Models for selection bias, applied to controlled trials and observational studies of antiepileptic drugs

Lead Research Organisation: University of Warwick
Department Name: Statistics

Abstract

Epilepsy is a common disorder which affects about 3-4% of the populations over a lifetime: it often requires long term drug use. Information on the effectiveness and side effects of drugs is provided by various kinds of studies. It is not easy to combine the information from these different sources, as they refer to different groups of people with epilepsy. For example, response rates for an anti-epileptic drug, (gabapentin) range from 18% in controlled trials to 65% in studies which observed patients before and after they received the drug.

In order to provide reliable summaries across a variety of studies, methods to take into account the difference between patients and studies must be developed. Scientists are studying the issues using a variety of methods. This study concentrates on assessing treatments and side effects and models for the association between how well patients respond in clinical trials, and patients decisions to enter studies.

This research is important for doctors treating people with epilepsy, for managers within the National Health Service, and for scientists working in evidence-based medicine and evidence -based policy. The statistical methods can be used in health, social science and education.

Summaries of the results about anti-epileptic drugs will be made available through self-help groups, newsletters, doctors and specialist nurses. Summary information with also be available on the web, in association with the Department of Neurological Science at the University of Liverpool.

Technical Summary

Epilepsy is a common disorder which affects about 3-4% of the population over a lifetime; it often requires long term drug use. Information on the effectiveness and harms of drugs is provided by randomised controlled trials (RCTs), studies following RCTs, in which patients can change to active drug, and before & after observational studies. Response rates are often lower in RCTs than in follow-on and before & after studies (18%, 53% & 65% respectively for gabapentin). Case reports, cohort studies and case-control studies also contribute data on harms, as detecting harms requires data on the effects of long-term use.

Differentially selected populations raise substantial methodological problems of selection bias which complicate statistical analyses. Patients who enter the follow-on phase are self-selected and not necessarily representative of the main trial. Studies may be samples from different populations, and outcomes may be selectively reported, and published. These selection bias issues occur in many areas of statistics.

The aim is to develop better ways of assessing treatments and side effects in epilepsy studies, and to extend statistical methodology to cover selection bias problems in the analysis of experimental and observational studies. We will focus on problems relevant to anti-epileptic drugs, where selection bias is evident.

Objectives:
1. To synthesise information on treatment and side effects of anti-epileptic drugs provided by data from the different kinds of AED studies.
2. To extend existing work on selection bias to generic problems of differential selection of patients into studies and selection of outcomes.
3. To apply the models of selection bias developed to provide information on drug choice and dosing issues for clinicians, patients and purchasers of care.

People with epilepsy will benefit from better information on the benefits and risks of treatment, and lay leaflets summarising the findings will be used for dissemination. Clinicians and purchasers of care will be able to exploit the enhanced information for design of studies and assessment of treatment effects in chronic diseases, particularly for epilepsy. Statisticians, economists and social scientists will be able to exploit the new methods for investigating selection biases. Pharmaceutical industry will benefit from clarification of effectiveness and harm profiles, enhanced information for design of studies. The market shares of different companies might be variously affected.

Publications

10 25 50
publication icon
Al-Bachari S (2013) Gabapentin add-on for drug-resistant partial epilepsy. in The Cochrane database of systematic reviews

publication icon
Hemming K (2009) Intrapartum amnioinfusion for meconium-stained amniotic fluid: evidence for small study effect bias? in BJOG : an international journal of obstetrics and gynaecology

publication icon
Hemming K (2012) Bayesian sensitivity models for missing covariates in the analysis of survival data. in Journal of evaluation in clinical practice

publication icon
Hemming K (2013) Vigabatrin for refractory partial epilepsy. in The Cochrane database of systematic reviews

publication icon
Hemming K (2006) Long-term survival for a cohort of adults with cerebral palsy. in Developmental medicine and child neurology

publication icon
Hemming K (2008) Intrauterine growth and survival in cerebral palsy. in Archives of disease in childhood. Fetal and neonatal edition

publication icon
Hemming K (2011) Fetal growth and birthweight standards as screening tools: methods for evaluating performance. in BJOG : an international journal of obstetrics and gynaecology

publication icon
Hemming K (2008) Vigabatrin for refractory partial epilepsy. in The Cochrane database of systematic reviews

 
Guideline Title Diagnosis and management of epilepsy in adults
Description Cochrane reviews used as evidence
Geographic Reach National 
Policy Influence Type Citation in clinical guidelines
 
Guideline Title Diagnosis and Management of Epilepsy in Adults - A National Clinical Guideline
Description Cochrane reviews used as evidence.
Geographic Reach National 
Policy Influence Type Citation in clinical guidelines
 
Description NIHR Research Methods Opportunity Funding Scheme Application
Amount £28,260 (GBP)
Funding ID Award Ref: RMOFS 2012/02 
Organisation National Institute for Health Research 
Sector Public
Country United Kingdom
Start 11/2012 
End 04/2013
 
Title Meta-regression with partial information 
Description Methods for Meta-regression with partial information on summary trial or patient characteristics 
Type Of Material Model of mechanisms or symptoms - human 
Year Produced 2010 
Provided To Others? Yes  
Impact Cited in ten articles (excluding citation by Hemming). 
URL http://onlinelibrary.wiley.com/doi/10.1002/sim.3848/pdf
 
Title Vigabatrin benefits and harms data base 
Description Systematic review, detailed infromation on vigabatrin benefits and harms 
Type Of Material Database/Collection of data 
Provided To Others? No  
Impact The data and publications were relevant in a legal case on visual field restriction 
URL http://www.liv.ac.uk/translational-medicine/research/epilepsy
 
Description EpiPGX 
Organisation Imperial College London
Department Imperial College Trust
Country United Kingdom 
Sector Charity/Non Profit 
PI Contribution I contribute to discussion of research methods, and software developed by my former PhD student, Dr J K Rogers, is being extended by a current PhD student (BCLK) . Initial analyses of data have been provided to the EpiPGX WP02 analysis working group. December 2014: further analyses have been provided to the EpiPGx WP02 group, and a day agreed for Professor Hutton and BCLK to discuss the development of statistical models with statisticians from the group.
Collaborator Contribution Discussions of data and possible methods of analysis.
Impact Software is under development, and analyses will lead to journal articles on epilepsy and genetic variation.
Start Year 2012
 
Description EpiPGX 
Organisation The Walton Centre NHS Foundation Trust
Country United Kingdom 
Sector Public 
PI Contribution I contribute to discussion of research methods, and software developed by my former PhD student, Dr J K Rogers, is being extended by a current PhD student (BCLK) . Initial analyses of data have been provided to the EpiPGX WP02 analysis working group. December 2014: further analyses have been provided to the EpiPGx WP02 group, and a day agreed for Professor Hutton and BCLK to discuss the development of statistical models with statisticians from the group.
Collaborator Contribution Discussions of data and possible methods of analysis.
Impact Software is under development, and analyses will lead to journal articles on epilepsy and genetic variation.
Start Year 2012
 
Description EpiPGX 
Organisation University College London
Country United Kingdom 
Sector Academic/University 
PI Contribution I contribute to discussion of research methods, and software developed by my former PhD student, Dr J K Rogers, is being extended by a current PhD student (BCLK) . Initial analyses of data have been provided to the EpiPGX WP02 analysis working group. December 2014: further analyses have been provided to the EpiPGx WP02 group, and a day agreed for Professor Hutton and BCLK to discuss the development of statistical models with statisticians from the group.
Collaborator Contribution Discussions of data and possible methods of analysis.
Impact Software is under development, and analyses will lead to journal articles on epilepsy and genetic variation.
Start Year 2012
 
Description NIHR Programme Grant 
Organisation University of Liverpool
Department Centre for Health Evaluation and Medical Statistics
Country United Kingdom 
Sector Academic/University 
PI Contribution Statistical methodology, supervising research students and mentoring. 1. EPSRC funded student worked on analysis of epilepsy data, and development of further statistical models specifically for analysis of epilepsy data. 2. Conacyt funded student worked on analysis of epilepsy data, and development of further statistical models specifically for analysis of epilepsy data. 3. Metoring of two post-doctoral research fellows.
Collaborator Contribution Ongoing interaction with epilepsy research, with data and methodological challenges provided for PhD students, motivating further collaboration with statisticians in Lancaster University. Ongoing development of statistical methods, discussions with post-doctoral and other researchers in Liverpool.
Impact JK Rogers and JL Hutton and AG Marson and DW Chadwick (2012) Assessing the Risk of Subsequent Tonic-Clonic Seizures in Patients with a History of Simple or Complex Partial Seizures J of Neurology, Neurosurgery and Psych. 10.1136/jnnp-2011-300917 Disciplines: statistics, neurology JK Rogers and JL Hutton (2012) Joint modelling of pre-randomisation event counts and multiple post-randomisation survival times with cure rates: application to data for early epilepsy and single seizures J Appl. Stats 10.1080/02664763.2012.748720 Discipline: statistics
Start Year 2008
 
Description NIHR Programme Grant 
Organisation University of Liverpool
Department Department of Neurological Science
Country United Kingdom 
Sector Academic/University 
PI Contribution Statistical methodology, supervising research students and mentoring. 1. EPSRC funded student worked on analysis of epilepsy data, and development of further statistical models specifically for analysis of epilepsy data. 2. Conacyt funded student worked on analysis of epilepsy data, and development of further statistical models specifically for analysis of epilepsy data. 3. Metoring of two post-doctoral research fellows.
Collaborator Contribution Ongoing interaction with epilepsy research, with data and methodological challenges provided for PhD students, motivating further collaboration with statisticians in Lancaster University. Ongoing development of statistical methods, discussions with post-doctoral and other researchers in Liverpool.
Impact JK Rogers and JL Hutton and AG Marson and DW Chadwick (2012) Assessing the Risk of Subsequent Tonic-Clonic Seizures in Patients with a History of Simple or Complex Partial Seizures J of Neurology, Neurosurgery and Psych. 10.1136/jnnp-2011-300917 Disciplines: statistics, neurology JK Rogers and JL Hutton (2012) Joint modelling of pre-randomisation event counts and multiple post-randomisation survival times with cure rates: application to data for early epilepsy and single seizures J Appl. Stats 10.1080/02664763.2012.748720 Discipline: statistics
Start Year 2008
 
Description National Institute for Health Research Estimating time to next event for patients with recurrent events 
Organisation University of Liverpool
Country United Kingdom 
Sector Academic/University 
PI Contribution I am a mentor to the fellowship holder. We meet regularly to discuss theory and methodology relevant to the fellow's project, which addresses models for epilepsy data collected in clinical settings. We have submitted a joint article. I have included a former PhD student in the discussions.
Collaborator Contribution The fellow has considerably familiarity with data sets which inspire methodological developments.
Impact Manuscript submitted, but no outputs yet.
Start Year 2016
 
Description Expert witness, fetal anti-convulsant syndrome 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact With Dr Hemming, ongoing discussion and written reports on evidence for benefits and risks of anti-convulsant medication for women who are pregnant.

The other experts are from many countries, so I think it likely that other countries will be affected.
The case was eventually withdrawn.
Year(s) Of Engagement Activity 2008,2009
 
Description Expert witness, vigabatrin 
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 Other audience: mainly lawyers. I provided expert witness reports in a large class action case about the association between vigabatrin and visual field constriction. The patients who participated in the class action won compenstation.

I was not allowed to know the size of the settlement, nor whether this case affected other countries. However, a large pharmaceutical company was the defendant, I think it likely that other countries were affected. I was told that it was many millions.
Year(s) Of Engagement Activity 2006,2007
 
Description Oral Presentations 2006-2007 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Health professionals
Results and Impact Conference platforms; Maguire, M.J., Hemming, K., Hutton, J.L., Marson, A.G., 2006. Observational anti-epileptic studies: Can we trust them? Investigating publication and language bias. North of England Neurological Association meeting: Liversedge Prize presentation, Harrogate. Maguire, M.J., Hemming, K., Hutton, J.L., Marson, A.G., 2007. Overwhelming heterogeneity in systematic reviews of observational anti-epileptic studies, ILAE UK Chapter Gower's prize presentation, Southampton. Maguire, M.J., Hemming, K., Hutton, J.L., Marson, A.G., 2007. Reporting and evaluating efficacy outcomes from open-label extension studies of anti-epileptic drugs. XV Cochrane Colloquium conference proceedings, Cochrane collaboration, Brazil, 064/1.

Dissemination and discussion/ feedback on work
Year(s) Of Engagement Activity 2006
 
Description Poster Presentations 2006 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Health professionals
Results and Impact Conference presentations; Maguire, M.J., Hutton, J.L., Marson, A.G., 2006. The impact of including grey and non-English literature on effects estimates from non-randomized anti-epileptic drug studies. Epilepsia ILAE Finland, P189. Maguire, M.J., Hemming, K., Hutton, J.L., Marson, A.G., 2006. Observational anti-epileptic studies: Can we trust them? Investigating publication and language bias. XIV Cochrane Colloquium Conference Proceedings Dublin, P083.

Dissemination and critical review of work
Year(s) Of Engagement Activity 2006
 
Description Poster presentations 2007 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Health professionals
Results and Impact Conference presentations; Maguire, M.J., Hemming, K., Hutton, J.L., Marson, A.G., 2007. Interpreting efficacy outcomes from open-label extension studies of anti-epileptic drugs. American Epilepsy Society Conference Proceedings Epilepsia Philadelphia, P3.174.

Dissemination and critical review of work by peers
Year(s) Of Engagement Activity 2007
 
Description Poster presentations 2008 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Health professionals
Results and Impact Conference presentations; Maguire, M.J., Hemming, K., Hutton, J.L., Marson, A.G., 2008. Prevalence of visual field defects following exposure to vigabatrin: A systematic review. 8th European Congress on Epileptology, Berlin. Maguire, M.J., Hemming, K., Hutton, J.L., Marson, A.G., 2008. Prevalence of visual field defects following exposure to vigabatrin: A systematic review. Association of British Neurologists Annual Meeting, Aveimore.

Dissemination of work and critical review of work by peers
Year(s) Of Engagement Activity 2008
 
Description School visits, Kent, Geneva, Lausanne, Cornwall 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Schools
Results and Impact Talk to motivate students to study statistics, in order to address health and social issues.

Pupils engaged in discussion! The talks given in Switzerland were by invitation, and the invitations have been repeated.
Year(s) Of Engagement Activity 2007,2012,2014