Developing robust common data models to guide safety in medicines in Europe

The European Medicines Agency held a 2 day international workshop in London [11th -12th December 2017] to define the opportunities and challenges around implementation of a common health data model in Europe to support regulatory decision making. The expected outcome of the workshop was agreement of guiding principles for the development of a Common Data Model (CDM) in Europe, including key criteria for validation in the context of regulatory decision-making.

A common data model could help harmonise healthcare data across multiple data sets and provide a mechanism to conduct pan-European studies in a timely manner to address regulatory questions. At the same time, applying a common model to European data has multiple challenges. The meeting brought together regulators with academia, data holders and the pharmaceutical industry.

Sessions included talks from experts from North America (FDA, Harvard, Duke, Georgia Tech ...) and the European region (Erasmus, Utrecht, CBG-MEB, EMA ...) discussing lessons learned and current challenges in very large current clinical data resources, regulatory verification and related issues. Common data model case studies considered included Sentinel – the Harvard-based FDA system for accessing patient data from 16 health data partners across the USA and CNODES (the Canadian Network for Operational Drug Effect Studies) which can access data on 100 million patients – a similar scale to Sentinel.

The U.S. Food and Drug Administration's (FDA) Sentinel Initiative is a long term approach which uses a common health data model to improve the FDA’s ability to identify and explore safety issues for medical products. Sentinel actively surveys pre-existing electronic healthcare data from multiple sources.

Consistent themes included ensuring the relevance of evolving common data models to health policy, keeping timelines as short as practical, interoperability, consent and related ethical issues (data custodians, patient data protection and privacy), and careful internal and external validation of clinical definitions, data, software and analytical models.

From the perspective of health professionals, policy makers, regulators and the public, key questions included whether clinical outcomes from common data models are generalisable or only relevant to specific sub-populations based on geography, genetics, demographics and/or complex co-morbidity.  In the era of precision medicine there is the clear need is to avoid “right” answers from the wrong clinical populations and “wrong” answers from the right populations.

Further key points considered included: what is the cost of developing and maintaining validated CDMs; who should pay; whether updating existing databases is a sufficient approach or rather new more robust databases are needed.

See more about the workshop