Disproportionality signal detection methods can help identify potential adverse outcomes of medication exposure. Such methods scan all medication–outcome pairs within a dataset to identify possible associations requiring further investigation. Many signal detection methods use separate statistical tests for each pair, however, information sharing between groups of ‘similar’ medications or outcomes may produce more sensitive and specific detection of harmful effects. Data-defined groups identified through clustering may provide a better reflection of dependencies within the data. We present a new signal detection method that combines the information component, biclustering and Bayesian hierarchical modelling (IC–Bic–BHM) to identify groups within the data and use these groups within the analysis. We use simulations to demonstrate the importance of using groupings that accurately reflect dependency within the data; and compare sensitivity, specificity, positive predictive value and negative predictive value of the IC–Bic–BHM with other methods. Results of application to 536 medications and 61 types of congenital anomaly in babies/foetuses from 87,513 EUROmediCAT pregnancies were also compared. Sensitivity and specificity of Bayesian hierarchical models are dependent upon how accurately groupings reflect dependency within the data. The IC-Bic-BHM is the first method to use data-defined dependencies between medications and outcomes in signal detection. Simulations suggest it performs well in datasets with large numbers of medication–outcome pairs but requires adequate power to detect groups. The IC-Bic-BHM identified signals involving several known teratogenic medications within EUROmediCAT data. This study presents a novel approach to signal detection but requires further investigation before adoption.
Retrieval of adverse event reports based on coded drug–event co-occurrence enables large-scale pharmacovigilance analyses but yields candidate reports rather than clinically adjudicated cases. Without explicit case identification, heterogeneous reports may be combined within the same case series and misinterpreted in downstream analyses. The aim of the study was to develop a structured framework for distinguishing clinically meaningful cases from candidate reports retrieved through coded drug–event co-occurrence, and to illustrate its application across information-rich and routine-review conditions. We conducted a sequential framework-development study using two case studies. Drug-induced impulsivity was used to iteratively develop and refine the framework under information-rich conditions. Drug-induced suicidality was used to examine its application under routine-review conditions with variable information availability and explicit classification of nonassessability. In case 1, nonrelevant reports were more frequent for drugs with uncertain or no established evidence ( 20–40
In the absence of a gold standard for assessing the completeness of spontaneous adverse drug reaction (ADR) reports, most studies have relied on assessing the proportion of reports containing information on predefined categories. The same principle applies to the Vigi4Eudra-score, which can be used to analyse the completeness of ADR reports from EudraVigilance. Moreover, ADR reports with a high level of information provided may be more useful in pharmacovigilance practice. The first aim of our study was to further develop the already existing Vigi4Eudra-score by modifying its penalties and by adding additional relevant categories (= EudraScore). The second aim was to assess whether the EudraScore performed better than the Vigi4Eudra-score to identify reports with at least a possible causal relationship as “well documented”. An expert panel was established comprising assessors from the pharmacovigilance and the research department of the German Federal Institute for Drugs and Medical Devices, with many years of experience in the assessment of ADR reports. Based on expert consensus, the Vigi4Eudra-score was modified by adapting existing categories and incorporating additional causality-relevant criteria, resulting in the EudraScore. A use case was conducted to compare the values calculated using the EudraScore with those obtained from the Vigi4Eudra-score. The performance of both scores in identifying reports with an “at least possible” causal relationship as “well documented” was evaluated by calculating sensitivity and specificity. Compared to the Vigi4Eudra-score, some penalties were adjusted for the EudraScore, with reduced penalties for missing country of origin and drug dose (from 10 to 5
The spontaneous reporting system is the cornerstone of pharmacovigilance but is limited by substantial underreporting, which may delay safety signal detection. The aim of this study was to evaluate the electronic health record-adverse event search method for identifying and reporting additional potential adverse drug reactions in electronic health records to strengthen routine signal detection. Between August 2022 and August 2025, targeted searches for adverse drug reactions were conducted in approximately 2,750,000 electronic health records from two Dutch hospitals using a data collection text-mining tool. Search criteria were derived from information in spontaneous reports of specific drug–adverse event associations. Structured (e.g. laboratory results, medication prescriptions) and unstructured (e.g. clinical notes, hospital discharge letters) data were used to identify potential adverse drug reactions. Identified cases were manually verified and, when relevant, reported to the Netherlands Pharmacovigilance Centre Lareb and analysed alongside spontaneous reports for signal detection. A total of 147 electronic health record-derived cases were reported for 41 drug–adverse event associations. These contributed to 34 signal detection analyses, supporting the assessment of seven safety signals. In 75
