Quality of Reports on Drug Toxicity in Eudravigilance: A Safety Physician's Perspective

European legislation requires Marketing Authorization Holders (MAHs) to continuously monitor Eudravigilance (EV) data and inform the European Medicines Agency and national competent authorities of validated safety signals. The process follows the Good Pharmacovigilance Practice Module IX and is based on the review of individual case safety reports (ICSR) both from the MAH's internal database and from EV. The data is reviewed and evaluated by a medically trained person, who should, based on the information provided in the case, determine the causal relationship between the suspect drug and the reported event. In order to do this with a certain degree of confidence, the case needs to report enough high-quality information on the drug, patient and adverse event, without significant confounders.

For this purpose, we performed an evaluation of data quality of ICSRs within the Eudravigilance database, focusing on drug toxicity cases for five commonly implicated substances: paracetamol, diazepam, fentanyl, quetiapine, and fluoxetine. A medically driven review was conducted on 500 randomly selected ICSRs from 2015 to 2024. A detailed quality assessment framework was developed and applied, scoring cases across several criteria (including data on suspect drugs, patient demographics, adverse event description, time to event, and case narratives), resulting in a maximum quality score of 25.

Main study findings revealed a generally low data quality, with an average score of 11.57 out of 25. Key quality deficiencies included improper classification of drugs as suspects (e.g., reporting concomitant medications or treatments as suspects), reporting of underlying diseases and indications as adverse events, lack of information on patients' medical history and missing time-to-event information. Cases from non-European Economic Area (non-EEA) countries and consumer-reported cases exhibited the lowest quality, while regulatory agency-reported cases were of higher quality. The study also identified a frequent misclassification of non-prescription or illicit substances (e.g., fentanyl) as prescription products, complicating signal detection and causality assessments. The analysis highlights a very important gap in pharmacovigilance signal detection and evaluation processes, underscoring risks for misleading results, increased workload, and potential misinterpretation of product safety profiles. The results highlight the need for enhanced case reporting trainings, improved quality control, better follow-up processes, and a collective mindset shift across stakeholders to prioritize data quality.

In conclusion, significant improvements in the completeness, accuracy, and clinical relevance of ICSRs are essential to support effective safety signal detection and benefit-risk assessment in the post-marketing surveillance of medicinal products.