Statistical Methodology Programme

Lead Research Organisation: MRC Clinical Trials Unit

Abstract

The statistical methodology group in the CTU works to benefit the ongoing clinical research programmes and from March 2009 has formed part of the London Hub for Trials Methodology Research based at CTU. This includes input both to the design and conduct of a trial or study and also to the analysis of the resulting data, leading to high quality data which is analysed and interpreted optimally. This input is provided at an expert level and where appropriate on the basis of methodological research. This research may involve comparing current statistical methods relevant to a particular scientific issue, updating or developing new methods if required, and recommending methods appropriate for use in clinical settings. The results of the research are reported in the journals read by medical researchers including medical statisticians. Where necessary the group will also produce computer programs to be disseminated widely through the medical statistics community. A good example is a freely-available program for complex sample size calculation which allows researchers to design studies to include an appropriate number of participants. The methodology underlying the program is more advanced than the methods typically used previously, allowing a wider range of study types, and the program has now been adopted for standard use at the CTU.

Technical Summary

Statistical methodology research in the CTU is driven by problems arising from the ongoing clinical research programmes and from March 2009 has formed part of the activity of the London Hub for Trials Methodology Research based at CTU. The careful, and where appropriate innovative, design and conduct of a trial or study lead to high quality data and hence results which are reliable, clear and understandable. However, expert and often innovative analyses are fundamental to optimal use of the data and to proper interpretation of the results, and finally, therefore, to maximum - including spaces impact on clinical practice. Furthermore, scientific questions may arise post hoc from trials (especially those with complex designs and several endpoints), and additional or novel analyses may give valuable insights beyond the original remit. Knowledge and sometimes development of appropriate methodology is therefore paramount. In general, statistical research at CTU may involve comparing current statistical methods relevant to a particular scientific issue, updating or developing new methods if required, assessing advantages and disadvantages, and recommending methods appropriate for use in clinical settings. The results of the research are reported in peer-reviewed clinical and/or biostatistical journals. Where necessary methods will be implemented as computer software within a standard statistical package disseminated widely through the medical statistics community, for example as freeware through the Stata Journal or the Web resource Statlib [see below for examples].|Current research themes include:|1) The use and extension of methods known as causal modelling in the analysis of trials and observational studies. For example, an analysis was performed with data from the DART trial to assess the causal impact of cotrimoxazole on mortality.||2) Methods to develop and validate prognostic and predictive models, under the general aim of stratified medicine. For example, methods were developed to create a single model concerning superficial bladder cancer from several MRC and EORTC trials and then validate that model with respect to the nine datasets||3) Developments in the design of trials, including multi-arm multi-stage trials, such as the ICON6 trial in ovarian cancer and the STAMPEDE trial in prostate cancer, and sample size calculation for complex designs, research which led to a freely available Stata package (ART Assessment of Resources for Trials), which has been adopted as a standard trial design tool at the CTU.||4) Methods to deal with missing data, including multiple imputation methods implemented in Stata package, and combined imputation and weighting methods for drop-out which when applied to the INITIO trial provided some reassurance that drop-out had not substantially affected the analysis of primary outcome.||5) More-flexible parametric methods to analyse survival data. Recent research has resulted in a paper proposing a new way of assessing treatment effects in trials with time-to-event outcomes but non-proportional hazards.||6) Extending current methods for the analysis of longitudinal data, for example we developed a method to assess how HIV-1 virus subtype influences CD4 response to starting HIV treatment

Publications

10 25 50
 
Title ART sample size 
Description software to calculate sample size in realistically complex scenarios 
Type Of Material Data analysis technique 
Year Produced 2006 
Provided To Others? Yes  
Impact software is used at Oxford university, widely within the MRC Clinical Trials Unit and is freely available 
 
Title Multivariable model-building 
Description A book on multivariable model-building: A pragmatic approach to regression analysis based on rfactional polynomials for modelling continuous variables 
Type Of Material Data analysis technique 
Year Produced 2008 
Provided To Others? Yes  
Impact Parctocal guide to regression analysis based on rfactional polynomials for modelling continuous variables 
 
Title Stata multiple imputation updates 
Description Updates to software in Stata to perform multiple imputation, a teachnique for dealing with missing data. Developed in collaboration with others. 
Type Of Material Data analysis technique 
Year Produced 2007 
Provided To Others? Yes  
Impact Used internationally by stats users when analysing medical research data 
 
Description PR & DA 
Organisation University of Oxford
Department Department of Statistics
Country United Kingdom 
Sector Academic/University 
PI Contribution Joint collaborative research
Collaborator Contribution Joint collaborative research
Impact 16947139, 16675816, 19477892
 
Description PR & IW, AC & SS 
Organisation University of Cambridge
Department MRC Biostatistics Unit
Country United Kingdom 
Sector Public 
PI Contribution Joint collaborative research
Collaborator Contribution Joint collaborative research
Impact 18203127, 19153970, and other publications
 
Description PR & PL 
Organisation University of Leicester
Department Department of Health Sciences
Country United Kingdom 
Sector Academic/University 
PI Contribution Joint collaborative research
Collaborator Contribution Joint collaborative research
Impact paper in Stata Journal, other submitted
Start Year 2008
 
Description PR & WS 
Organisation Albert Ludwigs University of Freiburg
Department Centre for Medical Biometry and Medical Informatics
Country Germany 
Sector Academic/University 
PI Contribution Joint collaborative research
Collaborator Contribution Joint collaborative research
Impact 18349388, plus several publications in the Stata Journal