EU: Lessons Learned from PROTECT on Common Protocols for Multi-Database Studies

Insights into common protocol studies – including a list of recommendations – which were conducted in the framework of the PROTECT project.
Presented by: Olaf Klungel, Utrecht Universiteit
Presented at: ENCePP Plenary Meeting, 24 November 2015

PROTECT GOAL

To Strengthen the monitoring of benefit-risk of medicines in Europe by developing innovative methods
To enhance early detection and assessment of adverse drug reactions from different data sources (clinical trials, spontaneous reporting and OBSERVATIONAL STUDIES)
To enable the integration and presentation of data on benefits and risks

These methods are being tested in real-world situations.

Explain differences in drug-adverse event associations due to choices in methodology and databases
Replication program of studies
- Same study EU database
- Different study database, specifically US data source

(Abbing-Karahagopian V, et al. Curr Clin Pharmacol 2014;9:130-138)

Common protocol for each drug-AE pair
Extensive sensitivity analyses on main methodological issues
Common standards, templates, procedures
Detailed data specification including definitions of exposures, outcomes, and confounders for each database.
Blinding of results of in-parallel and replication analyses
Stepwise unblinding after completion of each design
Registration of protocols at ENCePP to guarantee transparency

Substantial time (+/-1 year) reach agreement on common protocol – Consensus/buy-in on approach between stakeholders!
Research question determined choice of database and design of data collection
Detailed data-specification documents are needed to harmonize procedures and analyses
Frequent communication between research centres to reduce variation in “interpretation” of protocol
During programming and analysis phase further clarification is required and needs documentation

Develop common protocol with great detail to reduce methodological differences and “interpretation” by researchers
Solid infrastructure fo rcommunication/collaboration
Conduct analysis in parallel in multiple DBs versus “apriori” pooling of DBs
Cherish heterogeneity and explore its sources
To test robustness of findings conduct multiple sensitivity analyses:
Multiple designs (e.g. Cohort/case-control vs case-only)
Exposure (e.g. Individual AEDs), outcome (e.g SUI), confounding adjustment
Replication needed if parallel analysis consistent?