Innovative analytics

Bridging the gap between science and patient

Play video

Overview

Testing a potential new medicine in people for the first time is a key moment in drug development. By improving the design of our early clinical trials and the value of the information they give us, we are aiming to make better and faster decisions.

Through innovative analytics, we want to integrate and interpret ‘big data’ from different elements of our clinical trials, from genomic indicators to clinical signs and symptoms and response to treatment. We are also bringing our clinical trials to patients so they can participate in their homes.

Innovative analytics are helping us to manage the uncertainty we encounter when we first take our drug candidates into patients and they are contributing to the increased value of all the data we generate, with the aim of matching our potential drugs to those who are most likely to benefit.

 

Play video
01

Introduction

At AstraZeneca, we are changing the way we design and perform our early clinical trials – integrating innovative analytics and aiming to make studies smarter, shorter, smaller and more efficient.

“Innovative analytics help us ask questions in our early clinical trials with greater precision, so we can make better decisions. They allow us to use fewer patients and get results sooner, so we can accelerate our most promising agents, which is better for patients and more cost effective for us,” explains Professor Stephen Rennard, Vice President and Head of Early Clinical Development, AstraZeneca.

In one type of innovative analytics, we use previous data and models to predict likely treatment response. Earlier data analysis is assisting us in adapting trials and making important ‘next step’ decisions before trials are completed.

We are starting to incorporate artificial intelligence and machine learning to help us manage, interpret and ultimately integrate high density genomic, clinical and other data.

“These very dense datasets have the potential to revolutionise our clinical insights into diagnosis and treatment and, by linking them together, we can maximise their value,” says Rennard.

We are using real-time data analytics to monitor the safety and effectiveness of our novel candidate treatments and recruiting patients who actively contribute their knowledge and experiences. We are making our research more flexible for patients, so they can have tests and check ups at local clinics or even at home, instead of travelling to clinical trial centres.

 

 

Innovative analytics are the future, and they have taken us to the cusp of remarkable changes in the way we conduct our early phase clinical trials

Professor Stephen Rennard Vice President and Head of Early Clinical Development, AstraZeneca


02

Improving clinical trial design

At the heart of the change in our early clinical trial design are advances in our ability to predict the many possible outcomes of our studies and agree how we will respond to each of these scenarios – before the first patient even sets foot in clinic. 

“We now know the decision-making criteria for our trials before we even recruit patients, and we have plans in place to adapt over 80% of our early phase trials while they are underway, in response to accumulating data. In this way, we only take forward the most promising drug candidates, with the right doses and for the right patients,” says James Matcham, Head of Early Clinical Development Biometrics, AstraZeneca.

Being able to adapt the design of early phase trials in response to initial results is a major advantage to researchers as lack of effectiveness is the most common reason why drugs fail in clinical trials.
 

Adopting Bayesian principles to make best use of data

At AstraZeneca, we were early adopters of Bayesian principles which predict the probability of an event on the basis of previous findings. In our Phase I cancer trials, this approach helps us to make important decisions, such as choice of drug dose, based on data from previous high quality clinical studies. As patients start treatment, we update our predictions in line with their responses to treatment and, if necessary, revise our decisions.

“In the past, we started Phase I studies on the lowest possible dose, based on pre-clinical research, often knowing that it was unlikely to show any effect. Bayesian methodology is a superior way to predict and adapt dosing decisions, often for combinations of drugs, with the need for fewer patients,” says Matcham.

Bayesian dose response finding has proved so useful in Phase I and Phase IIa studies that we are now planning to use it for the first time in the critical Phase IIb studies when drug doses are finalised.

The Bayesian approach is also enabling us to reduce the number of patients recruited to the ‘control’ arms of standard or placebo treatment in some of our trials by as much as 50%.

By using historical control data, we can reduce recruitment time for Phase II trials and reduce the number of patients who do not get the new treatment. This is good for patients, as they have an increased chance of receiving the new candidate treatment, and it is good for us because it reduces the size of these studies

James Matcham Head of Early Clinical Development Biometrics, AstraZeneca


To reduce the number of patients on placebo or ‘control’ treatment in clinical trials worldwide, we contribute to the Transcelerate Placebo and Standard of Care (PSoC) Initiative for re-use of control data by other companies. 
 

Introducing platform trials to test multiple agents

Thanks to the strength and breadth of our pipeline of exploratory agents, we are increasingly establishing platform studies in our early clinical trials to compare multiple potential drugs for a single disease. For example, patients can be recruited to a platform study in cancer or cardiovascular disease according to results of biomarker tests for different sub-types of their disease. Depending on their response to initial treatment, they can move to other drugs on the platform indicated by their disease biomarkers, or stay on their initial drug.

Platform trials, such as our HUDSON and ORCHARD studies in oncology, may be advantageous for patients because they offer more opportunities to take part. For us, they make more efficient use of resources. Just as importantly, they bring additional understanding about the disease being treated and the biomarkers that indicate likely response to potential innovative treatments.

 


03

Making better decisions

By improving the design of our early clinical trials, enhancing our ability to simulate and predict outcomes and rapidly integrating preliminary clinical data to update our models, we are improving our decision making and enhancing the efficiency of our research.


Joint modelling: predicting long-term outcomes from early responses

Joint modelling is a novel way of predicting a key outcome of a clinical trial based on an earlier effect of treatment. In cancer trials, that could mean predicting overall survival from initial tumour shrinkage. Joint modelling uses advanced statistical analysis for essential ‘go, no-go’ decisions about whether to continue clinical trials of a potential new drug.

It does not remove the need for large, longer term confirmatory Phase III studies but it may enable us to make early decisions about starting other studies or formulation development, with potential for accelerating effective new treatments, and reducing expensive late-stage failure.

In collaboration with software experts at the University of Columbia, New York, we have developed joint modelling in oncology and it is also being explored in respiratory.
 

By working at the forefront of developments in this fast-moving field of joint modelling, we are committed to optimising the methodology and software as an open source tool, so that they can be made available and accessible to everyone working in this important area

Dr Craig Lambert Quantitative Clinical Pharmacology Enterprise Lead, AstraZeneca

Joint modelling is not the only technique being used to predict outcomes from early responses. In cardiovascular disease, we used a combination of statistical methods to predict outcomes for a novel agent to modify cholesterol levels in patients with heart disease, thus allowing Phase III studies to start early.

Real time analytics for 24/7 decision making

To further support rapid, robust decision making in our clinical trials, REACT (REal Time Analytics for Clinical Trials) is a cutting edge, data crunching system that is changing the way we interpret clinical trial results. The system has so far incorporated data from over 200,000 trial participants from some 50 trials in our key therapy areas, helping to support rapid, robust decision making. REACT is a major part of iDecide, our clinical informatics collaboration with Cancer Research UK Manchester Institute, at the University of Manchester and The Christie.

REACT processes all the efficacy and safety data in these trials, with fortnightly and monthly updates of results, and presents anonymised top-line and deep-dive data in an accessible format available 24/7 to our researchers.

“REACT is not only giving us a deeper, earlier understanding of the emerging balance between risk and benefit of a new drug candidate, it also has potential to change our culture and behaviour, so that we speed up our processes for sample analysis and data collection to both improve our analytics capabilities and to enable earlier clinical trial decision making,” explains Dr Dónal Landers, Clinical Leader for the iDecide Programme and Director of the digital Experimental Cancer Medicine Team (digitalECMT).

REACT is proving particularly useful in platform studies which test multiple compounds.

“With such rapid clinical insights of these new compounds, REACT is giving us confidence in deciding when to open and close treatment arms, and when we have enough evidence to support decisions for taking agents from Phase I platform studies into later stage development,” says Landers.


Incorporating genomic data into REACT

In the latest version of REACT, a new visualisation tool, called Oncoprint, has integrated genomic data from patient tissue and blood samples alongside clinical findings, so that we can dynamically generate and visualise an individual patient or a population genomic profile to understand its significance in the context of their response to therapy. Identifying gene signatures associated with tumour response allows a study to be adapted so the most appropriate patients are included. 


04

Advancing clinical science with big data

Genomics is now an integral part of recruiting patients to our clinical trials – from the earliest Phase I/II safety and dose-finding studies with small numbers of patients to the large Phase III studies which underpin regulatory approvals for new medicines.

“There is a general shift from recruiting patients according to solely clinical characteristics of their disease towards recruitment based on the molecular causes of disease. This helps us to assess which patients are most likely to benefit from potential drugs targeting those specific molecular causes,” explains Dr Slavé Petrovski, Vice-President and Head of Genome Analytics & Informatics, AstraZeneca.

Through our company-wide genomics initiative, we are aiming to analyse up to two million genome sequences by 2026, including 500,000 from our clinical trials. We are currently hitting sequencing milestones faster than anticipated, with 130,000 sequences already analysed and an anticipated 600,000 by the conclusion of 2019.

Making sense of the petabytes of genomics data processed and housed in our secure cloud-based storage environments is a huge task that our Centre for Genomics Research is addressing. Multiple cutting-edge technologies are needed, including sequencing technologies, to generate high quality raw genomic data from patient samples, advanced informatics to effectively process the raw data, and innovative analytics to interpret this in a way that has the potential to impact future drug development.
 

Linking genomic data to clinical outcomes

For patients in selected clinical trials who consent to genomic analyses, genetic data are linked to the patient’s clinical outcomes such as how well they responded to and tolerated treatment. By integrating anonymised genomic and clinical data from many hundreds to thousands of participants in a clinical trial programme, we are aiming to identify the genetic profiles that predict disease progression and response to treatment.

“The big advantage is that we not only learn why some patients respond to treatment while others don’t, we are also discovering more about the biology of the disease and the mode of action of our treatments – all of which helps in the design of new trials and the identification of new drug targets,” explains Dr Athena Matakidou, Head of Clinical Genomics at the Centre for Genomics Research at AstraZeneca, and Consultant Medical Oncologist at Cambridge University Hospitals.

As co-researchers at the cutting edge of genomic exploration, the patients who generously participate in our trials are protected by multi-layered security measures that ensure the confidentiality of their data at all stages of collection and analysis.

The Triangle Project: from genomics to multi-omics

Interpreting genomic data is just the first step in helping us improve the design of our early phase trials.

In the Triangle Project, carried out through a collaboration with Imperial College London, we are using new technologies to recognise patterns in very large amounts of radiomic, metabolomic and exosomal RNA sequencing data from completed clinical trials to guide patient recruitment to subsequent studies, based on likely benefits:

  • Radiomics digs deep into digitised data from radiological images of patient tumours or other abnormal tissues to detect smaller variations in shape, size, texture and volume than the human eye can discriminate.
  • Metabolomics uses data from samples of blood, tissue and even breath to obtain information on the metabolic processes that feed and energise cells and may change in disease or in response to treatment.
  • Exosomal RNA sequencing analyses expression of RNA shed from cells into body fluids as an indicator of their molecular blueprint which determines how those cells can function – in health and disease or in response to treatment.

 


The Triangle Project is linking patterns of data from these three sources with clinical outcomes from trials to attempt to discover novel indicators of response and tolerance to novel therapies. These indicators may then be used to enrich new studies with patients most likely to benefit, and enable those less likely to benefit to be considered for something more likely to help them.

 

By applying pattern recognition technologies to gain fresh insights on markers of efficacy and safety from previously completed, high quality clinical trials, we have the opportunity to make real-time decisions that help us to speed up and shorten new studies, optimising treatment opportunities for patients and reducing costs

Professor Hani Gabra Vice President, Clinical Discovery Unit at AstraZeneca, and Professor of Medical Oncology at Imperial College London.

05

Taking clinical trials into patients’ homes

Innovative analytics are not only changing clinical trials in laboratories and clinics, they are allowing us to take studies into patients’ homes.

Trials@home is a groundbreaking clinical research initiative that aims to give patients greater access to our clinical trials and the flexibility to participate from their homes, convenient health centres or conventional trial units – whichever they prefer. It is also designed to enrich data produced in early trials by collecting information more frequently.

“In conventional studies, we collect very little data between clinic visits, but trials@home may give us an opportunity to enhance the density of data we collect from each patient,” explains Natalie Fishburn, Head of Study Operations, Early Clinical Development, AstraZeneca.

In the last 12 months, we have started eight early phase clinical trials which allow patients to contribute some data from home, and within the next year, we expect to start the first fully remote early trial, with data sent by patients or local healthcare teams to trial centres.

 


Hybrid studies: partly in clinic, partly at home

At present most trials@home studies are hybrids, with some parts carried out in patients’ homes and some in trial centres. But if the results with techniques for use at home prove as accurate and robust as those used in trial centres, patients may be able to routinely choose where they want to participate – from home, local clinic or trial centre, or a mix of all three.

At the simplest level, patients are already using smartphone apps in cancer studies to remind them when to take medicines in what are often complicated dosing schedules for early clinical trials that might otherwise need to be repeatedly discussed at clinic visits.

At another level, participants in an early phase heart failure study in two countries are being offered mobile nurse visits at home for two of the 10 clinical appointments needed for the trial, instead of travelling to trial centres.

Developing a suite of home monitoring tests

Multiple innovative technologies are being investigated for the development of an industry-leading suite of home monitoring and sampling tests for day-to-day assessments. These include breath monitors to record cough around the clock, and monitors for blood pressure and mobility levels.

“We are developing a suite of home monitors that have the potential to be used across our clinical trials, some of which have already been validated and others we are taking steps to validate,” says Fishburn.

A pilot study is underway to compare the quality of data from finger-prick blood samples to those taken by traditional blood collection via a syringe for studies of absorption, distribution, metabolism and excretion of potential new drugs. The aim of these studies is to determine if a blood sample collected by a patient delivers the same results as traditional methods, as well as getting patient feedback on this type of involvement.

Patients will not only actively send us data, they may also be able to routinely wear or carry devices that automatically collect data during so-called ‘activity tracking’ of their ability to walk and exercise, for example if they have heart disease. This is an opportunity for monitoring the impact of treatment which has previously been unavailable to us.

 

The idea of patients taking samples and sending data to our clinics would have been science fiction when I started working in clinical trials, 21 years ago. But the children of today, who are constantly connected through their smartphones and tablets, are the potential patients, doctors and regulators of tomorrow and they may not find it acceptable to come into clinics and fill in paper forms. It is the reason we are integrating advanced, patient-friendly technologies as we evolve the clinical trials of the future.

Natalie Fishburn Head of Study Operations, Early Clinical Development, AstraZeneca