Research
LSHTM is a leading international centre for academic medical statistics, and the Medical Statistics Unit is a focus for its excellence in both research and teaching.
The Unit undertakes a broadly based research programme both in applied projects and in statistical methodology. The applied research is in collaborations worldwide with renowned centres in clinical medicine, epidemiology and public health, and often generates methodological questions. The methodological research is undertaken with a practical intent: to enhance the quality of future applied research in its design, conduct, analysis interpretation and reporting.
Much of the Unit's research can be classified under four broad areas: statistical methodology, epidemiological statistics, clinical trials, and pharmaco-epidemiology. This covers a very wide range of applications across medicine and public health, with particular expertise in cardiovascular and neurological diseases.
Highlights of current research in the four areas are:
Statistical Methodology: missing data, especially in longitudinal studies; propensity scores and other methods of adjustment for confounders; time-updated models relating disease events/biomarkers to prognosis; development of user-friendly prognostic risk scores; allowance for measurement error; developments in sensitivity analyses; small sample inference for mixed models; statistical issues in brain imaging and human growth studies; use of Figures in medical journals.
Epidemiological Statistics: modelling of intermediate outcomes in life-course epidemiology; dealing with missing linkage data; competing risks and multistate models in applications to infant mortality and chronic disease epidemiology; guidelines on reporting of epidemiological studies; methods for grouping quantitative risk factors.
Clinical Trials: applied research involves being the independent statistical centre, especially for data monitoring, in major international industry-sponsored trials, trial co-ordination and leadership in public-funded trials, a statistical resource for planning, analysis and interpretation of specific trials, use of large-scale trial databases for investigations of patient prognosis and treatment effects. Methodological activities include methods for data monitoring, reporting of trials, controversies in trial design including adaptive designs, non-inferiority trials and surrogate endpoints, cross-over trials; qualitative research into views of trial participants, multiplicity of data (e.g. subgroup analyses, composite endpoints, repeated measures) in trials, methods for systematic reviews.
Pharmaco-epidemiology: applications include collaborations in pharmaco-epidemiology and pharmaco-vigilance, uses of large primary care databases, e.g. GPRD, for evaluating specific drug safety issues. Methodological activities include development of signal detection methods in pharmaco-vigilance, guidance on the role of observational studies, (as compared to randomised trials) for evaluation of benefits and harms of drugs, innovative designs in pharmaco-epidemiology e.g. case series methods, the development of hierarchical Bayesian methods and cluster analysis for grouping of medically related events.
The Unit's faculty are also very active in teaching to both statisticians and other medical/epidemiological professions. This includes PhD degree students, the School's Masters degree courses, notably in Medical Statistics and the Distance learning MSc in Clinical Trials, and short courses, especially in medical statistics, epidemiological methods, clinical trials and pharmaco-epidemiology. Faculty members are also frequently engaged in other international short course teaching programmes.
