Status: Ongoing
First registered on:
29/06/2023
Last updated on:
02/10/2023
1. Study identification
EU PAS Register NumberEUPAS105544
Official titleUse cases for development, optimisation and implementation of artificial intelligence methods for real world data analyses in regulatory decision-making and health technology assessment along the product lifecycle
Study title acronymReal4Reg
Study typeObservational study
Brief description of the studyThe use of real world data (RWD) is established in regulatory processes such as safety monitoring, but evidentiary value for further use cases, especially in the pre-authorisation and evaluation phase of medicinal products, is rudimentary. Also, the use of RWD in post-authorisation steps is constrained by data variability and by challenges in analysing data from different settings and sources. Moreover, there are emerging opportunities in the use of artificial intelligence (AI), but there is a lack of knowledge on its appropriate application to heterogeneous RWD sources to increase evidentiary value in the regulatory decision-making and health technology assessment (HTA) context. Thus, the development of new and optimised AI-supported methodologies for RWD analyses is essential. In addition, the four use cases to be investigated in this study contain phenotype-specific open questions of high regulatory interest.
Was this study requested by a regulator?No
Is the study required by a Risk Management Plan (RMP)?
Not applicable
Regulatory procedure number (RMP Category 1 and 2 studies only)
Other study registration identification numbers and URLs as applicable
2. Research centres and Investigator details
Coordinating study entity
Department/Research groupResearch/ Pharmacoepidemiology
Organisation/affiliationBfArM
Details of (Primary) lead investigator
Title Professor
Last name Haenisch
First name Britta
Is this study being carried out with the collaboration of a research network?
No
Other centres where this study is being conducted
Multiple centres
In total how many centres are involved in this Study?10
School of Pharmacy, University of Eastern Finland, Kuopio, Finland
Danish Medicines Agency, Data Analytics Centre, Copenhagen, Denmark
National Authority of Medicines and Health Products, I.P. (Infarmed), Lisboa, Portugal
Fraunhofer Society, Fraunhofer Institute for Algorithms and Scientific Computing (SCAI), Research Group AI and Data Science, Sankt Augustin, Germany
IT Centre for Science (CSC), Espoo, Finland
German Centre for Neurodegenerative Diseases (DZNE), Research Group Pharmacoepidemiology, Bonn, Germany
European Organization for Professionals and Patients with ALS (EUpALS), Leuven, Belgium
European Institute of Women’s Health (EIWH), Dublin, Ireland
Countries in which this study is being conducted
International study
Denmark
Finland
Germany
Portugal
3. Study timelines: initial administrative steps, progress reports and final report
PlannedActual
Date when funding contract was signed05/11/202211/11/2022
Start date of data collection01/09/202311/09/2023
Start date of data analysis
Date of interim report, if expected
Date of final study report31/12/2026
4. Sources of funding
Please provide estimates of the percentage of funding by source for this study
Names(s)Approximate % funding
Pharmaceutical companies
Charities
Government bodyHORIZON Europe (EMA)100
Research councils
EU funding scheme
5. Contact details for enquiries
Scientific Enquiries
Title Professor
Last name Haenisch
First name Britta
Address line 1Kurt-Georg-Kiesinger-Alle 3
Address line 2
Address line 3
CityBonn
Postcode53175
CountryGermany
Phone number (incl. country code)49228993075721
Alternative phone number
Fax number (incl. country code)
Public Enquiries
Title Ms
Last name Fernandes
First name Joana
Address line 1Parque da Saúde de Lisboa, Avenida do Brasil, 53
Address line 2
Address line 3
CityLisboa
Postcode1749-004
CountryPortugal
Phone number (incl. country code)351217987167
Alternative phone number
Fax number (incl. country code)
6. Study drug(s) information
Substance class (ATC Code)J01MA (Fluoroquinolones)
Substance class (ATC Code)A10BK (Sodium-glucose co-transporter 2 (SGLT2) inhibitors)
Substance class (ATC Code)A10BH (Dipeptidyl peptidase 4 (DPP-4) inhibitors)
7. Medical conditions to be studied
Medical condition(s)Yes
Breast neoplasm NOS
Type 2 diabetes mellitus
Amyotrophic lateral sclerosis
Additional Medical Condition(s)
Cardiac arrythmias, Peripheral neuropathies, Drug-induced liver injury, Cardiac failure, Sudden cardiac death
8. Population under study
Age
Adults (18 - 44 years)
Adults (45 - 64 years)
Adults (65 - 74 years)
Adults (75 years and over)
Sex
Male
Female
Other population
Renal impaired
Hepatic impaired
Immunocompromised
Pregnant women
9. Number of subjects
Estimated total number of subjects88000000
Additional information
Estimated sum of patients across all 4 countries based on each country's study period:
Breast cancer incident cases: 709,000 (+7.09 mil. controls);
ALS incident cases: 30,000 (+ 300,000 controls);
Users of antibacterials for systemic use: 88 mil.;
Users of SGLT2-inhibitors and DPP4-inhibitors: 5.43 mil.;
10. Source of data
Is this study being carried out with an established data source?Yes
Data sources registered with ENCePP
Data sources not registered with ENCePP
Registo Oncológico Nacional (RON, Portugal
Individual-level data collected from the Finnish national healthcare, mortality and population registers and censuses, Finland
Health Data Lab, Germany
Sources of data
Disease/case registry
Administrative database, e.g. claims database
Routine primary care electronic patient registry
Pharmacy dispensing records
11. Scope of the study
What is the scope of the study?
Disease epidemiology
Risk assessment
Drug utilisation study
Effectiveness evaluation
Methods development/improvement
Primary scope : Methods development/improvement
12. Main objective(s)
What is the main objective of the study?
The study will develop tools and technologies for the effective analyses of real-world data (RWD) in regulatory decision-making and HTA based on four highly relevant uses cases along the pre- and post-authorisation steps of the product life cycle. Portugal, Finland, Denmark, and Germany provide routine care RWD.
Are there primary outcomes?Yes
1. Investigate the potentials of European health claims data for describing study populations and extracting a historical control arm to be used in RWD study designs
2. Optimise/develop AI-based methods for study population selection, standardised result reporting, clustering of risk profiles etc
3. Knowledge for guideline improvement and training activities in regulatory authorities& HTA bodies
Are there secondary outcomes?Yes
Evidentiary insights into
1. natural history&epidemiology of ALS(amyotrophic lateral sclerosis) &BC(breast cancer)
2. changes in fluorquinolone(FQ) prescribing behavior(PB) & patient characteristics after FQ authorisation changes, estimated risk of ADRs before/after these changes
3. trends in SGLT2 inhibitor PB against the background of new labelling for heart failure, comparative effectiveness
13. Study design
What is the design of the study?
Cohort study
Case-control study
Drug utilisation study
External control arm study
14. Follow-up of patients
Will patients be followed up?Yes
Please describe duration of follow up
Patients are generally followed up starting in year 2000 in Denmark(DEN), Finland(FIN), and Portugal(POR), and starting in 2008 in Germany(GER). They are followed up to 2021 in FIN, POR, and GER and up to 2022 in DEN. A washout window will be applied starting in 2000 (2008 in GER), the length of which varies according to use case.
E.g., ALS patients: incident date=first diagnosis since 2005(2013)
15. Data analysis plan
Please provide a brief summary of the analysis method
- Define a common data model based on OMOP to harmonise metadata
- Develop a workflow to subset and display RWD by user definable inclusion/exclusion criteria
- Construct synthetic control arms using RWD by implementing different propensity score (PS) weighting and matching algorithms
- Apply AI based language models (ExMedBERT) to vectorize the data for subsequent analysis; possible alternative: traditional one-hot vector encoding of diagnosis and prescription codes.
- Predict drug safety related issues and post-marketing effectiveness: PS algorithms select two groups of statistically comparable patients, AI/ML algorithms are trained to predict the future occurrence of an event. Either classical ML algorithms (random forests, XGboost, CatBoost,…) are used or fine-tuning of ExMedBERT deep neural network is performed
- Develop generative AI algorithms for synthetic data generation applying AI based language models
- External validity is high due to use of RWD.
16. ENCePP seal
Are you requesting the ENCePP seal for this study?
Yes
17. Full protocol
18. Study Results
Not submitted
Please list the 5 most relevant publications using data from your study
ReferenceLink to web-publication
None
19. Other relevant documents
Other documentsNot
submitted
Signed Code of
Conduct Checklist
Submitted
Signed Code of Conduct Declaration
Submitted
Signed Checklist for Study
Protocols
Submitted
