Screening of 16 major drug glucuronides for time-dependent inhibition of nine drug-metabolizing CYP enzymes - detailed studies on CYP3A inhibitors.

Time-dependent inhibition of cytochrome P450 (CYP) enzymes has been observed for a few glucuronide metabolites of clinically used drugs. Here, we investigated the inhibitory potential of 16 glucuronide metabolites towards nine major CYP enzymes in vitro. Automated substrate cocktail methods were used to screen time-dependent inhibition of CYP1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2J2 and 3A in human liver microsomes. Seven glucuronides (carvedilol ß-D-glucuronide, diclofenac acyl-ß-D-glucuronide, 4-hydroxyduloxetine ß-D-glucuronide, ezetimibe phenoxy-ß-D-glucuronide, raloxifene 4'-glucuronide, repaglinide acyl-ß-D-glucuronide and valproic acid ß-D-glucuronide) caused NADPH- and time-dependent inhibition of at least one of the CYPs investigated, including CYP2A6, CYP2C19 and CYP3A. In more detailed experiments, we focused on the glucuronides of carvedilol and diclofenac, which inhibited CYP3A. Carvedilol ß-D-glucuronide showed weak time-dependent inhibition of CYP3A, but the parent drug carvedilol was found to be a more potent inhibitor of CYP3A, with the half-maximal inhibitor concentration (IC50) decreasing from 7.0 to 1.1 µM after a 30-min preincubation with NADPH. The maximal inactivation constant (kinact) and the inhibitor concentration causing half of kinact (KI) for CYP3A inactivation by carvedilol were 0.051 1/min and 1.8 µM, respectively. Diclofenac acyl-ß-D-glucuronide caused time-dependent inactivation of CYP3A at high concentrations, with a 4-fold IC50 shift (from 400 to 98 µM after a 30-min preincubation with NADPH) and KI and kinact values of >2,000 µM and >0.16 1/min. In static predictions, carvedilol caused significant (>1.25-fold) increase in the exposure of the CYP3A substrates midazolam and simvastatin. In conclusion, we identified several glucuronide metabolites with CYP inhibitory properties. Based on detailed experiments, the inactivation of CYP3A by carvedilol may cause clinically significant drug-drug interactions.

Hydroxychloroquine is Metabolized by Cytochrome P450 2D6, 3A4, and 2C8, and Inhibits Cytochrome P450 2D6, while its Metabolites also Inhibit Cytochrome P450 3A in vitro.

This study aimed to explore the cytochrome P450 (CYP) metabolic and inhibitory profile of hydroxychloroquine (HCQ). Hydroxychloroquine metabolism was studied using human liver microsomes (HLMs) and recombinant CYP enzymes. The inhibitory effects of HCQ and its metabolites on nine CYPs were also determined in HLMs, using an automated substrate cocktail method. Our metabolism data indicated that CYP3A4, CYP2D6, and CYP2C8 are the key enzymes involved in HCQ metabolism. All three CYPs formed the primary metabolites desethylchloroquine (DCQ) and desethylhydroxychloroquine (DHCQ) to various degrees. Although the intrinsic clearance (CLint) value of HCQ depletion by recombinant CYP2D6 was > 10-fold higher than that by CYP3A4 (0.87 versus 0.075 µl/min/pmol), scaling of recombinant CYP CLint to HLM level resulted in almost equal HLM CLint values for CYP2D6 and CYP3A4 (11 and 14 µl/min/mg, respectively). The scaled HLM CLint of CYP2C8 was 5.7 µl/min/mg. Data from HLM experiments with CYP-selective inhibitors also suggested relatively equal roles for CYP2D6 and CYP3A4 in HCQ metabolism, with a smaller contribution by CYP2C8. In CYP inhibition experiments, HCQ, DCQ, DHCQ, and the secondary metabolite didesethylchloroquine were direct CYP2D6 inhibitors, with 50% inhibitory concentration (IC50) values between 18 and 135 µM. HCQ did not inhibit other CYPs. Furthermore, all metabolites were time-dependent CYP3A inhibitors (IC50 shift 2.2-3.4). To conclude, HCQ is metabolized by CYP3A4, CYP2D6, and CYP2C8 in vitro. HCQ and its metabolites are reversible CYP2D6 inhibitors, and HCQ metabolites are time-dependent CYP3A inhibitors. These data can be used to improve physiologically-based pharmacokinetic models and update drug-drug interaction risk estimations for HCQ. SIGNIFICANCE STATEMENT: While CYP2D6, CYP3A4, and CYP2C8 have been shown to mediate chloroquine biotransformation, it appears that the role of CYP enzymes in hydroxychloroquine (HCQ) metabolism has not been studied. In addition, little is known about the CYP inhibitory effects of HCQ. Here, we demonstrate that CYP2D6, CYP3A4, and CYP2C8 are the key enzymes involved in HCQ metabolism. Furthermore, our findings show that HCQ and its metabolites are inhibitors of CYP2D6, which likely explains the previously observed interaction between HCQ and metoprolol.

An automated cocktail method for in vitro assessment of direct and time-dependent inhibition of nine major cytochrome P450 enzymes - application to establishing CYP2C8 inhibitor selectivity.

We developed an in vitro high-throughput cocktail assay with nine major drug-metabolizing CYP enzymes, optimized for screening of time-dependent inhibition. The method was applied to determine the selectivity of the time-dependent CYP2C8 inhibitors gemfibrozil 1-O-ß-glucuronide and clopidogrel acyl-ß-D-glucuronide. In vitro incubations with CYP selective probe substrates and pooled human liver microsomes were conducted in 96-well plates with automated liquid handler techniques and metabolite concentrations were measured with quantitative UHPLC-MS/MS analysis. After determination of inter-substrate interactions and Km values for each reaction, probe substrates were divided into cocktails I (tacrine/CYP1A2, bupropion/CYP2B6, amodiaquine/CYP2C8, tolbutamide/CYP2C9 and midazolam/CYP3A4/5) and II (coumarin/CYP2A6, S-mephenytoin/CYP2C19, dextromethorphan/CYP2D6 and astemizole/CYP2J2). Time-dependent inhibitors (furafylline/CYP1A2, selegiline/CYP2A6, clopidogrel/CYP2B6, gemfibrozil 1-O-ß-glucuronide/CYP2C8, tienilic acid/CYP2C9, ticlopidine/CYP2C19, paroxetine/CYP2D6 and ritonavir/CYP3A) and direct inhibitor (terfenadine/CYP2J2) showed similar inhibition with single substrate and cocktail methods. Established time-dependent inhibitors caused IC50 fold shifts ranging from 2.2 to 30 with the cocktail method. Under time-dependent inhibition conditions, gemfibrozil 1-O-ß-glucuronide was a strong (>90% inhibition) and selective (<< 20% inhibition of other CYPs) inhibitor of CYP2C8 at concentrations ranging from 60 to 300 µM, while the selectivity of clopidogrel acyl-ß-D-glucuronide was limited at concentrations above its IC80 for CYP2C8. The time-dependent IC50 values of these glucuronides for CYP2C8 were 8.1 and 38 µM, respectively. In conclusion, a reliable cocktail method including the nine most important drug-metabolizing CYP enzymes was developed, optimized and validated for detecting time-dependent inhibition. Moreover, gemfibrozil 1-O-ß-glucuronide was established as a selective inhibitor of CYP2C8 for use as a diagnostic inhibitor in in vitro studies.

VOTRAGE study: Phase I dose-escalation study of pazopanib in unfit older patients.

BACKGROUND: Pazopanib is a tyrosine kinase inhibitor given at the approved dose of 800 mg orally once daily (OD), but often requiring individual dose adjustment due to toxicity. Limited data is available to guide prescription in older patients especially the unfit according to geriatric assessment. PATIENTS AND METHODS: VOTRAGE is a 3 + 3 dose-escalation, open-label phase I trial of continuous OD oral administration of pazopanib to evaluate safety, PK and PD data in unfit older patients with advanced solid tumors. The primary objective was to determine the maximum tolerated dose (MTD). PK data were compared with those obtained in younger adult patients in a population PK analysis. RESULTS: Eighteen patients with a median age of 82.5 years (range 75-91) were included in three dosing cohorts (400, 600, and 800 mg daily). Three dose-limiting toxicities (DLT) were observed in five patients at 800 mg and one DLT at 600 mg in six evaluable patients. MTD was defined as level 2 dose (600 mg). Individual oral clearance was not correlated with age. A relationship was observed between the occurrence of DLT and pazopanib plasma exposure. Decreased oral bioavailability of pazopanib when given with proton-pump inhibitors was confirmed in this group of patients. CONCLUSION: We recommend performing geriatric assessment in patients older than 75 and starting pazopanib at 600 mg per day in unfit older patients. Therapeutic drug monitoring appears very helpful in this population.

Formulae recently proposed to estimate renal glomerular filtration rate improve the prediction of carboplatin clearance.

PURPOSE: While doses of carboplatin are mostly individualized according to the Calvert equation based on estimated Glomerular Filtration Rate (eGFR), there is still uncertainty regarding the best formula to predict GFR. Since Janowitz et al. recently proposed a new equation predicting GFR in cancer patients, we aimed to compare this equation to other carboplatin clearance (carboCL) predicting formulae. METHODS: The actual carboCL of 491 patients was compared to predicted carboCL according to the Calvert formula using several equations to predict GFR (Janowitz, Cockcroft-Gault, MDRD, CKD-EPI, CKD-EPI with cystatin C (CKD-EPI-cysC)); and according to two others that directly predict carboCL (Chatelut and Thomas). The formulae were compared on Mean Percentage Error (MPE), Mean Absolute Percentage Error (MAPE) and percentage of patients with a prediction error above 20% (P20). RESULTS: The MPE, MAPE and P20 were, respectively, within the ranges - 5.2 to + 5.9%; 14.0-21.2% and 23-46%. The MAPE and P20 of Calvert-CKD-EPI-cysC were the lowest. The performance of Calvert-CKD-EPI was better than that of other creatinine-based formulae although not significantly different from the Calvert-Janowitz formula. Among formulae based on creatinine only, Calvert-CKD-EPI and Calvert-Janowitz are the least influenced by patient characteristics. CONCLUSION: Whereas CysC improves carboplatin CL prediction, the Calvert-CKD-EPI equation seems the most suitable creatinine-based formula to predict carboCL homogeneously in all subgroups of patients.

Involvement of Pazopanib and Sunitinib Aldehyde Reactive Metabolites in Toxicity and Drug-Drug Interactions in Vitro and in Patient Samples.

Tyrosine kinase inhibitors (TKI) are targeted anticancer drugs that have been successfully developed over the past 2 decades. To date, many of them (around 70%) require warnings for liver injury and five of them, including pazopanib and sunitinib, have Black Box Warning (BBW) labels. Although TKI-induced hepatotoxicity is the first cause of drug failures in clinical trials, BBW labels, and market withdrawals, the underlying mechanisms remain unclear. However, the recent discovery of new reactive metabolites (RM) with aldehyde structures during pazopanib and sunitinib metabolism offers new perspectives for investigating their involvement in the toxicity of these two TKI. These hard electrophiles have a high reactivity potential toward proteins and are thought to be responsible for cytochrome P450 inactivation, drug-drug interactions (DDI), and liver toxicity. We report here, for the first time, the presence of these aldehyde RM in human plasma samples obtained during drug monitoring. Docking experiments in the CYP3A4 active site were performed and showed that pazopanib and sunitinib fitting in the catalytic site are in accordance with their regioselective oxidation to aldehydes. They also suggested that aldehyde RM may react with lysine and arginine residues. Based on these results, we studied the reactivity of the aldehyde RM toward lysine and arginine residues as potential targets on the protein framework to better understand how these RM could be involved in liver toxicity and drug-drug interactions. Adduct formation with different hepatic and plasma proteins was investigated by LC-MS/MS, and adducts between pazopanib or sunitinib aldehyde derivatives and lysine residues on both CYP3A4 and plasma proteins were indeed shown for the first time.

Identifying the reactive metabolites of tyrosine kinase inhibitors in a comprehensive approach: Implications for drug-drug interactions and hepatotoxicity.

Tyrosine kinase inhibitors (TKI) are small heterocyclic molecules targeting transmembrane and cytoplasmic tyrosine kinases that have met with considerable success in clinical oncology. TKI are associated with toxicities including liver injury that may be serious and even life-threatening. Many of them require warnings in drug labeling against liver injury, and five of them have Black Box Warning (BBW) labels. Although drug-induced liver injury is a matter of clinical and industrial concern, little is known about the underlying mechanisms that likely involve reactive metabolites (RM). RM are electrophiles or radicals originating from the metabolic activation of particular functional groups, known as structural alerts or toxicophores. RM are able to covalently bind to proteins and macromolecules, causing cellular damage and even cell death. If the adducted protein is the enzyme involved in RM formation, time-dependent inhibition of the enzyme-also called mechanism-based inhibition (MBI) or inactivation-can occur and lead to pharmacokinetic drug-drug interactions. To mitigate RM liabilities, common practice in drug development includes avoiding structural alerts and assessing RM formation via RM trapping screens with soft and hard nucleophiles (glutathione, potassium cyanide, and methoxylamine) in liver microsomes. RM-positive derivatives are further optimized to afford drug candidates with blocked or minimized bioactivation potential. However, different structural alerts are still commonly used scaffolds in drug design, including in TKI structures. This review focuses on the current state of knowledge of the relations among TKI structures, bioactivation pathways, RM characterization, and hepatotoxicity and cytochrome P450 MBI in vitro.

Pharmacodynamic Therapeutic Drug Monitoring for Cancer: Challenges, Advances, and Future Opportunities.

In the modern era of cancer treatment, with targeted agents superseding more traditional cytotoxic chemotherapeutics, it is becoming increasingly important to use stratified medicine approaches to ensure that patients receive the most appropriate drugs and treatment schedules. In this context, there is significant potential for the use of pharmacodynamic biomarkers to provide pharmacological information, which could be used in a therapeutic drug monitoring setting. This review focuses on discussing some of the challenges faced to date in translating preclinical pharmacodynamic biomarker approaches to a clinical setting. Recent advances in important areas including circulating biomarkers and pharmacokinetic/pharmacodynamic modeling approaches are discussed, and selected examples of anticancer drugs where there is existing evidence to potentially advance pharmacodynamic therapeutic drug monitoring approaches to deliver more effective treatment are discussed. Although we may not yet be in a position to systematically implement therapeutic drug monitoring approaches based on pharmacodynamic information in a cancer patient setting, such approaches are likely to become more commonplace in the coming years. Based on ever-increasing levels of pharmacodynamic information being generated on newer anticancer drugs, facilitated by increasingly advanced and accessible experimental approaches available to researchers to collect these data, we can now look forward optimistically to significant advances being made in this area.

Clinical Pharmacokinetics and Pharmacodynamics of Dabrafenib.

Dabrafenib is a potent and selective inhibitor of BRAF-mutant kinase that is approved, as monotherapy or in combination with trametinib (mitogen-activated protein kinase (MAPK) kinase (MEK) inhibitor), for unresectable or metastatic BRAF-mutated melanoma, advanced non-small cell lung cancer and anaplastic thyroid cancer harbouring the BRAFV600E mutation. The recommended dose of dabrafenib is 150 mg twice daily (bid) under fasted conditions. After single oral administration of the recommended dose, the absolute oral bioavailability (F) of dabrafenib is 95%. Dabrafenib shows a time-dependent increase in apparent clearance (CL/F) following multiple doses, which is likely due to induction of its own metabolism through cytochrome P450 (CYP) 3A4. Therefore, steady state is reached only after 14 days of daily dose administration. Moreover, the extent of this auto-induction process is dependent on the dose, which explains why dabrafenib systemic exposure at steady state increases less than dose proportionally over the dose range of 75-300 mg bid. The main elimination route of dabrafenib is the oxidative metabolism via CYP3A4/2C8 and biliary excretion. Among the three major metabolites identified, hydroxy-dabrafenib appears to contribute to the pharmacological activity. Age, sex and body weight did not have any clinically significant influence on plasma exposure to dabrafenib. No dose adjustment is needed for patients with mild renal or hepatic impairment, whereas the impacts of severe impairment on dabrafenib pharmacokinetics remain unknown. Considering that dabrafenib is a substrate of CYP3A4/2C8 and is a CYP3A4/2B6/2C inducer, drug-drug interactions are expected with dabrafenib. The relationship between clinical outcomes and plasma exposure to dabrafenib and hydroxy-dabrafenib should be investigated more deeply.

Safety of oral hydration after cisplatin infusion in an outpatient lung cancer unit.

PURPOSE: Hydration is needed before and after cisplatin infusion for reducing the risk of nephrotoxicity. Even though there is no standard regimen, patients receive mostly intravenous hydration before and after cisplatin leading hospitalization during at least one night. Since the feasibility has been published, oral hydration after cisplatin was implemented in our practice. The safety of this new way of hydration needs to be assessed in clinical practice. METHODS: We collected medical records from patients treated by cisplatin for lung cancer in our unit between 2010 and 2016. We retrospectively analyzed the incidence of cisplatin induced nephrotoxicity between after and before the change of hydration regimen. RESULTS: Our patient cohort included 241 patients hydrated by intravenous regimen (IV/IV group) and 276 patient hydrated by intravenous and oral regimen (IV/PO group). Grade ≥ 1 nephrotoxicity occurred in 39.4 and 25.7% in the IV/IV and IV/PO groups respectively (p = 0.001). Age over 70 at baseline was a predictive factor for nephrotoxicity, but not estimated glomerular filtration rate nor cisplatin-associated drugs. After a multivariate analysis, age remained a predictive factor for nephrotoxicity and IV/PO hydration associated with a decrease in nephrotoxic risk. CONCLUSION: The implementation of oral hydration in our practice was not associated with an increase in nephrotoxicity. Our observation based on large data from clinical practice shows that oral hydration after cisplatin is safe.

Metalloporphyrin-Catalyzed Oxidation of Sunitinib and Pazopanib, Two Anticancer Tyrosine Kinase Inhibitors: Evidence for New Potentially Toxic Metabolites.

Oxidation of two tyrosine kinase inhibitors (TKIs) sunitinib and pazopanib, using a chemical catalytic system able to mimic the cytochrome P450 type oxidation, allowed us to prepare  putative reactive/toxic metabolites of these anticancer drugs. Among these metabolites, aromatic aldehyde derivatives were unambiguously characterized. Such biomimetic oxidation of TKI-type drugs was essential to facilitate the identification of low amounts of aldehydes generated from these TKIs when incubated with human liver microsomes (HLM), which are classical models of human hepatic metabolism. These TKI derivative aldehydes quickly react in vitro with amines. A similar reaction is expected to occur in vivo and may be at the origin of the potentially severe hepatotoxicity of these TKIs.

[5-fluorouracil therapeutic drug monitoring: Update and recommendations of the STP-PT group of the SFPT and the GPCO-Unicancer]. / Suivi thérapeutique pharmacologique du 5-fluorouracile : mise au point et recommandations du groupe STP-PT de la SFPT et du GPCO-Unicancer.

Despite being 60-years old now, 5-FU remains the backbone of numerous regimen to treat a variety of solid tumors such as breast, head-and-neck and digestive cancers either in neo-adjuvant, adjuvant or metastatic settings. Standard 5-FU usually claims 15-40% of severe toxicities and up to 1% of toxic-death. Numerous studies show a stiff relationship between 5-FU exposure and toxicity or efficacy. In addition, 5-FU pharmacokinetics is highly variable between patients. Indeed, 80% of the 5-FU dose is catabolized in the liver by dihydropyrimidine dehydrogenase (DPD) into inactive compounds. It is now well established that DPD deficiency could lead to severe toxicities and, thus, require dose reduction in deficient patients. However, despite dosage adaptation based on DPD status, some patients may still experience under- or over-exposure, leading to inefficacy or major toxicity. The "Suivi thérapeutique pharmacologique et personnalisation des traitements" (STP-PT) group of the "Société française de pharmacologie et de thérapeutique" (SFPT) and the "Groupe de pharmacologie clinique oncologique" (GPCO)-Unicancer, based on the latest and most up-to-date literature data, recommend the implementation of 5-FU Therapeutic Drug Monitoring in order to ensure an adequate 5-FU exposure.

Simultaneous monitoring of pazopanib and its metabolites by UPLC-MS/MS.

Pazopanib is a multi-targeted tyrosine kinase inhibitor (TKI) approved as first-line treatment for patients with advanced renal cell carcinoma (RCC) and as second-line treatment for patients with advanced soft tissue sarcoma (STS) previously treated with chemotherapy. The most common adverse events, observed during the RCC and STS trials, were gastrointestinal disorders, hypertension, fatigue, elevated ALAT and ASAT, but the molecular mechanisms explaining pazopanib toxicity remain unclear. Therapeutic activity is considered to be mainly dependent on pazopanib exposure as the primary metabolites are inactive or display low plasma concentrations, but metabolites may be involved in toxicity as relationships between metabolite profiles and toxicity have not been evaluated. We report here, for the first time, the validation of a method for the simultaneous quantification of pazopanib and semi-quantification of its metabolites (relative determination). As there are no standards available, pazopanib metabolites were generated with human liver microsomes (HLM) to provide controls in the development of an UPLC-MS/MS method for monitoring both pazopanib and metabolites. The optimised method was validated for specificity, linearity, sensitivity, precision, accuracy, matrix effect and stability. The coefficient of variation (CV%) for intra-day and inter-day precision varied from 2.1% to 7.9% and 5.6% to 13.1% respectively. The biases varied from -12% to 2.3% (intra-day) and 3.8% to 13.1% (inter-day) for accuracy evaluation. Intra-day and inter-day precision CV were respectively 20.1% and 19.6% and accuracy biases were between 20.7% (intra-day) and 3.8% (inter-day) at the limit of quantification. The recoveries from matrix samples spiked with pazopanib were respectively 102.6 ±â€¯12.9% and 102.5 ±â€¯1.2% at low and high levels of calibration range. No matrix effect was evidenced as demonstrated by the normalised matrix factor values: 1.3 ±â€¯0.1 and 1.2 ±â€¯0.2 respectively measured at low and high part of calibration range. A good stability of pazopanib was observed during short term, long term and in process storage conditions and after three freeze/thaw cycles. The method was applied to clinical samples from three patients treated with pazopanib to establish the metabolite profiles (semi-quantitative data) during treatment. The assessment of metabolite profiles could be useful to improve our understanding of the occurrence of adverse events and to improve pazopanib pharmacokinetic-pharmacodynamic relationships.

Determination of unbound fraction of pazopanib in vitro and in cancer patients reveals albumin as the main binding site.

INTRODUCTION: Pazopanib exhibits wide inter-patient pharmacokinetic variability which may contribute to differences in treatment outcome. Unbound drug concentrations are believed to be more relevant to pharmacological responses than total concentrations. Thus it is desirable to evaluate pazopanib binding on plasma proteins and different factors potentially affecting this process. METHODS: An equilibrium dialysis method coupled with UPLC-MS/MS assay has been optimized and validated for the determination of pazopanib unbound fraction (fu%) in human plasma. Pazopanib binding in the plasma of healthy volunteers and in isolated protein solutions was investigated. The unbound fraction was determined for 24 cancer patients treated daily with pazopanib. RESULTS: We found that pazopanib was extensively bound in human plasma (>99.9 %) with a mean fu% value of 0.0106 ± 0.0013 % at 40 µg/mL. Protein binding was concentration independent over a clinically relevant range of concentrations. In isolated protein solutions, pazopanib at 40 µg/mL was mainly bound to albumin (40 g/L) and to a lesser extent to α1-acid glycoprotein (1 g/L) and low density lipoproteins (1.2 g/L), with a mean fu% of 0.0073 ± 0.0022 %, 0.992 ± 0.44 % and 7.4 ± 1.7 % respectively. Inter-patient variability (CV%) of fu% in cancer patients was limited (27.2 %). A correlation was observed between individual unbound fraction values and albuminemia. CONCLUSIONS: Pazopanib exhibits extensive binding to plasma proteins in human plasma. Variable albumin concentrations, frequently observed in cancer patients, may affect pazopanib unbound fraction with implications for inter-patient variability in drug efficacy and toxicity.

Is spontaneous reporting always the most important information supporting drug withdrawals for pharmacovigilance reasons in France?

PURPOSE: The objective of our study was to determine the nature of scientific evidence leading to drug withdrawal for safety reasons in France (between 2005 and 2011). METHODS: Drugs (i.e., active ingredients) withdrawn were identified from the Web site of the French Health Products Agency. Additional information allowed us to classify these withdrawals according to the nature of evidence as clinical trials (CT), case reports/case series (CR/CS), case-control studies (CC), cohort, animal, or observational studies. RESULTS: A total of 22 active ingredients were withdrawn from the French market between 2005 and 2011. The nature and type of adverse drug reactions (ADRs) leading more frequently to drug withdrawal were cardiovascular (10-fold), neurological (5-fold), or hepatic, cutaneous, or psychiatric (3-fold each) ADRs. CR (19/22; 86.4%) and CT (13/22; 59.1%) were the most frequently involved methods. In 5 of 22 (23%) cases, CR were the sole evidence. However, 68% (15/22) of regulatory decisions were based on multiple sources of evidence: For example, data from CR + CT were used in eight cases. CC or cohort studies were used in only five cases. CONCLUSION: This study underlines that spontaneous reporting remained the most important source of drug withdrawals between 2005 and 2011. However, its relative importance decreased in comparison with that in 1997-2004. The importance of pharmacoepidemiological methods slightly increased but remained low. Finally, regulatory authorities seem to have more frequently based their safety decisions on multiple sources of evidence than before.