PSI Conference Webinar: Intersection of Clinical Trials and Real World Data

Add to:

Sponsored by Amgen_4_Blue_PC

Date: Wednesday 10th June 2020       

Time: 10:00 - 11:30
Speakers: Elizabeth Williamson (Dept. of Medical Statistics), Christen Gray (IQVIA) & Kirsty Hicks (GSK)


This webinar is part of our 2020 Conference Webinar Series. Further information including details of other webinars that are included in the Conference package can be found here.
Members receive all webinars in the conference series for free.
Non-members receive all webinars in the conference series for £100+VAT, which includes complimentary membership* of PSI until the 31st December 2020. 

To register your place for this event, and others in the Conference webinar series, please click here.

Speaker Details:



Session Abstract

Elizabeth Williamson, 
Dept. of Medical Statistics.

Elizabeth Williamson is a biostatistician working in the Department of Medical Statistics. Her work focuses on statistical methods for addressing causal questions using electronic health records. Following an undergraduate degree in mathematics at King's College Cambridge and an MSc from the University of Leicester she undertook a PhD at the London School of Hygiene & Tropical Medicine (LSHTM) investigating the use of propensity scores to address confounding in observational data. From 2007 to 2014, Elizabeth lived in Australia, working in a range of biostatistical roles at the University of Sydney, Murdoch Childrens Research Institute, Monash University and the University of Melbourne. In 2014, she returned to LSHTM to take a position jointly between LSHTM and the Farr Institute of Health Informatics, London.

Using RWD to emulate trials.

Trials remain the gold standard for establishing harms and benefits of drugs. However, in certain circumstances it is useful to use observational data to attempt to emulate a particular randomised trial. This might be prior to running the trial in question, for the purposes of hypothesis generation or to obtain better estimates of the parameters required for power calculations. Alternatively, once the trial has been completed, trial emulation approaches might be used to extend results of the trials to relevant patient populations who are not included or less represented in the trial.

This talk will explore different approaches to trial emulation using RWD using two examples both using data from the UK Clinical Practice Research Datalink (CPRD), a large database of UK primary care records. In the first example, data from the CPRD was used to emulate a trial of macrolide antibiotics on all-cause mortality prior to the relevant trial being conducted. In the second example, data from CPRD was used to generalise the results of the TORCH COPD trial to a patient group less represented among the original trial participants –those with mild COPD.

Christen Gray

Christen is a Sr Consultant in Biostatistics for the RWS team at IQVIA. Prior to joining IQVIA, Christen spent four years at the Foundation for Innovation New Diagnostics (FIND) in Geneva, Switzerland as the Biostatistics and Data Manager in the clinical trials team. This team was responsible for the design, coordination, and analysis of multi-national clinical trials of diagnostics for tuberculosis, malaria, and other neglected diseases. Her work included the analysis and reporting for a WHO expert review submission for TB diagnostics. Prior work experience also includes field studies and analysis of interventions for the reduction of indoor air pollution from wood and charcoal cookstoves in developing countries at Berkeley Air in Berkeley, CA, USA. Christen also completed a PhD in Medical Statistics at the London School of Hygiene and Tropical Medicine (LSHTM). Her thesis focused on correction for exposure measurement error using Bayesian methods. Christen holds a Bsc in Molecular Biology from MIT as well as an MPH in Epidemiology & Biostatistics from UC Berkeley. She also has experience in the laboratory side of drug development after working in the Infectious Diseases Department atNovartis Institutes of Biomedical Research in Cambridge, MA.

Comparing the impact of unmeasured confounding due to selection bias in external comparator studies using RWD.

Background: Augmentation of the control arm of a randomized controlled trial (RCT) with external data has been proposed in recent years where standard RCTs face enrolment restrictions. Using real-world data (RWD) for external controls is a natural next step. However, in order to do so, there need to exist accessible methods for researchers which can minimize the risk from unmeasured confounding in this setting. Bayesian borrowing methods, which discount the external data dependent upon the similarity of the outcomes to the internal controls, have been applied when the external data is prior control arms of clinical trials. The simplest of these approaches is the power prior. In using RWD, greater variation in the underlying population and measured variables is expected.

Objective: To assess the ability of simple analytical methods to reduce bias stemming from unmeasured confounding due to selection bias in an augmented RCT.

Methods: A simulation study of an augmented RCT was performed with unmeasured confounding to assess the use of propensity score (PS) methods alone, the use of a normalized power prior method alone, and the use of a combined PS-adjusted power prior. Results: When no unmeasured confounding is present, the use of traditional PS minimizes bias as expected. In the presence of even a weak uncontrolled confounder, the power prior method is necessary to appropriately discount the external data to minimize bias.

Conclusions: Using a combination of simple analytical methods, rather than a single complex method alone, may provide a way for researchers to implement augmented RCTs.

Kirsty Hicks

Kirsty is a Senior Statistics Director at GSK, having joined in 2001 from Roche where she worked as a late phase statistician for a number of years. Throughout her time at GSK Kirsty has supported all gastrointestinal assets in the early phase portfolio covering diseases such as ulcerative colitis and irritable bowel syndrome; and supported numerous immunological-inflammatory assets across a wide range of diseases. Throughout this time Kirsty has managed an ever-growingteam of statisticians in addition to being heavily involved in numerous non-projectinitiatives. More recently these have included exploring experimental medicine analysis methods; facilitating prior elicitations and leading the biostatistics team investigating the incorporation and analysis of complex digital data into clinical trials. Kirsty is currently leading the UK Oncology Biostatistics team, and heads up the Cancer Epigenetics Biostatistics group.

Advanced Analytics of Digital Data: A focus on sensor data.

What is the evolving area of Digital data you may ask? Nowadays most of us will be wearing a watch monitoring some aspect of our movement or use phone apps that collect health information. The data collected can be referred to as digital biomarkers, and this data, in particular sensor technology is becoming a common feature in clinical trials. Digital data can be used to collect information on sleep patterns, respiration rate, step count and continuous monitoring of heart rate and energy expenditure. Collecting data through digital devices also increases patient engagement and can provide real time compliance monitoring. The regulations for use of digital technologies in clinical trials are still evolving, and current recommendations for the analysis of such data are limited. Utilising sensors, and collecting actigraphy data, is a non-invasive method of monitoring activity. An actigraph sensor is worn for a time period (eg a week or more) to measure activity. The movements the actigraph undergoes are continually recorded; where 300 data points per second can be collected. As a statistician initial questions that spring to mind include ‘How can this volume of data be analysed or represented visually?’ Obvious summary measures would include average step count over a day; or the percentage of time a patient is active or sedentary. But a lot of information is lost, what about the raw data? For example, the energy used at the minute level or even second level? Statistical methods that can be applied to this data are under investigation and this presentation will introduce some of these.


Upcoming Events

Latest Jobs