Statistics for Stratified Medicine and Analysis of Complex Phenotypes

Many diseases and conditions, such as cancers, dementia and rheumatic disorders, are multi-factorial; exhibiting a range of biological manifestations that reflect the contribution of genetic and environmental factors. Even across individuals with rare diseases, which may result from single gene mutations, varying symptoms, health outcomes and treatment responses are seen. The complexity of diseases leads to many challenges ranging from understanding disease mechanisms and susceptibility to risk prediction and development and application of treatments. Stratified medicine, where "homogeneous" groups of people likely to respond similarly to treatment or have similar underlying disease mechanism or risk are sought based on biomarker information, is aimed at targeting therapies and optimal decision making for groups of patients who have shared biological characteristics. That genetic, molecular and imaging technologies have transformed biology has meant the need for those trained in analysing complex data, accounting for uncertainty and making evidence-based conclusions is paramount. Our programme will address statistical issues arising from using complex biological and clinical data for further understanding of diseases and to develop models aimed at risk prediction, stratified prevention, projection of disease course, treatment response, safety and the likelihood of adverse events and the tailoring of therapeutics to individuals.

Stratified Medicine is a strategic priority for the MRC (and other organisations), and a substantial programme of investment by the MRC has begun and will continue for the coming years. The stratified medicine paradigm acknowledges the enormous biotechnological advances in generating and collecting detailed biological genotypic and phenotypic information across the various levels of organisation of the (human) biological system with the capacity to match the complexity of diseases and unmask patterns which can identify new and more subtle strata of disease. The unprecedented availability and progress of “omics” technologies for characterizing individual DNA mutations, metabolic profiles, protein expression and localisation, mRNA expression and epigenetic signatures in specific cell types, coupled with imaging and more traditional clinical, lifestyle and environmental data, have presented opportunities for further understanding of mechanisms, outcomes and molecular heterogeneity of diseases. This paves the way for individual-level models aimed at risk assessment and stratified prevention, projection of disease course, treatment response, safety and the likelihood of adverse events.

We will undertake a programme of statistical methodological research that encompasses risk stratification, prediction and validation, integrative and joint modelling of molecular and clinical data of various kinds and complexities (and their conversion into meaningful outputs that can inform health care decisions), mechanistic understanding and causality, treatment strategies and the design of innovative/purposeful clinical trials for biomarkers, rare diseases and dynamic treatment regimes. Both Bayesian and non-Bayesian methodology will be developed.

We are research partners in a number of MRC and other funded consortia in stratified medicine, and have close links with teams in North America, Europe and the UK. Our collaborations cover many diseases, medical conditions and areas. For example, we are involved in projects in rheumatology, dementia, rare diseases, HIV, antibiotic resistance and blood donation. Some of the goals of these projects are shared. We therefore envisage that methodology developed may be transferrable across diseases and medical areas, although tailoring may be required.