The Centre has been involved in several successful Fellowship applications. Project outlines for some of the Fellowships are detailed here:
- Rhian Daniel
The Wellcome Trust and The Royal Society
Sir Henry Dale Fellowship
Statistical methods for studying multidimensional mediators of genetic associations with chronic diseases
5 years (1 Oct 2015 - 30 Sep 2020)
Recent advances in OMICs technology have transformed the nature of biomedical research. Highly-correlated, very high-dimensional biomarker and imaging data are increasingly available in many large prospective datasets such as UK-Biobank. In this fellowship, I will focus on the extension of causal inference methods, particularly mediation analysis with multiple mediators, to these high-dimensional data settings. I plan to apply these methods to address mediation analysis questions in several rich datasets, including the UCLEB consortium of 15 British cohorts with access to 1H-NMR metabolomic data as well as genetic material and incident clinical events (totalling 35K participants with >7K cardiovascular events).
- Karla Diaz-Ordaz
UK Medical Research Council
Career Development Award in Biostatistics
Multiple imputation methods for valid causal treatment effects estimation after departures from protocol
4 years (April 2014 - March 2018)
Randomised controlled trials are considered the gold standard in determining the effect of an intervention. Two issues that often undermine the credibility of effectiveness measures are non-compliance with treatment allocation (participants who do not adhere to the protocol, e.g. by not receiving the intended treatment) and missing data (e.g. loss to follow-up).
Both these issues are relevant to the adoption of the Intention-to-treat (ITT) analysis, which aims to measure treatment effectiveness. The ITT principle states that all individuals randomised in a clinical trial should be included in the analysis, in the groups to which they were randomised, regardless of any departures from the randomised treatment.
Following this principle preserves the benefits of randomisation, i.e. having treatment groups that do not differ systematically on any factors except those assigned in the trial. But such pragmatic estimates may not be the only estimates of interest, and given the considerable costs and high number of patients involved in a typical confirmatory clinical trial, efforts should be made to obtain valid explanatory treatment effects, i.e. the “real” treatment effects that a patient taking the treatment 100% as prescribed can expect on average.
Ad hoc methods which ignore treatment allocation may lead to incorrect estimates of treatment effect. Existing statistical methods can handle some complex longitudinal settings, e.g. when compliance with treatment varies with time and is associated with the clinical outcome of interest (assumed to be a continuous measure). However, they rely on assumptions which are difficult to understand and involve sophisticated numerical iterative procedures to obtain the estimates of interest. It is therefore important that practical and efficient methods are available for handling non-compliance and missing data, especially when the data structure is complex, in a unified, transparent, and systematic manner.
The proposed research aims to develop new statistical methods to deal with these two issues within a single unifying framework and under transparent assumptions. This method is based on multiple imputation, which is a practical and flexible method already widely used to address missing data problems.
- Antonio Gasparrini
UK Medical Research Council
Methodology Research Fellowship
A general conceptual and statistical framework to model non-linear and delayed exposure-response relationships and combine such complex associations across studies
Epidemiological multi-city studies on the health effects of environmental factors, such as air pollution and temperature, have known an intense development in the last few years. Within this study design, two methodological issues are prominent among those that need to be addressed: how to specify exposure-response relationships describing potentially non-linear and delayed effects in each city, and how to combine such complex associations across different cities.
Interestingly, the same problems apply in different research areas involving different study designs and data structures. The extension and further development of statistical methodologies already proposed in time series analysis would provide a valuable tool in a wide range of research fields.
The aim of this project is to develop a general conceptual and statistical framework to model non-linear and delayed exposure-response relationships, and to propose appropriate approaches to combine such complex associations across studies.
- to define a general conceptual framework to formalize the idea of delayed effects
- to provide a unified statistical approach to model non-linear and delayed effects by extending existing methods developed in time series analysis
- to review, compare and extend meta-analytic techniques to synthesize non-linear and/or delayed associations
- to provide a complete software implementation of the methodologies described above in the R statistical software, together with detailed and comprehensive documentation
This research is comprised of sub-projects Sub-P1 and Sub-P2.
Sub-P1 addresses objectives 1-2-4. Existing methods in time series analysis, based on distributed lag non-linear models, will be improved and extended, algebraically and conceptually, to different designs such as cohort and longitudinal analysis. The framework will be implemented by extending the R package dlnm developed by the applicant.
Sub-P2 addresses objectives 3-4. In particular, two existing techniques, meta-smoothing and multivariate meta-analysis, will be compared and extended, assessing their flexibility, applicability, and reliability of their results through simulations and applications to real data. A specific R package will be developed to implement these models.
The research involves collaborations with high-calibre and experienced researchers across three different institutions (LSHTM, MRC-Cambridge and University of Southern California).
The research project aims to provide a unified methodological approach to model and pool non-linear and delayed dependencies, through a general conceptual and statistical framework. The research project emphasizes generality and usability, specifying a systematic and widely applicable methodology, which encompasses different study designs and data structures, and providing a comprehensive software implementation and documentation.
Scientific studies which assess the health effects of various risk factors usually depend on statistical models, used to quantify the relationship between the dose people are exposed to, and the health outcome. This relationship should concisely and reliably summarize the association, in order to correctly inform about the health consequences of specific risk factors. This association is usually estimated in multiple comparable studies, whose evidence needs then to be combined in order to improve the knowledge about related health risks.
Researchers at the London School of Hygiene and Tropical Medicine, in collaboration with the MRC-Cambridge and the University of Southern California, have promoted a project to provide the statistical tools suitable to describe complex relationships and combine them across studies. These methodologies will be implemented in statistical software, freely available and usable by all the researchers working in different biomedical fields. The aim of the project is to improve the analytical method to study the association between risk factors and human health.
- Manuel Gomes
UK Medical Research Council
Early Career Fellowship Economics of Health
Developing appropriate methods for handling missing data in health economic evaluation
Policy makers worldwide use health economic evaluation to help decide which health care interventions to provide. An important concern faced by economic evaluation studies is that there may be missing data. However, most published studies ignore this or use inappropriate methods to address the problem. If missing data are not addressed appropriately this can lead to misleading cost-effectiveness results and scarce health care resources being misallocated. Careful analytical methods are required to address missing data across different circumstances in health economic evaluation. I propose to conduct a comprehensive programme of research to address this gap in knowledge, using both simulation work and data from clinical areas of high policy relevance. The research will develop and test alternative methods for addressing missing data and show the impact that better methods can have in the evaluation of health interventions and health care providers. By helping improve the quality of cost-effectiveness studies which are used to inform policy making, this research will help ensure that scarce resources are allocated in the best ways for improving the population’s health.
- Ruth Keogh
UK Medical Research Council
Development and practical application of landmarking in studies of time-varying exposures and survival
4 years (April 2015 - March 2019)
The aim of this work is to develop of statistical methods which enable us to gain understanding of the effects of time-varying exposures on survival, for example to make predictions of individual features on survival, to gain insight into biological mechanisms or to inform treatment decisions. I am interested especially in methods that enable us to make best use of data such as electronic health records, which contain patient-level variables recorded longitudinally.
My research focuses on:
- Making dynamic predictions of survival to a future time horizon based on an individual’s measurements up to a given time
- Estimating the causal effects on survival of patterns in time-varying exposures, including continuous and binary exposures and intermediate events, taking into account time-varying confounding.
These investigations have to date required quite different complex statistical methods. A unified approach does not exist but is essential to enable applications from a wide range of researchers. In this project I am developing statistical methods based on an approach called ‘landmarking’ to provide a unified and intuitive approach. A key advantage of landmarking is that it is based on the familiar method of Cox regression modelling. However, landmarking has so far been limited to dynamic prediction and by lack of extensions to accommodate practical constraints.
The methods developed are being partly motivated by and also applied to data from the US and UK Cystic Fibrosis Patient Registries.
- Noemi Kreif
Medical Research Council
Early Career Fellowship in the Economics of Health
Improving statistical methods to address confounding in the economic evaluation of health interventions
Confounding is a major methodological challenge in health economic evaluations that use non-randomised studies. Currently recommended methods are not directly applicable for complex settings, such as when cost-effectiveness of a continuous or dynamic treatment regime is of interest. The causal inference literature proposes methods to address confounding in these settings. This fellowship aims to critically assess and extend these methods, to deal with the specific challenges of economic evaluation. Examples include the extension of the generalised propensity score method with machine learning, and the use of the synthetic control approach to evaluate health policies.
- Jenny Neuburger
National Institute for Health Research
I am carrying out research to evaluate improvements to hospital services for older people with a hip fracture, funded by an NIHR postdoctoral fellowship. Using large datasets from the Hospital Episode Statistics (HES) and the National Hip Fracture Database (NHFD), the aim is to find out which national policy changes, and which local changes to hospital services, have been effective in improving patient care and reducing mortality over the last decade.
The complex multilevel structure of the data allows several options for analysis. In this fellowship, I propose to adapt statistical techniques used for analysing complex data structures in other contexts. For example, techniques of multivariate meta-analysis (MV-meta) may be used to estimate and combine parameters that describe effects of organisational changes on mortality at the hospital level.
The HES database contains records of more than half a million hospital admissions for hip fracture over the last decade. This has been linked to ONS mortality data, which permits the analysis of deaths inside and outside hospital. It will also be linked to the NHFD, which contains more than 250,000 records with details of patients’ clinical care going back to 2007, and to hospital-level data on staffing and organisation of services.