DARWIN EU completed the first studies and needs new data partners

Yesterday, the EMA announced the Open Call for applications for new data partners to support real-world data evaluation through DARWIN EU. Additionally, three studies have been completed evaluating prevalence of rare hematological cancers, use of valproate and antibiotics. A fourth study evaluating serious adverse events in asthma is still ongoing.

Background on HMA-EMA’s joint Big Data Initiative

• The Big Data Task Force initially provided 10 priority recommendations for developing competencies and uses for big data. The task force, established in 2017, aims to explore how big data could be used to support research and innovation to expedite the development of new medicines. It operated in two phases: Phase 1 collected the necessary information on capabilities, expertise and challenges in utilizing big data. Phase 2 translated those findings into a final report and 10 actionable priority recommendations. The priority recommendations were meant to guide the work to develop ““a strengthened regulatory system that can efficiently integrate data analysis into its assessment processes to improve decision-making.” [For an update on Big Data activities, please read AgencyIQ’s 2021 DIA meeting report here.]
• A refresher: According to the EMA, Big Data are defined as “extremely large datasets which may be complex, multi-dimensional, unstructured and heterogeneous, which are accumulating rapidly and which may be analysed computationally to reveal patterns, trends, and associations. In general, big data sets require advanced or specialized methods to provide an answer within reliable constraints.”
• Real-world Data (RWD) is “an umbrella term for data regarding the effects of health interventions (e.g. safety, effectiveness, resource use, etc) that are not collected in the context of highly-controlled RCT’s. Instead, RWD can either be primary research data collected in a manner which reflects how interventions would be used in routine clinical practice or secondary research data derived from routinely collected data.”
• Real World Evidence (RWE) is “the evidence derived from the analysis and/or synthesis of real-world data (RWD).”
• In August 2021, DARWIN EU was highlighted by the EMAas one area of focus. DARWIN EU is the Data Analytics and Real-World Interrogation Network, which the EMA and HMA believe will speed patient access to innovative medicines. As designed, DARWIN EU is a federal network of data holders working under common governance using a common data model, and a priority deliverable of the workplan which supports delivery of the EMA Network Strategy to 2025. [See AgencyIQ’s analysis of the updated workplan].
• The DARWIN EU Coordination Centre was established by EMA at the Erasmus University Medical Center Rotterdam in February 2022. Erasmus will serve as a contractor, working with the EMA to set up the infrastructure for the Coordination Center. The center will “develop and manage a network of real-world healthcare data sources across the EU” and provide the necessary business services. [See AgencyIQ’s analysis of the launch of coordination center].
• In November 2022, the EMA announced the selection of only eight data partners including both public and private institutions. All selected partners have “access to real-world healthcare data from one or more sources such as hospitals, primary care, health insurance, biobanks, or disease-specific patient registries,” according to the announcement. The selected data partners include a mix of data hubs, a biobank, a disease registry and two universities: 1) Barcelona MAR Health Park (IMASIS, Spain), 2. University Hospital of Bordeaux (CHUBX, France), 3. Auria Clinical Informatics (ACI VARHA, Finland), 4. Estonian Biobank, 5. Integrated Primary Care Information Project (IPCI, Netherlands), 6. Netherlands Comprehensive Cancer Organisation ( IKNL), 7. Information System for Research in Primary Care (SIDIAP, Spain) and 8. Clinical Practice Research Datalink GOLD (CPRD GOLD, U.K.). IQVIA Disease Analyzer (Germany) and IQVIA Longitudinal Patient Database (Belgium) joined as data partners during Phase 1. [See AgencyIQ’s analysis of the updated workplan and data partner selection.]

DARWIN EU is now ready to select new data partners

• Today, the EMA announced that it plans to recruit 10 more data partners for DARWIN EU. Data custodians are invited to respond to the Open Call for Data Partners. These partners can be from public or private institutions that hold real-world data such as primary and secondary care data, claims databases, biobanks, and registries. Interested parties must complete an Expression of Interest (EoI) form and email it to [email protected]. All accepted EoIs will be evaluated at the next review dates of either April 30 or October 31, 2023. However, applications will be accepted on a continuous basis.
• Starting the onboarding process does not mean the interested custodian is an official DARWIN EU data partner. The DARWIN EU Coordination Center (CC) may contact interested parties with requests for more information on the submitted application, but this does not mean that party has been selected. Shortlisted interested data partners will start the “onboarding process” of technical and quality checks to ensure “readiness for DARWIN EU studies.” Data partners are only officially a member of the network once the onboarding process has been completed successfully and an agreement has been signed with the CC.
• Interested data partners will remain on the list of potential data partners. Applicants not selected in a review round will remain on the list and could be selected in future rounds of selection. The review will evaluate data partners on specific factors such as size, type of data, data quality, geographical coverage, and added value to the network. Interested data partners will only be removed from the list of applicants when they request the removal.
• The move to enlist more data partners comes as DARWIN EU published its first round of studies, which “start to demonstrate the benefits” of the collaboration, according to an EMA announcement. Topics for the studies were chosen based on previous RWE requests from committees of the EMA.

The first study evaluated the prevalence of rare blood cancers in five E.U. countries

• Improved survival and aging population have increased the prevalence of rare cancers, leading to uncertainty during orphan drug designation assessment. There must be less than 5 cases of a specific condition in 10,000 E.U. persons at the time of orphan designation application to meet the prevalence criterion for the designation, and over 40% of orphan designations from 2000-2015 were granted for rare hematological cancers. However, true prevalence wasn’t known because of a lack of epidemiological studies, meaning that regulators had to turn to less-certain measures to calculate prevalence. Additionally, longer life expectancy may mean higher prevalence of previously rare conditions.
• The DARWIN EU study aimed to estimate prevalence in five hematological cancers: follicular lymphoma (FL), diffuse large B-cell lymphoma (DLBCL), multiple myeloma (MM), chronic lymphocytic leukemia (CLL), acute myeloid leukemia (AML) and acute lymphocytic leukemia (ALL). A 5-year partial prevalence was used for primary analysis. The study data sources included the IPCI, SIDIAP, CPRD GOLD, Longitudinal Patient Database (LPD, IQVIA, Belgium), and the Disease Analyser (DA, IQVIA Germany).
• All evaluated cancers still met orphan designation prevalence requirements, according to the 5-year partial prevalence technique used in the study. CLL and MM had the highest prevalence, but still were lower than the required 5 in 10,000.
• Outcome reporting, uncertainty on outcome definitions and how long after diagnosis a person is still considered a prevalent case are key considerations. The report of study results noted some underreporting of “outcome events of interest,” although the databases were regarded as strong. Results from datasets with patient-level hospital linkage and those without were comparable. The findings aligned with those of the Surveillance, Epidemiology, and End Results (SEER) Program at the National Cancer Institute and the Haematological Malignancy Research Network of the U.K. Although the study drew from patient data in several E.U. countries, patient and health system factors may impact prevalence in other E.U. countries, limiting generalization of the data from this study.

The second study evaluated use of valproate-containing medicines in women of childbearing potential

• Valproic acid/valproate-containing medicines (VPA) have dose-dependent teratogenicity and should be limited in women of childbearing age. VPAs are used for seizure disorders and bipolar disorder, as well as for migraine prevention. VPA exposure prenatally can cause “neurodevelopmental impairments and congenital malformation,” according to the report for this study. In 2014, the EMA’s risk mitigation actions limited use of VPA and derivatives in women of childbearing age to patients with intolerance of alternatives or for whom the alternatives were ineffective. EMA’s Pharmacovigilance Risk Assessment Committee (PRAC) implemented new use restrictions and a pregnancy prevention program in 2018.
• The EMA requested a DARWIN EU study to evaluate population-level and patient-level VPA utilization. The study aimed to address two objectives: 1. “to characterise the prevalence and incidence of use of VPA and alternative antiepileptic therapies among women aged 12 to 55 years of age.” and 2. “to characterise the use of VPA” in the same population. VPA use was evaluated for 2010 to 2021 using datasets from ACI VARHA, CPRD GOLD, IPCI, SIDIAP, IQVIA Belgium LPD, and IQVIA Germany DA.
• Overall, incidence of valproate use declined, while the prevalence initially declined and then remained stable during the years studied. New VPA treatment initiation (i.e., incidence) decreased in all datasets throughout the study period. Overall, prevalence (i.e., proportion of women using VPAs during a specific time) decreased for the first few years and stabilized at the lower level for the last few years. Generally, prevalence in women aged over 45 years was 2 to 3 times higher and decreased less notably over the study period than prevalence in younger age groups. Observed differences in systemic hormonal contraception use across data sets may be due to different national reimbursement practices.
• Data quality and incomplete recording of prescribing information pose study limitations. Since the study used “routinely collected health data,” the report noted that data quality was a consideration. Prescriptions do not indicate actual use and dose information was not available for certain datasets, necessitating assumptions regarding actual use and duration of use. Additionally, the indication for use was inferred from a proxy based on pre-defined conditions noted on the therapy initiation data, since the reason for prescribing VPAs was not noted in the database. Individual datasets also had limitations. For example, Finland’s Social Insurance Institution required valproate prescribing before stepping up to any other anti-epileptic drugs, resulting in prescriptions for never-used valproate that were issued solely to circumvent the rules. This report also highlighted that results are representative for the specific study population in the countries evaluated but cannot be generalized.

A third study characterized the use of antibiotics on the World Health Organization Watch List

• Antibiotic stewardship is necessary to limit the risk of antibiotic resistance development. The World Health Organization’s (WHO’s) AWaRe Classification of antibiotics provides a tool for antibiotic stewardship by classifying these medicines to treat bacterial infections into three groups: Access, Watch and Reserve depending on the specific antibiotic’s impact on antimicrobial resistance. WHO stresses the importance of appropriate use for the 258 products currently on this list. For example, the antibiotics on the Watch list have higher resistance potential.
• This DARWIN EU study evaluated the use of antibiotics on the Watch list during routine healthcare and use trends over time. The study (report) had two objectives: a) “to investigate the incidence and prevalence of use of antibiotics” on the Watch list and b) “to explore duration of antibiotic use” and indication for prescribing. Watch list antibiotic use was evaluated for 2012 to 2021 using datasets from CPRD GOLD, IPCI, SIDIAP, CHUBX, IMASIS and IQVIA Germany DA.
• Incidence of antibiotic prescribing was generally below 100 per 100,000 persons per year with few exceptions. In most databases the use of ciprofloxacin, clarithromycin, fosfomycin and azithromycin was higher in primary care. A higher use of ceftriaxone, vancomycin and meropenem was seen in secondary care. “The incident rate remained stable or decreased over time” except for some antibiotics used in secondary care (e.g., ceftriaxone, vancomycin). Antibiotic use incidence generally increased with age but was comparable by sex.
• The study limitations were very similar to those in the VPA study. The report noted data quality issues in “routinely collected health care data.” Similarly, use of prescription data meant researchers had to make assumptions about actual use and duration of use; further, indications had to be inferred through a proxy based on pre-defined conditions. Additionally, the number of prescriptions with missing or unknown indications was high in this study, and not all infections fell into the SNOMED dictionary categories. This report also observed that findings can’t be generalized across the E.U. since antibiotic use varies from country to country.

A fourth study, still in progress, assesses serious adverse events in severe asthma

• This study was prompted by a clinical trial where the rate of serious adverse events differed between the investigational and control arms, calling for a population study. The retrospective cohort study will determine “background rates of selected health outcomes in patients with severe asthma,” to provide context for these unexpected findings. The overall prevalence of asthma is 9.8%, of which about 3 – 10% of patients have severe uncontrolled asthma. This non-interventional study will evaluate rates of severe adverse events in a broad population to potentially inform future safety assessments.
• The study will examine four outcomes in three different cohorts according to the protocol. The four outcomes include a) all-cause mortality, b) mortality from fatal infection, c) mortality from cardiovascular events and d) the incidence rate of serious cardiovascular events. Individuals with severe asthma and at least one year of history present in the database between 2015 and 2021 will be evaluated in three cohorts: patients with severe asthma fulfilling 1) inclusion criteria, 2) inclusion and exclusion criteria and 3) inclusion criteria plus at least one exclusion criterion. The datasets include CPRD GOLD, IPCI, SIDIAP, Parc Salut Mar Barcelona (PSMR), Hospital del Mar (IMASIS, Spain), and the Estonian Biobank.
• Study results are expected in spring 2023.

What’s next

• DARWIN EU expects to onboard an additional ten data partners throughout 2023. Additionally, it hopes to initiate another 16 studies. Interested data custodians should complete the EoI and apply to become a data partner. Stakeholders should keep in mind a) the two selection dates in April and October and b) the process involved in selecting partners.
• Although the three completed studies provide some insight on the respective study objectives there are still various limitations. Data quality and completeness raised some considerations, although some of these limitations are inherent to the use of certain types of RWD, such as using prescribing information as a proxy for medication use. There are other data sources, such as prescription fill and refill data from pharmacies, that can come closer to actual medication use; recognizing the strengths and limitations of various data sources through these and other RWE studies may further inform the buildout of the European Health Data Space to optimize its use for research.
• The relatively small number of data sets may limit the interpretation and generalization of data. Last year, DARWIN EU selected eight of the planned ten data partners. However, the studies thus far have only uses five or six of the selected data partners. For each study, investigators note that data can’t be generalized across other populations not studied and the E.U. due to national differences – a standard disclaimer in medical research, but one that use of large amounts of diverse data can overcome, at least partially. Presumably, study findings will become more generalizable once more data partners with a wider geographic spread representing the E.U. population have been onboarded and are being used in DARWIN EU studies.
• DARWIN EU is expected to deliver 150 real-world evidence studies per year by 2025. EMA further notes that “by then use of RWE [real-world evidence] will have been enabled and value established across the spectrum of regulatory use cases.” This ambitious goal would require that most of the limitations of currently available data are overcome. For real-world evidence developed through DARWIN EU to have its greatest possible utility in regulatory decision making, the studied population should be as similar as possible to that used for the evaluation of the investigational medicinal product. Differences in population characteristics and regional health care utilization differences, as well as data quality issues, will otherwise impede the use of such data to support regulatory applications.

To contact the author of this analysis, please email Kirsten Messmer.
To contact the editor of this analysis, please email Kari Oakes.

Key Documents and Dates

• EMA press release – DARWIN EU has completed its first studies and is calling for new data partners – published March 28, 2023
• DARWIN EU – How to join the data network? and Open Call for new data partners – last updated March 29, 2023
• EMA website – Data Analysis and Real World Integration Network (DARWIN EU) – last updated March 28, 2023