PharmBERT: a domain-specific BERT model for drug labels.

Human prescription drug labeling contains a summary of the essential scientific information needed for the safe and effective use of the drug and includes the Prescribing Information, FDA-approved patient labeling (Medication Guides, Patient Package Inserts and/or Instructions for Use), and/or carton and container labeling. Drug labeling contains critical information about drug products, such as pharmacokinetics and adverse events. Automatic information extraction from drug labels may facilitate finding the adverse reaction of the drugs or finding the interaction of one drug with another drug. Natural language processing (NLP) techniques, especially recently developed Bidirectional Encoder Representations from Transformers (BERT), have exhibited exceptional merits in text-based information extraction. A common paradigm in training BERT is to pretrain the model on large unlabeled generic language corpora, so that the model learns the distribution of the words in the language, and then fine-tune on a downstream task. In this paper, first, we show the uniqueness of language used in drug labels, which therefore cannot be optimally handled by other BERT models. Then, we present the developed PharmBERT, which is a BERT model specifically pretrained on the drug labels (publicly available at Hugging Face). We demonstrate that our model outperforms the vanilla BERT, ClinicalBERT and BioBERT in multiple NLP tasks in the drug label domain. Moreover, how the domain-specific pretraining has contributed to the superior performance of PharmBERT is demonstrated by analyzing different layers of PharmBERT, and more insight into how it understands different linguistic aspects of the data is gained.

The Facioscapulohumeral Muscular Dystrophy-Health Index: Development and evaluation of a disease-specific outcome measure.

INTRODUCTION/AIMS: As promising therapeutic interventions are tested among patients with facioscapulohumeral muscular dystrophy (FSHD), there is a clear need for valid and reliable outcome tools to track disease progression and therapeutic gain in clinical trials and for clinical monitoring. Our aim was to develop and validate the Facioscapulohumeral Muscular Dystrophy-Health Index (FSHD-HI) as a multifaceted patient-reported outcome measure (PRO) designed to measure disease burden in adults with FSHD. METHODS: Through initial interviews with 20 individuals and a national cross-sectional study with 328 individuals with FSHD, we identified the most prevalent and impactful symptoms in FSHD. The most relevant symptoms were included in the FSHD-HI. We used patient interviews, test-retest reliability evaluation, known groups validity testing, and factor analysis to evaluate and optimize the FSHD-HI. RESULTS: The FSHD-HI contains 14 subscales that measure FSHD disease burden from the patient's perspective. Fourteen adults with FSHD participated in semistructured beta interviews and found the FSHD-HI to be clear, usable, and relevant to them. Thirty-two adults with FSHD participated in test-retest reliability assessments, which demonstrated the high reliability of the FSHD-HI total score (intraclass correlation coefficient = 0.924). The final FSHD-HI and its subscales also demonstrated a high internal consistency (Cronbach α = 0.988). DISCUSSION: The FSHD-HI provides researchers and clinicians with a reliable and valid mechanism to measure multifaceted disease burden in patients with FSHD. The FSHD-HI may facilitate quantification of therapeutic effectiveness, as demonstrated by its use as a secondary and exploratory measure in several clinical trials.

The FDA Metformin Label Change and Racial and Sex Disparities in Metformin Prescription among Patients with CKD.

BACKGROUND: In 2016, the Food and Drug Administration (FDA) changed labeling regarding metformin contraindications in patients with diabetes and CKD from using serum creatinine-based thresholds to using eGFR-based thresholds. Because race and sex affect serum creatinine levels independently of GFR, the earlier creatinine-based contraindication may have inadvertently caused racial and sex disparities in metformin prescription among patients with low eGFR. METHODS: In an analysis of 15,946 Black and White primary care patients with diabetes and eGFR≥30 ml/min per 1.73 m2 in a large health system (the primary cohort), we assessed the association of race and sex with metformin prescription across eGFR level before and after the FDA label change. For a replication cohort, we meta-analyzed data from 36 cohorts with 1,051,723 patients from OptumLabs Data Warehouse. RESULTS: In the primary cohort, before the label change, Black patients with eGFR of 30-44 ml/min per 1.73 m2 were prescribed metformin less often than White counterparts (adjusted prevalence ratio [aPR], 0.65; 95% confidence interval [95% CI], 0.52 to 0.82); this disparity was significantly attenuated after the label change (aPR, 0.90; 95% CI, 0.74 to 1.09; P value for interaction by period =0.04). Results were consistent in the replication cohorts. Men with eGFR of 30-44 ml/min per 1.73 m2 received metformin prescriptions less often than women counterparts before the label change; this was nonsignificantly attenuated after the label change, but we found significant attenuation in the replication cohorts (aPRpre-label change, 0.76; 95% CI, 0.73 to 0.79; aPRpost-label change, 0.85; 95% CI, 0.83 to 0.88; P value for interaction by period <0.001). CONCLUSIONS: The metformin label change to an eGFR-based contraindication may have reduced racial and sex disparities in metformin prescription in moderate kidney dysfunction.

Pharmacogenetics in Neuroblastoma: What Can Already Be Clinically Implemented and What Is Coming Next?

Pharmacogenetics is one of the cornerstones of Personalized Precision Medicine that needs to be implemented in the routine of our patients' clinical management in order to tailor their therapies as much as possible, with the aim of maximizing efficacy and minimizing toxicity. This is of great importance, especially in pediatric cancer and even more in complex malignancies such as neuroblastoma, where the rates of therapeutic success are still below those of many other types of tumors. The studies are mainly focused on germline genetic variants and in the present review, state of the art is presented: which are the variants that have a level of evidence high enough to be implemented in the clinic, and how to distinguish them from the ones that still need validation to confirm their utility. Further aspects as relevant characteristics regarding ontogeny and future directions in the research will also be discussed.

Fall-risk-increasing adverse reactions-is there value in easily accessible drug information? A case-control study.

PURPOSE: The individual fall risk of a patient is often multifactorial. Polymedication contributes to an additional risk of fall-risk-increasing adverse reactions (FRIARs). Previous studies have not sufficiently investigated the complexity facing prescribers when balancing the therapeutic benefits of individual drugs against their potential fall risk. METHODS: An expert panel identified drugs with FRIARs based on the Summary of Product Characteristics (SmPC). These FRIARs and other parameters (such as the total number of drugs, dosage, dose adjustments, and drug changes) were then analyzed for their impact on falls in a case-control study using logistic regression. RESULTS: During a 1-year period, 112 (1%) of 11,481 hospital patients experienced at least one fall event. Complete data was available for evaluation from 87 of them (case group). We matched these patients to another 87 patients who had no fall events (control group). FRIAR drugs were more frequently prescribed in the case group (4.26 (Q25-Q75, 3.75-4.78) per patient; p = 0.033) than in the control group (3.48 (2.97-3.99)). Drugs with FRIARs (ß = 0.137; p = 0.035) and the total number of FRIARs (ß = 0.033; p = 0.031) increased the fall risk. The total number of drugs, dosage, dose adjustments, and drug changes showed no influence. CONCLUSIONS: FRIARs were associated with a higher number of falls. To consider FRIARs offers a chance to address the complexity of the individual medication. This data can support future computerized physician order entries with clinical decision support.

Pharmacogenomic information in the Warning section of drug labels: A comparison between labels in the United States and those in five other countries/regions.

WHAT IS KNOWN AND OBJECTIVE: Clinically validated pharmacogenomic information useful for patient selection and/or dose adjustment is included in drug labels. However, the label information may differ among countries. This commentary summarizes the pharmacogenomic information on drug labels in different countries. COMMENT: We selected six drugs, namely, clopidogrel, atomoxetine, irinotecan, mercaptopurine, abacavir and carbamazepine and compared the pharmacogenomic information in the "Warning" section of these drug labels in the United States and 5 other countries/regions. WHAT IS NEW AND CONCLUSION: The pharmacogenomic information in drug labels is not well harmonized across countries/regions, possibly due to differences in population characteristics such as relevant allele frequencies, variable genetic test availability and differences in insurance coverage. Further and periodical investigations of this issue would be useful.

Definition of variables required for comprehensive description of drug dosage and clinical pharmacokinetics.

PURPOSE: Electronic clinical decision support systems (CDSS) require drug information that can be processed by computers. The goal of this project was to determine and evaluate a compilation of variables that comprehensively capture the information contained in the summary of product characteristic (SmPC) and unequivocally describe the drug, its dosage options, and clinical pharmacokinetics. METHODS: An expert panel defined and structured a set of variables and drafted a guideline to extract and enter information on dosage and clinical pharmacokinetics from textual SmPCs as published by the European Medicines Agency (EMA). The set of variables was iteratively revised and evaluated by data extraction and variable allocation of roughly 7% of all centrally approved drugs. RESULTS: The information contained in the SmPC was allocated to three information clusters consisting of 260 variables. The cluster "drug characterization" specifies the nature of the drug. The cluster "dosage" provides information on approved drug dosages and defines corresponding specific conditions. The cluster "clinical pharmacokinetics" includes pharmacokinetic parameters of relevance for dosing in clinical practice. A first evaluation demonstrated that, despite the complexity of the current free text SmPCs, dosage and pharmacokinetic information can be reliably extracted from the SmPCs and comprehensively described by a limited set of variables. CONCLUSION: By proposing a compilation of variables well describing drug dosage and clinical pharmacokinetics, the project represents a step forward towards the development of a comprehensive database system serving as information source for sophisticated CDSS.

Strength of evidence for labeled dosing recommendations in renal impairment.

BACKGROUND/AIMS: Renally excreted medications often require dose adjustment in patients with kidney impairment. While drug development and approval in the United States are typically based on several Phase I and II studies and one or more larger Phase III randomized trials, the basis for labeled dosing recommendations for patients with renal impairment is less well known. In response, we aimed to quantify the level of evidence used to recommend labeled dosing adjustments for newly approved drugs in patients with renal impairment. METHODS: We reviewed publicly available drug labels and approval packages for new molecular entities approved in the United States between 2012 and 2014. The sample was restricted to 29 renally excreted new molecular entities that were not granted orphan drug status. We extracted data regarding approved indications, normal dosing, dosing adjustments for patients with mild (estimated glomerular filtration rate >60 mL/min/1.73 m2), moderate (estimated glomerular filtration rate 30-<60 mL/min/1.73 m2), and severe (estimated glomerular filtration rate <30 mL/min/1.73 m2) renal impairment, characteristics of studies used to justify dosing adjustments, and numbers of subjects in each study. RESULTS: In all, 14 of 29 (48%) new molecular entities had labels that recommended dosing adjustments for patients with mild, moderate, and/or severe renal impairment. Among these 14 new molecular entities, 4 (29%) used only pharmacokinetic studies to justify the recommendations, with no examination of clinical outcomes for patients with renal impairment. Where data were available, the median number of patients with renal impairment evaluated in studies used for dosing adjustment was 34 (range, 4-5976). Of the 15 new molecular entities with no recommended dosing adjustments for this population, 2 (13%) did not report assessing the effects of renal impairment. CONCLUSION: Nearly half of newly approved renally excreted drugs include dosing adjustments for kidney impairment on the label, but the recommendations are usually based on very small numbers of patients and often utilize pharmacokinetic studies alone. More research is needed to understand the benefits and risks of new drugs in patients with renal impairment.

Mediation effects of medication information processing and adherence on association between health literacy and quality of life.

BACKGROUND: To examine whether medication related information processing defined as reading of over-the-counter drug labels, understanding prescription instructions, and information seeking-and medication adherence account for the association between health literacy and quality of life, and whether these associations may be moderated by age and gender. METHODS: A sample of 305 adults in South Korea was recruited through a proportional quota sampling to take part in a cross-sectional survey on health literacy, medication-related information processing, medication adherence, and quality of life. Descriptive statistics and structural equation modeling (SEM) were performed. RESULTS: Two mediation pathways linking health literacy with quality of life were found. First, health literacy was positively associated with reading drug labels, which was subsequently linked to medication adherence and quality of life. Second, health literacy was positively associated with accurate understanding of prescription instructions, which was associated with quality of life. Age moderation was found, as the mediation by reading drug labels was significant only among young adults whereas the mediation by understanding of medication instruction was only among older adults. CONCLUSION: Reading drug labels and understanding prescription instructions explained the pathways by which health literacy affects medication adherence and quality of life. The results suggest that training skills for processing medication information can be effective to enhance the health of those with limited health literacy.

A cross-national comparison of 17 countries' insulin glargine drug labels.

PURPOSE: Type 2 diabetes mellitus has reached epidemic proportions worldwide. Many patients with type 2 diabetes mellitus will require insulin, and the evidence-based use of insulin is described in the prescription drug label. Product labels in different countries may provide inconsistent information. We evaluated the variability in drug label content for one brand of basal insulin across diverse settings. METHODS: We examined the drug label content pertinent to effective and safe use of insulin glargine across 17 countries: Abu Dhabi (United Arab Emirates), Argentina, Brazil, Canada, China, Germany, Israel, Italy, Japan, Mexico, Russia, Saudi Arabia, South Korea, Spain, Turkey, UK, and the USA. We compared label characteristics in settings where drug labels were governed by a local regulatory authority versus countries where labels were administered by a regional body or adopted from another locale. RESULTS: All 17 labels cautioned that providers should consider age, illness, diet, and exercise when prescribing. Only two (12%) described care of the fasting patient. Caution was urged for patients with renal or hepatic impairment in 16 (94%) labels. Four (24%) did not describe responses to missed doses, and five (29%) failed to recommend patient counseling about the risk of hypoglycemia. Labels emerging from regional or adopted regulatory bodies reported fewer patients in efficacy studies than did labels from settings with their own drug regulatory agencies (365 ± 0 patients vs. 3560 ± 2938, p = 0.04). CONCLUSIONS: There is substantial variation in the content of drug labels for glargine, which may lead to international inconsistency in quality of care for diabetic patients.

A Longitudinal Analysis of Black Box Warnings: Trends and Implications for Drug Safety.

A black box warning, signaling potential life-threatening adverse effects of medications or medical devices, is crucial for public and healthcare professional awareness. Comprehending and adhering to these warnings can prevent serious harm. This review aims to elucidate their significance. Data on drugs with black box warnings were collected from the Food and Drug Administration's (FDA's) official website using the search term 'Boxed warnings' from January 1, 2015, to January 31, 2024. A Microsoft Excel spreadsheet (Microsoft Corporation, Redmond, WA, USA) containing black box warnings for this period was downloaded from the FDA's website. Additional parameters, such as drug class and whether the warnings were new or existing, were added to the downloaded spreadsheet. The collected data were organized by year, categorizing new and existing warnings, along with details on the evidence source, system-wise classification, and black box warnings for commonly used drugs, including their clinical significance. Results show that in the past decade, 40% of black box warnings were issued in 2023, followed by 12% in 2022. Most warnings (67%) comprised existing ones with minor revisions while 29% were new. Nine existing warnings were removed during the period. Post-marketing studies predominantly provided evidence for these warnings. Neuropsychiatric concerns like addiction potential (31%), suicidal tendency (7%), and hypersensitivity reactions (12%) were the frequently encountered black box warnings. Black box warnings play a crucial role in highlighting the serious adverse effects of medications. Neuropsychiatric warnings have been frequent over the past decade. Awareness of these warnings is essential to prevent adverse effects and enhance patient care, especially concerning drugs like guaifenesin/hydrocodone bitartrate, zolpidem, and montelukast commonly encountered in clinical practice.

ADME Gene-Related Pharmacogenomic Labeling of FDA-Approved Drugs: Comparison with Clinical Pharmacogenetics Implementation Consortium (CPIC) Evidence Levels.

Pharmacogenomics (PGx) can facilitate the transition to patient-specific drug regimens and thus improve their efficacy and reduce toxicity. The aim of this study was to evaluate the overlap of PGx classification for drug absorption, distribution, metabolism, and elimination (ADME)-related genes in the U.S. Food and Drug Administration (FDA) PGx labeling and in the Clinical Pharmacogenetics Implementation Consortium (CPIC) database. FDA-approved drugs and PGx labeling for ADME genes were identified in the CPIC database. Drugs were filtered by their association with ADME (pharmacokinetics)-related genes, PGx FDA labeling class, and CPIC evidence level. FDA PGx labeling was classified as either actionable, informative, testing recommended, or testing required, and varying CPIC evidence levels as either A, B, C, or D. From a total of 442 ADME and non-ADME gene-drug pairs in the CPIC database, 273, 55, and 48 pairs were excluded for lack of FDA labeling, mixed CPIC evidence level provisional classification, and non-ADME gene-drug pairs, respectively. The 66 ADME gene-drug pairs were classified into the following categories: 10 (15%) informative, 49 (74%) actionable, 6 (9%) testing recommended, and 1 (2%) testing required. CYP2D6 was the most prevalent gene among the FDA PGx labeling. From the ADME gene-drug pairs with both FDA and CPIC PGx classification, the majority of the drugs were for depression, cancer, and pain medications. The ADME gene-drug pairs with FDA PGx labeling considerably overlap with CPIC classification; however, a large number of ADME gene-drug pairs have only CPIC evidence levels but not FDA classification. PGx actionable labeling was the most common classification, with CYP2D6 as the most prevalent ADME gene in the FDA PGx labeling. Health professionals can impact therapeutic outcomes via pharmacogenetic interventions by analyzing and reconciling the FDA labels and CPIC database.

Learning with an evolving medicine label: how artificial intelligence-based medication recommendation systems must adapt to changing medication labels.

INTRODUCTION: Artificial intelligence or machine learning (AI/ML) based systems can help personalize prescribing decisions for individual patients. The recommendations of these clinical decision support systems must relate to the "label" of the medicines involved. The label of a medicine is an approved guide that indicates how to prescribe the drug in a safe and effective manner. AREAS COVERED: The label for a medicine may evolve as new information on drug safety and effectiveness emerges, leading to the addition or removal of warnings, drug-drug interactions, or to permit new indications. However, the speed at which these updates are made to these AI/ML recommendation systems may be delayed and could influence the safety of prescribing decisions. This article explores the need to keep AI/ML tools 'in sync' with any label changes. Additionally, challenges relating to medicine availability and geographical suitability are discussed. EXPERT OPINION: These considerations highlight the important role that pharmacoepidemiologists and drug safety professionals must play within the monitoring and use of these tools. Furthermore, these issues highlight the guiding role that regulators need to have in planning and oversight of these tools.

Artificial intelligence or machine learning (AI/ML) based systems that guide the prescription of medications have the potential to vastly improve patient care, but these tools should only provide recommendations that are in line with the label of a medicine. With a constantly evolving medication label, this is likely to be a challenge, and this also has implications for the off-label use of medicines.

The FDA ought to change Plan B's label.

This commentary defends 3 arguments for changing the label of levonorgestrel-based emergency contraception (LNG EC) so that it no longer supports the possibility of a mechanism of action after fertilization. First, there is no direct scientific evidence confirming any postfertilization mechanisms. Second, despite the weight of evidence, there is still widespread public misunderstanding over the mechanism of LNG EC. Third, this FDA label is not a value-free claim, but instead it has functioned like a political tool for reducing contraceptive access. The label is laden with antiabortion values (even though EC is contraception, not abortion), and it imposes these values on potential users, resulting in barriers to access such as with Burwell v. Hobby Lobby. These 3 arguments together provide scientific, social, and ethical grounds for the FDA to take the initiate in changing Plan B's drug label.

Implementation of Pharmacogenomic Information on Stevens-Johnson Syndrome and Toxic Epidermal Necrolysis.

Drug-related Stevens-Johnson syndrome and toxic epidermal necrolysis (SJS/TEN) are rare but severe adverse drug reactions, termed as idiosyncratic reactions; however, predicting their onset remains challenging. Pharmacogenomic information associated with SJS/TEN has accumulated on several drugs in the last 15 years, with clinically useful information now included on drug labels in several countries/regions or guidelines of the Clinical Pharmacogenetics Implementation Consortium (CPIC) for implementation. However, label information might be different among countries. This mini-review summarizes pharmacogenomic information on drug labels of five drugs in six countries and compared descriptions of drug labels and CPIC guidelines. Finally, we discuss future perspectives of this issue. Pharmacogenomic information on drug labels is not well-harmonized across countries/regions, but CPIC guidelines are a scientifically sound goal for future pharmacogenomic implementation.

Clinical Approach to Cardiovascular Toxicity of Oral Antineoplastic Agents: JACC State-of-the-Art Review.

Precision medicine has ushered in a new era of targeted treatments for numerous malignancies, leading to improvements in overall survival. Unlike traditional chemotherapy, many molecular targeted antineoplastic agents are available in oral formulation, leading to enhanced patient convenience and a perception of reduced risk of adverse effects. Although oral antineoplastic agents are generally well-tolerated, cardiovascular toxicities are being reported with increasing frequency in part due to U.S. Food and Drug Administration and manufacturer recommended cardiac monitoring. Monitoring strategies have focused on left ventricular dysfunction, hypertension, and QT prolongation/arrhythmias. Given the rapid pace of development and availability of new oral antineoplastic agents, the purpose of this review is to provide clinicians with an up-to-date practical approach to monitoring and management of cardiovascular toxicities with the aim of improving overall outcomes for patients with cancer.

Comparison of signal detection of tumour necrosis factor-α inhibitors using the Korea Adverse Events Reporting System Database, 2005-2016.

OBJECTIVES: There are no pharmacovigilance studies on adverse event (AE) data for tumour necrosis factor alpha (TNFα) inhibitors in South Korea. We analysed AEs induced by adalimumab, infliximab, and etanercept METHODS: We used data from the Korea Institute of Drug Safety and Risk Management-Korea Adverse Events Reporting System Database (KIDS-KD) collected between 2005 and 2016. We used three different signal detection methods: proportional reporting ratio (PRR), reporting odds ratio (ROR), and information component (IC). The drug was compared with drug labels in the USA and Korea. Logistic regression analysis identified AEs that are more likely to occur with drug use. RESULTS: Of the 5594 AEs identified, 350, 452, and 71 were related to adalimumab, infliximab, and etanercept, respectively. We identified seven new signals, which were not listed on drug labels in either Korea or the USA, for AEs associated with the study drugs: two for adalimumab (medication error and drug failure), two for infliximab (palpitation and temperature sensation change), and three for etanercept (hyperkeratosis, acne, and thyroid neoplasm malignant). Injection site pain (OR 6.14, 95% CI 1.15-32.74) and alopecia (OR 4.54, 95% CI 1.16-17.77) for adalimumab, chest pain (OR 6.01, 95% CI 1.35-26.77) for infliximab, and uveitis (OR 10.11, 95% CI 1.13-90.77) for etanercept were more likely to be reported in patients with each TNFα inhibitor than in those without, respectively. CONCLUSIONS: Seven new signals that were not included in the current label were identified for TNFα inhibitors and should be updated and monitored.Key Points• Large-scale spontaneous AE reporting data and data mining techniques are useful for detecting signals of rare AEs as well as common AEs induced by drugs.• Drug labels should be updated to reflect signals that are newly discovered by continuous monitoring.

[Pharmacogenetics for patient care in France: A discipline that evolves!] / La pharmacogénétique au service du soin en France : une discipline qui évolue !

Pharmacogenetics, which concepts are known for a long time, is entering a new period at least as far as its practical applications for patients are concerned. In recent years there have been more and more initiatives to promote widespread dissemination, and health authorities are increasingly incorporating these concepts into drug labels. In France, the national network of pharmacogenetics (RNPGx) works to promote these activities, both with health actors (biologists, clinicians) and health authorities. This article reviews the current situation in France and the milestones of the year 2018. It highlights recent advances in this field, in terms of currently recommended analyses, sharing of information or technological developments, and the prospects for future developments in the near future from targeted pharmacogenetics to eventually preemptive approaches.

Colorectal cancer: pharmacogenetics support for the correct drug prescription.

Pharmacogenetics (PGx) in clinical practice is a tool that the clinician can use to guide, in a personalized way, the most suitable treatment that will be administered to the patient. The objective of this review is to summarize in a practical and accessible rational way, the advances that currently exist for the application of PGx in colorectal cancer chemotherapy management through the study of the patients' germline polymorphisms. To define the polymorphisms that can be applied, we rely on three fundamental cornerstones: the recommendations of drug regulatory agencies; the implementation guidelines prepared by expert consortia in PGx and information from clinical annotations (the drug/polymorphism relation) according to the scientific level of evidence assigned by PharmGKB experts.