AIM:Machine learning (ML) applications in pharmacovigilance remain limited and underexplored. Using data from the French National pharmacovigilance database (FNPV), this proof-of-concept study aimed to assess the feasibility of using a ML algorithm-eXtreme Gradient Boosting (XGBoost)-combined with SHapley Additive exPlanations (SHAP) analysis, to classify a well-defined drug-adverse reaction pair (warfarin-associated gastrointestinal bleeding) and identify risk factors for reporting this adverse reaction. METHODS:We extracted individual case safety reports (ICSRs) involving warfarin from the FNPV. Following data preprocessing, the dataset was randomly split into a training set (75%) and an independent test set (25%). The XGBoost algorithm was trained to classify reports as either gastrointestinal bleeding or non-gastrointestinal bleeding adverse reactions. Model performance was assessed on the test set using multiple metrics: accuracy, recall, precision, F1-score and the area under the receiver operating characteristic curve (AUC-ROC). Feature importance and directional effects were interpreted using SHAP values. RESULTS:A total of 2025 ICSRs involving warfarin were extracted from the FNPV. After preprocessing, 1045 ICSRs were retained. On the test set, the XGBoost model demonstrated an accuracy of 0.68, recall of 0.80, precision of 0.68, F1-score of 0.74 and an AUC-ROC of 0.716. SHAP analysis revealed the four most influential features in predicting gastrointestinal bleeding reporting: age, gastrointestinal bleeding risk, antiplatelet and history of renal dysfunction. CONCLUSION:We demonstrated the feasibility of using an XGBoost algorithm combined with SHAP analysis to pharmacovigilance data for the classification of adverse drug reactions and identify predictive features of reporting, following post hoc model interpretation by pharmacologists.

Drug hypersensitivity reactions (DHRs) are a significant clinical concern, leading to severe complications, prolonged hospital stays and increased healthcare costs. Despite their importance, DHRs are under-reported, limiting the ability to estimate their true prevalence and impeding effective pharmacovigilance. Current methods for detecting DHRs, such as spontaneous reporting systems and manual reviews, are either insufficient or labour intensive, emphasizing the need for automated approaches. This study evaluated Llama 3.1 (70b), an open-source large language model (LLM), for detecting allergic adverse drug events in hospital discharge summaries. In total 305 randomly selected French discharge summaries were analysed, with parameters set to achieve 90–95% sensitivity for DHR detection. Llama 3.1 (70b) identified the presence of DHRs, the causative drug, the allergic mechanism (immediate or delayed) and pharmacovigilance reporting. Four board-certified dermatologists, blinded to each other’s evaluations and the model’s output, reviewed the summaries to create a consensus-based gold standard. Out of 305 summaries, 27 DHRs were identified, primarily delayed reactions (70%). Notably, none of these were reported to pharmacovigilance agencies. While dermatologists required 2.5–4 h for review, Llama 3.1 completed the task in 17 min. The model achieved 85% sensitivity and 84% specificity, closely mirroring expert performance. Interestingly, all DHRs missed by Llama were also overlooked by at least one expert. Llama 3.1 demonstrated strong accuracies in identifying the responsible drug (88%), the allergic mechanism (96%) and pharmacovigilance reporting (88%). Its performance underscores the potential of LLMs as efficient, cost-effective tools for large-scale DHR detection. The study also highlighted the absence of prompt engineering, reflecting real-world application scenarios, and emphasized the model’s open-source nature, which ensures privacy and scalability. These findings suggest that integrating LLMs like Llama 3.1 into clinical workflows could enhance drug safety surveillance and streamline pharmacovigilance processes. By automating DHR detection, healthcare systems could improve reporting, enhance patient safety, and optimize prescribing practices. Future studies should explore scalability to larger datasets and integration into clinical settings. In conclusion, this study demonstrates the feasibility of using privacy-preserving LLMs for DHR detection, paving the way for improved pharmacovigilance and healthcare practices.

The drug authorization process is shifting towards a policy aimed at shortening time-to-market. While this policy facilitates early access to new treatments, it can also result in potentially insufficient knowledge of both efficacy and safety at the time of marketing. The latter is particularly true for long-term outcomes or in specific populations (e.g. children and the elderly). Yet, French pharmacoepidemiology is currently not designed to address these challenges, despite recognized expertise. In this context, we aim: (i) to define a strategy for strengthening pharmacoepidemiology in France; and (ii) to identify the associated human, technical, and financial requirements to ensure its success. In this paper, we present the French Pharmacoepidemiology Initiative (https://frenchpharmacoepi.org/), i.e. a network of independent academic teams to complement existing institutions. It will provide coordinated expertise and a workforce to meet national and regional needs for pharmacoepidemiological monitoring and drug-related decision-making. Leveraging the existing expertise of university hospital pharmacoepidemiology units would enable rapid operational deployment to inform the decisions and policies of national regulatory agencies.

OBJECTIVE:The risk of mucocutaneous ulcerations associated with nicorandil remains a well-described adverse effect (AE). In case of a suspected AE, early diagnosis and immediate discontinuation of nicorandil are recommended. The aim of this study is to update pharmacovigilance data. METHODS:Two sources of data were used for this study over the period from January 2017 to the end of November 2024: pharmacovigilance reports registered in the French PharmacoVigilance Database (FPVD) and data related to this AE and nicorandil extracted from the Toulouse hospital discharge database [programme de médicalisation des systèmes d'information (PMSI)]. RESULTS:We collected a total of 62 cases: 28 cases were registered in the FPVD and 34 additional cases could be identified in PMSI (n=62). None of these cases were reported to the Toulouse Pharmacovigilance Center. Patients were aged 56 to 97 years (sex-ratio 0.94). Nicorandil was discontinued for 36 patients (11 immediately). Six patients died, three of them despite nicorandil discontinuation, mainly due to digestive complications or sepsis. In 61% of FPVD cases (n=17) the AE was classified as severe. The median time to onset of the ulceration was 407 days (IQR: 123 to 1826 days). Cutaneous ulcerations were mainly localized on the lower limbs and mucosal ulcerations mainly affected the oral and digestive mucosa. CONCLUSION:Despite a decline in nicorandil sales since 2017 and several communications from the health authorities, our findings indicate the persistence of serious adverse reactions with nicorandil. Delayed discontinuation of the drug results in unnecessary investigations and potentially fatal outcomes. This study has shown that the PMSI and then the Clinical Data Warehouse (CDW), operational at the Toulouse Universitary Hospital Center since 18 June 2025, is a data source that contributes to reducing the under-reporting rate of unknown, albeit "expected" AE and to confirming the persistence of a validated pharmacovigilance signal.

Objectif Le risque d’ulcérations cutanéo-muqueuses associées au nicorandil reste un effet indésirable (EI) bien décrit. En cas de suspicion de cet EI, un diagnostic précoce et un arrêt immédiat du nicorandil sont recommandés. L’objectif de ce travail est d’actualiser les données de pharmacovigilance. Méthodes Deux sources de données ont été utilisées sur la période de janvier 2017 à fin novembre 2024 : les déclarations de pharmacovigilance enregistrées dans la base nationale de pharmacovigilance (BNPV) et les dossiers en lien avec cet EI et le nicorandil à partir des données du PMSI (programme de médicalisation des systèmes d’information) du centre hospitalier universitaire de Toulouse. Résultats Nous avons collecté un total de 62 cas : 28 cas enregistrés dans la BNPV et 34 cas supplémentaires identifiés dans le PMSI (n=62). Aucun des cas du PMSI n’a été signalé au centre de pharmacovigilance de Toulouse. Les patients étaient âgés de 56 à 97ans (sex-ratio 0,94). Le nicorandil a été arrêté chez 36 patients (immédiatement pour 11 d’entre eux). Six patients sont décédés, dont trois malgré l’arrêt du nicorandil, principalement en raison de complications digestives ou de sepsis. Dans 61 % des cas de la BNPV (n=17) l’EI était classé comme grave et le délai médian de survenue de l’EI était de 407jours (IQR : 123 à 1826jours). Les ulcérations cutanées étaient principalement localisées sur les membres inférieurs et les ulcérations muqueuses affectaient majoritairement les muqueuses buccale et digestive. Conclusion Malgré une baisse des ventes de nicorandil depuis 2017 et les différents communiqués des autorités de santé, nos résultats témoignent de la persistance d’effets indésirables graves sous nicorandil. L’arrêt non immédiat du médicament conduit à des investigations inutiles et à des issues potentiellement fatales. Cette étude a montré que les données du PMSI et donc l’entrepôt des données de santé (EDS) opérationnel au CHU de Toulouse depuis 18 juin 2025, s’avère une source de données permettant d’améliorer la sous-notification des effets indésirables méconnus bien qu’« attendu » et de confirmer la persistance d’un signal de pharmacovigilance validé.

Background: Fluoroquinolones (FQ) are the second most frequent antibiotics responsible for hypersensitivity reaction (HSR). Data concerning the interest of allergological work-ups are scarce. Objectives: We reported a university center experience of the interest and safety of skin tests (ST) and drug provocation tests (DPT) for the exploration of immediate hypersensitivity reaction (IHR) and delayed HSR (DHR) to FQ. Materials and Methods: This retrospective study included patients tested for at least 1 FQ between 2008 and 2023. The primary outcome was the frequency of positive tests for FQ. Secondary outcomes aimed to characterize results, cross-reactivity, and tolerance of the tests. Results: Among the 113 patients included (32% in IHR, 68% in DHR), the overall test positivity rate was 27% (29% in DHR, 23% in IHR). Negative predictive value was 85% (80% in DHR, 95% in IHR). The most frequently positive ST was the intradermal test (7%). Allergy was more frequently confirmed for ofloxacin (OFX) (38%). Positive tests were more frequent for fixed drug eruption (67%). Cross-reactivity was detected in 15%, involving OFX and LFX or CPX and OFX. DPT was responsible for nonsevere recurrence of initial HSR for 13% of patients who had DPT and recurrence leading to a short hospitalization for 2% of them. 3% of contacted patients experienced a nonsevere recurrence of HSR in real life. Conclusion: This cohort demonstrates the benefit and safety of allergological exploration for FQ allowing subsequent use in 73% of cases.

OBJECTIVE:Metformin-associated lactic acidosis (MALA) is a rare but serious adverse drug reaction (ADR). The aim of the study was to identify clinical situations associated with the onset of MALA in patients hospitalised in the Nord Pas de Calais regional intensive care units (ICUs), and to assess its preventability. MATERIAL AND METHODS:We included all cases of MALA, identified by metformin accumulation >2.3mg/dL and lactate >2.2mmol/L, reported by the regional ICU physicians to the Regional Centre of Pharmacovigilance and registered in the French Pharmacovigilance Database between 1 January 2017 and 30 December 2018. RESULTS:One hundred and ninety-eight (198) cases of MALA were included. 38 patients died in direct association with MALA (19.2%). There was a correlation between metformin plasma accumulation and acute renal failure and with the severity of MALA (P<0.0001). All patients presented an acute intercurrent event favouring MALA, dehydration for 87 (43.9%) patients, severe infection for 65 (32.8%) patients. For 172 patients (86.7%), the prescription was not adapted to the intercurrent medical situation as recommended. Seventy (40.5%) patients consulted their general practitioner for the acute intercurrent event, 1 temporarily stopped metformin and 34.3% had been referred directly to hospital. The remaining 65.7% presented to the hospital around 4 days later due to worsening symptoms. MALA was identified as preventable in 160 patients (80.8%). CONCLUSIONS:MALA in ICUs often follow acute dehydration or infection, and these high-risk situations must be signals to prevent this serious ADR. Specific education programmes for physicians and patients could also reduce this risk.

Introduction: Ischemic or hemorrhagic stroke can occur to patients treated with oral anticoagulants (OAC), through lack of effectiveness or overdosing. Objective: To evaluate the impact of clinical pharmacist's intervention on pharmacovigilance (PV) reporting for OAC-treated patients hospitalized for stroke. Methods: Monocentric prospective study in which a clinical pharmacist's intervention was performed in a stroke unit, with a focus on patients treated by OAC prior admission. A PV report was made with all data collected for cases of stroke suspected to be related to OAC therapy. Data provided by pharmacist were compared with data initially available in the patient's electronic medical records. PV reports with pharmacist intervention were compared to those without. Results: During the study period, 48 patients were included in the study: 43 (89.6%) ischemic strokes with an embolic or unknown etiology, four hemorrhage strokes (8.33%), and one medication error (2.08%). A clinical pharmacist intervention was performed for 19 patients (39.6%) and provided significant additional data in all of them (100%). The information was related to adherence to treatment for 17 cases (89.5%), OAC's initial prescription date for 11 cases (57.9%) and identifying event(s) that could have interfered with the efficacy of the OAC in five cases (26.3%). For patients with pharmacist intervention, PV reports were significantly more informative in terms of date's introduction of anticoagulant, adherence to treatment, reference to weight change or concomitant event. Conclusions: clinical pharmacist's intervention with patients taking oral anticoagulants and hospitalized for acute stroke contributes to collect high-quality data for pharmacovigilance reporting.

Dermatitis®Vol. 34, No. 2 LettersDuodenal Stenosis Linked to Drug Reaction With Eosinophilia and Systemic SymptomsMarion Douxami, Emmanuel Faure, Charlotte Fievet, Sébastien Buche, Selma Azib-Meftah, Aurélie Cuypers-Tilmant, Johana Béné, Sophie Gautier, Delphine Staumont-Salle, and Frédéric DezoteuxMarion DouxamiService de Dermatologie, CHU Lille, Lille, France. E-mail Address: [email protected]Search for more papers by this author, Emmanuel FaureService de Maladies Infectieuses, CHU Lille, Lille, FranceSearch for more papers by this author, Charlotte FievetService de Dermatologie, CHU Lille, Lille, FranceService de Dermatologie, CH Seclin, Seclin, FranceSearch for more papers by this author, Sébastien BucheService de Dermatologie, CHU Lille, Lille, FranceService de Dermatologie, CH Seclin, Seclin, FranceSearch for more papers by this author, Selma Azib-MeftahService de Dermatologie, CH Seclin, Seclin, FranceSearch for more papers by this author, Aurélie Cuypers-TilmantCentre de Biologie pathologie, Institut de Pathologie, CHU Lille, Lille, FranceSearch for more papers by this author, Johana BénéService de Dermatologie, CHU Lille, Lille, FranceSearch for more papers by this author, Sophie GautierCentre Régional de Pharmacovigilance, CHU Lille, Lille, FranceSearch for more papers by this author, Delphine Staumont-SalleService de Dermatologie, CHU Lille, Lille, FranceU1286 Inserm, INFINITE-Institute for Translational Research in Inflammation, Lille, FranceSearch for more papers by this author, and Frédéric DezoteuxService de Dermatologie, CHU Lille, Lille, FranceU1286 Inserm, INFINITE-Institute for Translational Research in Inflammation, Lille, FranceSearch for more papers by this authorPublished Online:16 Mar 2023https://doi.org/10.1089/derm.2022.29001.mdoAboutSectionsView articleView Full TextPDF/EPUB Permissions & CitationsPermissionsDownload CitationsTrack CitationsAdd to favorites Back To Publication ShareShare onFacebookTwitterLinked InRedditEmail View articleFiguresReferencesRelatedDetails Volume 34Issue 2Apr 2023 Information© 2023 American Contact Dermatitis Society. All Rights Reserved.To cite this article:Marion Douxami, Emmanuel Faure, Charlotte Fievet, Sébastien Buche, Selma Azib-Meftah, Aurélie Cuypers-Tilmant, Johana Béné, Sophie Gautier, Delphine Staumont-Salle, and Frédéric Dezoteux.Duodenal Stenosis Linked to Drug Reaction With Eosinophilia and Systemic Symptoms.Dermatitis®.Apr 2023.169-170.http://doi.org/10.1089/derm.2022.29001.mdoPublished in Volume: 34 Issue 2: March 16, 2023Online Ahead of Print:January 19, 2023PDF download

Ethanol is an excipient with known effect whose presence is regulated because it can cause adverse effects, notably a misuse. In order to raise awareness of this risk, this study searched all oral drugs with ethanol as an excipient from the Theriaque (R) database. All drugs marketed in France with a unit dose ethanol intake of 0.1 g or more were identified and analyzed, according to the maximum unit and daily dosage recommended by the manufacturer. This research revealed 106 pharmaceutical specialties responsible for a unit intake of ethanol of 0.1 g or more among the 8532 oral drugs containing ethanol (1.2 %): 2 at a daily dose > 13 g and the majority (57/106; 54 %) at a daily dose < 1 g. These are mainly oral solutions (97/106; 91 %) of phytotherapy (45/97; 46 %). The most frequently found therapeutic class was anti-tussive (12/106; 11 %). The majority of drugs are over-the-counter medication (56/106; 53 %). Overall, 106 drugs on the French market can be associated with a risk of misuse and cause adverse effects in vulnerable populations such as children and pregnant women. Vigilance and appropriate monitoring is required for these drugs (especially those over-the-counter ones), and their substitution should be preferred if possible. (c) 2022 Societe francaise de pharmacologie et de therapeutique. Published by Elsevier Masson SAS. All rights reserved.

Immune checkpoint inhibitor (ICI)-related cytopenias have been poorly described. This study aimed to further characterize ICI-related cytopenias, using the French pharmacovigilance database. All grade ≥ 2 hematological adverse drug reactions involving at least one ICI coded as suspected or interacting drug according to the World Health Organization criteria and reported up to 31 March 2022, were extracted from the French pharmacovigilance database. Patients were included if they experienced ICI-related grade ≥ 2 cytopenia. We included 68 patients (75 ICI-related cytopenias). Sixty-three percent were male, and the median age was 63.0 years. Seven patients (10.3%) had a previous history of autoimmune disease. Immune thrombocytopenia (ITP) and autoimmune hemolytic anemia (AIHA) were the most frequently reported (50.7% and 25.3%, respectively). The median time to onset of ICI-related cytopenias was 2 months. Nearly half were grade ≥ 4, and three patients died from bleeding complications of refractory ITP and from thromboembolic disease with active AIHA. Out of 61 evaluable responses, complete or partial remission was observed after conventional treatment in 72.1% of ICI-related cytopenias. Among the 10 patients with ICI resumption after grade ≥ 2 ICI-related cytopenia, three relapsed. ICI-related cytopenias are rare but potentially life-threatening. Further studies are needed to identify risk factors of ICI-related cytopenias.

L’éthanol est un excipient à effet notoire dont la présence est réglementée car il peut causer des effets indésirables, et notamment l’apparition de mésusage. Afin de sensibiliser à ce risque, cette étude a recherché les médicaments destinés à la voie orale contenant de l’éthanol en tant qu’excipient depuis la base de données Thériaque®. Les médicaments commercialisés engendrant une prise unitaire d’éthanol >0,1g ont été identifiés et analysés, en fonction de la posologie maximale unitaire et journalière recommandée par le fabricant. Cette recherche met en évidence 106/8532 spécialités orales contenant de l’éthanol (1,2 %) : 2 à une dose journalière >13g et la majorité (57/106 ; 54 %) à une dose journalière <1g. Ce sont principalement des solutions buvables (97/106 ; 91 %) de phytothérapie (45/97 ; 46 %). La classe thérapeutique la plus fréquemment retrouvée et celle des antitussifs (12/106 ; 11 %). La majorité des médicaments sont en vente libre en pharmacie d’officine (56/106 ; 53 %). Au total, 106 spécialités peuvent être associées à un risque de mésusage et engendrer une exposition néfaste chez des populations particulières comme l’enfant et la femme enceinte. Une vigilance et une surveillance adaptée sont nécessaires pour ces traitements (surtout ceux non listés), et leur substitution doit être privilégiée.

BACKGROUND The antidepressant escitalopram is widely prescribed for the treatment of depression. It is generally well-tolerated, and cholestasis is not mentioned in its summary of product characteristics (SmPC). We present a case of cholestatic and cytolysis liver injury due to escitalopram and a VigiBase® study. CASE SUMMARY A 68-year-old man was admitted to our emergency unit due to clinical jaundice associated with hepatitis, pruritus and dark urine. We tested the patient for the most common etiologies of jaundice, including hemolysis, viral hepatitis, cirrhosis, carcinoma, cholangitis, cholelithiasis and intrahepatic or extrahepatic obstruction. The etiological study was negative, and an adverse drug reaction was the sole possible explanation. The patient was receiving treatment with escitalopram. Two days after its withdrawal, pruritus was resolved. Ten days after withdrawal, clinical jaundice disappeared. It took a month and three weeks after withdrawal for the patient to have normalized liver function tests. To our knowledge, this is the first reported case of cholestasis where treatment with escitalopram was the only possible cause, with a highly probable causality. In addition, we determined whether escitalopram is associated with hepatotoxicity and cholestasis by performing a disproportionality analysis. All cases of hepatobiliary disorders induced by escitalopram and reported in the World Health Organization pharmacovigilance database (VigiBase®) were analyzed to characterize this toxicity. We found that patients treated with escitalopram had an increased risk of hepatitis [odds ratio (OR) = 1.938(1.186-3.166)] and cholestasis [OR = 1.866(1.279-2.724)] [OR (95% confidence interval)]. The median duration between the introduction of escitalopram and the occurrence of acute hepatitis and/or cholestasis was ten days +/- seven days. CONCLUSION Although extremely rare, this case report, the review of the literature and the pharmacovigilance update confirm that escitalopram can cause drug-induced hepatotoxicity and cholestasis, generally within a week after initiation. Thus, escitalopram should be withdrawn immediately if an iatrogenic cause cannot be excluded. If its responsibility is ascertained, escitalopram should be consequently contraindicated. In addition, serotoninergic antidepressants in patients with non-severe depression are ineffective and harmful. Finally, the SmPC of escitalopram should be updated to alert for this risk and give clear clinical guidelines.

Enteroviruses are a frequent source of infection and among the most common central nervous system viral pathogens. Enteroviruses – in particular, the Coxsackie B viruses – are a known cause of myocarditis. Rituximab is a genetically engineered chimeric anti-CD20 monoclonal antibody. Many reports in the literature suggest a higher risk of infection following repeated rituximab therapy, including viral infection. However, observations of enterovirus-related myocarditis in the context of rituximab treatment are scarce. The authors describe the case of a patient with neuromyelitis optica spectrum disorder who developed severe and fatal enterovirus-related myocarditis after rituximab therapy with a difficult differential diagnosis of autoimmune or giant-cell myocarditis. This case highlights the importance of complete diagnostic workup in difficult cases of myocarditis, including endomyocardial biopsies.