Analysis of trials, meta-analyses and observational studies

Lead Research Organisation: University College London

Abstract

We cover issues such as survival analysis methods that allow for non-proportional hazards in both trials and IPD meta-analyses, modelling of prognostic and predictive factors, and analysis of longitudinal and clustered data. On the application of causal models, CTU trials are a rich resource for evaluating aspects of patient management other than the randomised comparison, such as the impact of second-line or concomitant treatments. Although often left to clinician discretion, the trial will usually collect information on the basis for these decisions. Exploiting such data in a major causal analysis of the DART trial in HIV, we showed that 24-weekly and 12 weekly CD4 monitoring give similar results, and that a single CD4 count at 48 weeks leads to better survival than no CD4 monitoring. Finally, on missing data, even modest amounts of missing data can lead to bias and make study conclusions unreliable and/or imprecise. Some methods to deal with it can lead to further bias or imprecision, yet prevail in many disease areas, and are recommended by some regulators. In collaboration with the London School of Hygiene and Tropical Medicine (LSHTM), we have proposed a new framework for the analysis of clinical trials with missing data, which has been adopted by a Drug Information Association working group and for a pharmaceutical company regulatory submission.

Technical Summary

Appropriate analysis of trials is vital to obtain a full return on the investment made through efficient design and excellent conduct. However, this is often far from straightforward. We are focusing on those which arise directly from our clinical studies. Thus we are investigating how to analyse multi-arm multi-stage (MAMS) trials, and how best to analyses time-to-event outcomes and recurrent events within trials, as well as using causal models to answer questions not addressed by randomisation. We are also examining the analysis, validation and handling of missing data in multivariable prognostic models; and one versus two-stage and network meta-analysis approaches using individual participant data. Clustered data often arise in trials and observational studies so another focus is the Appropriate analysis of longitudinal and clustered data.
 
Description Missing data course
Geographic Reach Local/Municipal/Regional 
Policy Influence Type Influenced training of practitioners or researchers
 
Description Missing data course (Philadelphia)
Geographic Reach Local/Municipal/Regional 
Policy Influence Type Influenced training of practitioners or researchers
 
Description Missing data course: LSHTM
Geographic Reach Local/Municipal/Regional 
Policy Influence Type Influenced training of practitioners or researchers
 
Description Missing data course: Swiss Winter Epidemiology School
Geographic Reach Europe 
Policy Influence Type Influenced training of practitioners or researchers
 
Description Missing data course: focus on electronic health record data
Geographic Reach Local/Municipal/Regional 
Policy Influence Type Influenced training of practitioners or researchers
 
Description Cost-effectiveness analysis with informative missing data: tools and strategies
Amount £283,777 (GBP)
Funding ID DRF-2015-08-047 
Organisation National Institute for Health Research 
Department NIHR Fellowship Programme
Sector Academic/University
Country United Kingdom
Start 10/2015 
End 09/2018
 
Description Cross-Unit Appointment: Using causal analyses to add value to large RCTs
Amount £185,000 (GBP)
Organisation Medical Research Council (MRC) 
Department MRC Population Health Sciences Research Network (PHSRN)
Sector Public
Country United Kingdom
Start 04/2014 
End 03/2018
 
Description MRC Methodology Research Panel: Missing data in propensity score analyses of Electronic Health Records Data
Amount £400,000 (GBP)
Funding ID MR/M013278/1 
Organisation Medical Research Council (MRC) 
Sector Academic/University
Country United Kingdom
Start 09/2015 
End 09/2018
 
Description MRC Methodology Research Panel: Multiple imputation by chained equations for data that are missing not at random
Amount £164,000 (GBP)
Funding ID MR/M025012/1 
Organisation Medical Research Council (MRC) 
Department MRC/NIHR Methodology Research Programme
Sector Academic/University
Country United Kingdom
Start 02/2016 
End 01/2019
 
Description Collaborative research with MRC BSU 
Organisation University of Cambridge
Department MRC Biostatistics Unit
Country United Kingdom 
Sector Public 
PI Contribution Joint collaborative research in design and analysis of clinical trials
Collaborator Contribution Joint collaborative research in design and analysis of clinical trials
Impact Several papers including PUBMED id: 19153970; 19452569; 21225900
 
Description Collaborative research with the University of Freiburg 
Organisation Albert Ludwigs University of Freiburg
Department Centre for Medical Biometry and Medical Informatics
Country Germany 
Sector Academic/University 
PI Contribution Joint collaborative research on the analysis of clinical trials
Collaborator Contribution Joint collaborative research on the analysis of clinical trials
Impact Several papers including PUBMED ID 20191601
 
Description Collaborative research with the University of Leicester 
Organisation University of Leicester
Department Department of Health Sciences
Country United Kingdom 
Sector Academic/University 
PI Contribution Joint collaborative research on analysis of clinical
Collaborator Contribution Joint collaborative research on analysis of clinical trials
Impact Several papers including in the STATA journal and PUBMED ID 20213719
 
Description Methodology Research Collaboration with industry 
Organisation Amgen Inc
Country United States 
Sector Private 
PI Contribution Commitment to developing research activity in to design, conduct or analysis methodology in areas of mutual interest. Detailed discussions ongoing
Collaborator Contribution Commitment to developing research activity in to design, conduct or analysis methodology in areas of mutual interest. Detailed discussions ongoing
Impact None yet
Start Year 2014
 
Description Methodology Research Collaboration with industry 
Organisation AstraZeneca
Country United Kingdom 
Sector Private 
PI Contribution Commitment to developing research activity in to design, conduct or analysis methodology in areas of mutual interest. Detailed discussions ongoing
Collaborator Contribution Commitment to developing research activity in to design, conduct or analysis methodology in areas of mutual interest. Detailed discussions ongoing
Impact None yet
Start Year 2014
 
Description Methodology Research Collaboration with industry 
Organisation GlaxoSmithKline (GSK)
Country Global 
Sector Private 
PI Contribution Commitment to developing research activity in to design, conduct or analysis methodology in areas of mutual interest. Detailed discussions ongoing
Collaborator Contribution Commitment to developing research activity in to design, conduct or analysis methodology in areas of mutual interest. Detailed discussions ongoing
Impact None yet
Start Year 2014
 
Description Methods for multiple imputation with multiple rating scales 
Organisation Bangor University
Department Institute of Medical and Social Care Research
Country United Kingdom 
Sector Academic/University 
PI Contribution Expertise and intellectual input
Collaborator Contribution Expertise and intellectual input
Impact None yet
Start Year 2014
 
Description Multiple imputation by predictive mean matching using multilevel models 
Organisation National Institute of Health and Medical Research (INSERM)
Country France 
Sector Public 
PI Contribution Expertise and intellectual input
Collaborator Contribution Expertise and intellectual input
Impact No outputs yet to report, collaboration still active. Manuscript submitted for publication - currently under review (Nov 2014)
Start Year 2014
 
Description Strengthening the Analytical Thinking for Observational Studies: STRATOS initiative 
Organisation Albert Ludwigs University of Freiburg
Department Centre for Medical Biometry and Medical Informatics
Country Germany 
Sector Academic/University 
PI Contribution Member of the steering committee (James Carpenter) and co-author of the paper in Statistics in Medicine which introduced this initiative
Collaborator Contribution Besides steering group membership, chair of the Topic Group on missing data, responsible for leading the activities of this group
Impact One paper has been published
Start Year 2014
 
Description Using causal analyses to add value to large RCTs 
Organisation University of Bristol
Department MRC Centre for Causal Analyses in Translational Epidemiology
Country United Kingdom 
Sector Public 
PI Contribution To help develop "best practice" examples to guide statisticians working in clinical academia in the use of causal methods To help systematically survey the clinical literature, and to conduct a scoping exercise to identify types of questions which causal models have most potential to answer in RCTs To help apply causal inference methods to use RCT data to answer two new clinical questions of different types To help disseminate best practice through workshops etc
Collaborator Contribution To help develop "best practice" examples to guide statisticians working in clinical academia in the use of causal methods To help systematically survey the clinical literature, and to conduct a scoping exercise to identify types of questions which causal models have most potential to answer in RCTs To help apply causal inference methods to use RCT data to answer two new clinical questions of different types To help disseminate best practice through workshops etc
Impact Funding from the MRC Population and Health Sciences Network (to fund a cross-unit appointment for a joint appointment between CTU and CaiTE) commenced in April 2014 (until 2017)
Start Year 2012
 
Title Jomo 
Description R package for multilevel joint modelling imputation 
Type Of Technology Software 
Year Produced 2016 
Open Source License? Yes  
Impact Use of the jomo method by applied researchers 
URL https://cran.r-project.org/web/packages/jomo/jomo.pdf
 
Title R package for multiple imputation of missing data in IPD meta-analysis 
Description This is a software package developed in R that enables multiple imputation of missing data in IPD meta-analysis 
Type Of Technology Software 
Year Produced 2015 
Impact No impacts as yet 
 
Title Software for substantive model compatible multiple imputation 
Description Statistical software for performing multiple imputation, consistent with the assumptions made by the substantive statistical analysis model. 
Type Of Technology Software 
Year Produced 2015 
Open Source License? Yes  
Impact None so far. 
URL http://www.missingdata.org.uk
 
Title Stata package for reference based sensitvity analysis for longitudinal trials with protocol deviation via multiple imputation 
Description This software implements, in the statistical software package Stata, Reference Based Sensitivity Analysis for missing outcome data in clinical trials. Within Stata, simply type ssc install mimix to install the software 
Type Of Technology Software 
Year Produced 2016 
Impact None yet 
 
Title fp_select: model selection for univariable fractional polynomials 
Description The fp_select routine to perform model selection for univariable fractional polynomial models 
Type Of Technology Software 
Year Produced 2017 
Impact None yet - just released 
 
Title mfpa: extension of mfp using the acd covariate transformation for enhanced parametric multivariable modelling 
Description An extension of mfp using the acd covariate transformation for enhanced parametric multivariable modelling 
Type Of Technology Software 
Year Produced 2016 
Impact Use of this software 
URL http://www.homepages.ucl.ac.uk/~ucakjpr/stata
 
Description EME workshop 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact 30 professionals attended a training course on efficacy and mechanisms evaluation with relevance to the MRC board of the same name.
Year(s) Of Engagement Activity 2017,2018
 
Description Marie Curie workshop 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact 40 people from a range of backgrounds attended a workshop "Missing Data in Palliative Care Studies" at which I spoke. This led to a draft document for use by the Marie Curie charity and others.
Year(s) Of Engagement Activity 2017
 
Description PSI course 2019 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact 50 members of PSI (European Statistical Organisation focussing on pharmaceutical industry) attended a pre-confernece course on "Demystifying causal inference"
Year(s) Of Engagement Activity 2018