is a clear case in which large and complex data structures arise from, for example, functional genomic or microarray studies. Functional MRI and other imaging technologies also generate vast quantities of data and a demand for efficient and robust analysis.
- new forms of data collection: increased computerisation and systematization of data collection mean that what were previously considered as individual studies can now be linked and simultaneously analysed. Examples include the use of multiple performance indicators in monitoring a healthcare system, systematic reviews based on individual patient data, and linked longitudinal data for studies of disease natural history.
- new study designs: standard designs such as case-control studies or two-arm randomised trials are increasingly being supplemented by innovations such as life-course studies and cluster randomised trials, requiring appropriate methodology and guidance to practitioners.
- new application areas: for example, recent requirements by agencies such as the National Institute for Health and Clinical Excellence (NICE) for proof of cost-effectiveness has fuelled interest in long-term health economic evaluation of interventions, requiring substantial attention to statistical issues concerning models of disease processes, uncertainty in parameter estimates, and communicating uncertainty about cost-effectiveness.
All these developments present statistical challenges in design, analysis and presentation. In addition, classic problems such as missing data and measurement error apply in almost all areas and continue to present methodological challenges.
The Unit aims to maintain a balance between methodological development and application. The most useful new methodology is stimulated by applied studies, because attention is then focused on real problems rather than artificial abstractions, but this can still lead to generic ideas that are transferable between application areas. In turn, the Unit's involvement in applied studies is stimulated and enhanced by links to its methodological research programmes. Finally, efficient and wide-ranging dissemination is vital.
An additional generic research theme stems from the increasing reliance on 'probabilistic models' as a means of synthesizing all available evidence concerning a biomedical research problem or a health policy decision. Such models are particularly appropriate for complex policy decisions, and should encourage the use of all relevant information, and provide reasoned predictions and inferences about quantities of interest. A range of issues arises concerning, for example, the appropriate complexity of the model, incorporating parameter uncertainty, explanation of the conclusions in an accountable manner, ensuring robustness, and so on. The Unit is taking a leading role in the development of methods and software for handling both the construction and criticism of such models.
The Unit's distinct role
The Unit is an internationally acclaimed research institution with a primary emphasis on methodological research across a wide range of topics in medical statistics. It is also one of the largest groupings of biostatisticians in Europe. Its focus on medical applications makes it distinct from the vast majority of general statistical departments in UK universities. The Unit's methodologies are cross-disciplinary, and other jurisdictions that impact on public health are not ignored. Internationally, the Unit is one of only a few comparable centres for biostatistics research.
Historically, the strength of the UK medical research community has been based on sound collaboration between clinicians and statisticians, and the MRC (or its predecessor, the Medical Research Committee) has had a Statistical Unit, in various forms, since its inception in 1913-1914. We look forward to celebrating our centenary in 2013-2014.
Research themes
The Unit currently focuses on certain themes:
Disease progression modelling: Recent major contributions made by the Unit include research on progression in rheumatic diseases and in hepatitis C infection, clinical events after heart transplantation, and the development of old age disability.
Clinical trial methodology: Focusing on large-scale pragmatic trials, the Unit has made recent advances in methods for handling missing baseline and outcome data, non-adherence to randomised treatments, design and analysis of cluster randomised trials, and within-trial cost-effectiveness analysis.
Evidence synthesis: The Unit has made major contributions in methods for investigating heterogeneity in meta-analysis. Current research focuses on methodology for systematic review of gene-disease associations, and long-term health economic modelling (with recent important applications in population screening for abdominal aortic aneurysms, and in the assessment of implantable cardioverter defibrillators). Future work includes generic methods for modelling bias and relevance in evidence synthesis.
Bioinformatics and genetics: This more recent component of the Unit's work includes characterising DNA sequence patterns, predicting structural aspects of protein complexes, and identification of disease susceptibility genes in coronary artery disease, psoriatic arthritis, and multiple sclerosis. Work is expanding into the area of functional genomics.
Statistics in public policy: The Unit has had a major impact in delivering efficient designs for assessing public health risks, including HIV, vCJD and drug overdose deaths. Statistics in public policy is likely to be an increasingly important area of work, including research on performance indicators with the UK Healthcare Commission, criminal justice, and health economic decision making.