Résumé
Aim: Clinically, interstitial lung disease (ILD) is a heterogeneous group of respiratory disorders. Due to their low incidence, pharmacovigilance database analysis is useful to detect them. Precise diagnosis is challenging as well as coding in these databases. Query criteria are among the key elements for a good signal detection. This study aimed to investigate interpretation biases related to ILD coding in PV databases.Methods: MedDRA Preferred Terms (PT) included in the broad Standardized MedDRA Query (SMQ) 'ILD' for the top five known pneumotoxic drugs was described in VigiBase by reporting year (2000-2024), country, and reporter type. Then, a systematic literature review was conducted to identify PV studies using disproportionality analyses to detect drug-induced ILD reporting signals, assessing query strategies, analysis units, and bias consideration.Results: The most frequently implicated known pneumotoxic drugs were amiodarone, methotrexate, nivolumab, pembrolizumab, and everolimus. On average per year, the PTs 'ILD' represented <40% of all PTs in the broad SMQ 'ILD'. Terminology evolution (e.g., emergence of 'immune-mediated lung disease' after 2019) and national preferences (France/Japan vs. USA/UK/Germany) were observed. Among 22 reviewed accessible studies, 50% used PT-based queries, only 27% performed sensitivity analyses, and none discussed the impact of MedDRA coding evolution on analysis reliability.Discussion/conclusion: Our analysis highlights the substantial variability in MedDRA coding that may affect the reliability of drug safety analyses. Moreover, studies conducted on PV databases do not sufficiently assess the relevance of query criteria and their impact on the results. We propose a few practical recommendations.