Upload
others
View
12
Download
2
Embed Size (px)
Citation preview
Medidata Using AI & Analytics on Clinical Trial DataLife Science Forum Basel7th June 2018
2© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
Principal, Data & [email protected]
Sean McDonnell
3© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
RAVEEDC
MEDS: Clinical Trials Data Repository
Expertise Advance Clinical Development
Medidata Solutions
4© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
*Optional footnote goes here.
“…computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making…”
English Oxford Living Dictionary
Artificial Intelligence
5© 2018 Medidata Solutions, Inc. – Proprietary and Confidential © 2018 Medidata Solutions, Inc. – Proprietary and Confidential 5
Streamlining genomic biomarker discoveryRave Omicsw/ Biomarker Discovery
Outlier detection to ensure data quality
Edge Central Monitoringw/ Machine Learning &Edge Trial Assurance
Mitigate go/no-go decision risks
MEDS Synthetic Control Arms & Synthetic Control Data
Medical coding faster & more accurate via NLP
Rave Coder
AI already transforming clinical research
6© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
63.2%
30.7%
58.1%
85.3%
9.6%
0%
20%
40%
60%
80%
100%
Phase I to II Phase II to III
Phase III to NDA/BLA
NDA/BLA to Approval
Phase I to Approval
Success Rates 2006-2015(N = 9,985)2
© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
6
~10 YearsLong timelines1
~$2.6 BillionHigh development cost3
Industry performance will improve through AI
FDA ApprovedLow success rate
Source: 1. Biopharmaceutical Research & Development:, The Process Behind New
Medicines, PhRMA, 20152. Clinical Development Success Rates 2006-2015, BIO, Biomedtracker, Amplion3. Tufts Center for the Study of Drug Development, November 2014
7© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
76.7%
46.7%
76.5%
94.5%
25.9%
63.0%
28.8%
55.0%
83.9%
8.4%
0%
20%
40%
60%
80%
100%
Phase I to II Phase II to III Phase III to NDA/BLA NDA/BLA to Approval Phase I to Approval
With Selection Biomarkers
Without Biomarkers
Source: 1. Biotechnology Innovation Organization (BIO), Biomedtracker, AMPLION (2016, May 25). Clinical Development Success Rates 2006-2015. Retrieved January 20, 2018, from
https://www.bio.org/sites/default/files/Clinical%20Development%20Success%20Rates%202006-2015%20-%20BIO,%20Biomedtracker,%20Amplion%202016.pdf
Clinical trials with molecular biomarkers are better — 3X success rates
8© 2018 Medidata Solutions, Inc. – Proprietary and Confidential © 2018 Medidata Solutions, Inc. – Proprietary and Confidential 8
Only a small number of genomic biomarkers have reached clinical application, many more to be found!
9© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
Upload Link MachineLearning
BiomarkerDiscovery
Rave Omics
10© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
10
Rave Omics has already found evidence of new genomic biomarkersSponsor 1: Found gene associated with treatment response in oncology trial
Sponsor 2: Pharmacovigilance: Found strong evidence that mutations impact AEsData quality, scientific discovery, AE correlations on oncology trial
High specific gene levels in patient, relevant for drug in trial and response
Possible new disease sub-type
• Two specific genes reported as top signatures for metastatic melanoma
Sample biopsy issue where liver cells included with tumor
11© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
Monthscleaned and QC’ed data
3
Rave Omics discovered 6 patient subsets, with 3x differential response rate
Yearscollect data
8 MinutesIntegrated data
30 MinutesNovel patient clusters
15
12© 2018 Medidata Solutions, Inc. – Proprietary and Confidential © 2018 Medidata Solutions, Inc. – Proprietary and Confidential 12
improve data quality machine learning
13© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
13
Edge Central Monitoring w/ Machine Learning
Analyzes Millions of Data Points and Finds Thousands of Patterns
14© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
14
Using machine learning to find errors, some that humans missed
• Highly correlated measurements
• Anomalous data identified
One site looks different
15© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
Automating Medical Coding with Rave Coder & NLP
96.6%For Adverse Events
97.8%For Con Meds
Coverage 100% Coverage 100%
16© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
“Headache” going from verbatim to preferred
17© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
ThePromiseofPrecisionMedicine
Efficacy and Safety studyModerate asthma• No effect on FEV1 and
only a small numerical trend on exacerbations
Investigator-sponsored trials• Severe asthma with eosinophilic
inflammation
1997
IND opened
1999
Three pivotal Efficacy and Safety studiesSevere asthma with eosinophilic inflammation• Two studies – dramatic
reduction in exacerbations• One study – reduction in
corticosteroid use possible
2009 - 2012 2015
FDA ApprovalIn asthma patients with eosinophilic inflammation
KeytoProgramSuccessIdentificationofhigh-respondingsubgroup
No Go Go!
© 2018 Medidata Solutions, Inc. – Proprietary and Confidential 5
18© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
Synthetic Controls streamlines clinical development and mitigates Go/No-go decision risk
19© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
19
MEDS Synthetic Control Products
Product Details: • An aggregated dataset of hundreds (or
thousands) of patients in an oncology indication
• Selected from multiple recent trials chosen to match key customer inclusion/exclusion criteria
• Delivered via a visualization application allowing customer interaction with pre-built biostatanalyses and dynamic population subsetting
Product Options:
• MEDS Synthetic Control Arm (SCA)• MEDS Synthetic Control Data (SCD)
Synthetic Control Visualization Tool Familiar statistical views tailored to research question
Subgroup drill-down in every view
Comparison between experimental treatment arm and synthetic control group
© 2018 Medidata Solutions, Inc. – Proprietary and Confidential 120
21© 2018 Medidata Solutions, Inc. – Proprietary and Confidential © 2018 Medidata Solutions, Inc. – Proprietary and Confidential 21
Data Capture & Management
Trial Planning & Management
Study Design & Feasibility
Mobile Health
Images Genomics
RWE
Site &Lab Data
Random-ization & Supply
Biostats& Safety
Site Payments & Grants
Strategic Monitoring
CTMS
eTMF
RAVE
22© 2018 Medidata Solutions, Inc. – Proprietary and Confidential
Thank you.Visit us during the break to learn more!
Principal, Data & [email protected]
Sean McDonnell