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.

Hepatotoxicity of Antibiotics and Antifungals and Their Safe Use in Hepatic Impairment.

Idiosyncratic drug-induced liver injury (DILI) is a rare and unpredictable form of hepatotoxicity. While its clinical course is usually benign, cases leading to liver transplantation or death can occur. Based on modern prospective registries, antimicrobials including antibiotics and antifungals are frequently implicated as common causes. Amoxicillin-clavulanate ranks as the most common cause for DILI in the Western World. Although the absolute risk of hepatotoxicity of these agents is low, as their usage is quite high, it is not uncommon for practitioners to encounter liver injury following the initiation of antibiotic or antifungal therapy. In this review article, mechanisms of hepatoxicity are presented. The adverse hepatic effects of well-established antibiotic and antifungal agents are described, including their frequency, severity, and pattern of injury and their HLA risks. We also review the drug labeling and prescription guidance from regulatory bodies, with a focus on individuals with hepatic impairment.

Black Box Warning: Large Language Models and the Future of Infectious Diseases Consultation.

Large language models (LLMs) are artificial intelligence systems trained by deep learning algorithms to process natural language and generate text responses to user prompts. Some approach physician performance on a range of medical challenges, leading some proponents to advocate for their potential use in clinical consultation and prompting some consternation about the future of cognitive specialties. However, LLMs currently have limitations that preclude safe clinical deployment in performing specialist consultations, including frequent confabulations, lack of contextual awareness crucial for nuanced diagnostic and treatment plans, inscrutable and unexplainable training data and methods, and propensity to recapitulate biases. Nonetheless, considering the rapid improvement in this technology, growing calls for clinical integration, and healthcare systems that chronically undervalue cognitive specialties, it is critical that infectious diseases clinicians engage with LLMs to enable informed advocacy for how they should-and shouldn't-be used to augment specialist care.

Warning Labels as a Public Health Intervention: Effects and Challenges for Tobacco, Cannabis, and Opioid Medications.

Warning labels help consumers understand product risks, enabling informed decisions. Since the 1966 introduction of cigarette warning labels in the United States, research has determined the most effective message content (health effects information) and format (brand-free packaging with pictures). However, new challenges have emerged. This article reviews the current state of tobacco warning labels in the United States, where legal battles have stalled pictorial cigarette warnings and new products such as electronic cigarettes and synthetic nicotine products pose unknown health risks. This article describes the emerging research on cannabis warnings; as more places legalize recreational cannabis, they are adopting lessons from tobacco warnings. However, its uncertain legal status and widespread underestimation of harms impede strict warning standards. The article also reviews opioid medication warning labels, suggesting that lessons from tobacco could help in the development of effective and culturally appropriate FDA-compliant opioid warning labels that promote safe medication use and increased co-dispensing of naloxone.

Effectiveness of direct delabelling of allergy labels in type A adverse drug reactions to penicillin: a multicentre hospitalwide prospective cohort study.

BACKGROUND: Patient-reported penicillin allergy labels (PALs) are associated with adverse patient outcomes and inappropriate antibiotic prescribing. Removal of PALs via direct oral challenge (DOC) is associated with increased penicillin utilization post removal. OBJECTIVES: To assess the impact of direct delabelling (allergy label removal via medical reconciliation alone) of type A adverse drug reaction (ADR) PALs on inpatient prescribing. METHODS: From January 2019 to December 2022 at two tertiary hospitals in Melbourne, patients aged ≥18 years with type A ADR PALs, as defined by the validated Antibiotic Allergy Assessment Tool, were offered direct delabelling or single-dose DOC. The primary endpoint was antibiotic use pre- and post-assessment (during index admission and 90 days post assessment). The secondary endpoint was the proportion of patients delabelled in the direct delabelling and DOC cohorts in the electronic medical record at 90 days post assessment. RESULTS: Allergy labels (n = 4108) were assessed for 488 participants, with 490 individual type A ADR PAL assessments included. Three hundred and thirty-seven patients were directly delabelled, 69 underwent DOC and 84 were not delabelled. There was increased use of any penicillin following direct delabelling (OR 19.19, 95% CI 2.48-148.36) and DOC (OR 56.98, 95% CI 6.82-476.19) during the index admission, higher in the DOC group compared with direct delabelling (OR 2.97, 95% CI 1.39-6.37). Relabelling at 90 days was low with no statistically significant difference between direct delabelling (5/337; 1.5%) and DOC (0/69; 0%). CONCLUSIONS: Both direct delabelling and DOC of type A ADR PALs increased penicillin usage; however, the impact was greatest with DOC. Most patients remain delabelled at 90 days.

Pharmacogenomic biomarker information on drug labels of the Spanish Agency of Medicines and Sanitary products: evaluation and comparison with other regulatory agencies.

This work aimed to analyse the pharmacogenetic information in the Spanish Drug Regulatory Agency (AEMPS) Summary of Products Characteristics (SmPC), evaluating the presence of pharmacogenetic biomarkers, as well as the associated recommendations. A total of 55.4% of the 1891 drug labels reviewed included information on pharmacogenetic biomarker(s). Pharmacogenomic information appears most frequently in the "antineoplastic and immunomodulating agents", "nervous system", and "cardiovascular system" Anatomical Therapeutic Chemical groups. A total of 509 different pharmacogenetic biomarkers were found, of which CYP450 enzymes accounted for almost 34% of the total drug-biomarker associations evaluated. A total of 3679 drug-biomarker pairs were identified, 102 of which were at the 1A level (PharmGKB® classification system), and 33.33% of these drug-pharmacogenetic biomarker pairs were assigned to "actionable PGx", 12.75% to "informative PGx", 4.9% to "testing recommended", and 4.9% to "testing required". The rate of coincidence in the assigned PGx level of recommendation between the AEMPS and regulatory agencies included in the PharmGKB® Drug Label Annotations database (i.e., the FDA, EMA, SWISS Medic, PMDA, and HCSC) ranged from 45% to 65%, being 'actionable level' the most frequent. On the other hand, discrepancies between agencies did not exceed 35%. This study highlights the presence of relevant pharmacogenetic information on Spanish drug labels, which would help avoid interactions, toxicity, or lack of treatment efficacy.

Estimating the impact of label design on reducing the risk of medication errors by applying HEART in drug administration.

Medication errors are one of the biggest problems in healthcare. The medicines' poor labelling design (i.e. look-alike labels) is a well-recognised risk for potential confusion, wrong administration, and patient damage. Human factors and ergonomics (HFE) encourages the human-centred design of system elements, which might reduce medication errors and improve people's well-being and system performance. OBJECTIVE: The aim of the present study is twofold: (i) to use a human reliability analysis technique to evaluate a medication administration task within a simulated scenario of a neonatal intensive care unit (NICU) and (ii) to estimate the impact of a human-centred design (HCD) label in medication administration compared to a look-alike (LA) label. METHOD: This paper used a modified version of the human error assessment and reduction technique (HEART) to analyse a medication administration task in a simulated NICU scenario. The modified technique involved expert nurses quantifying the likelihood of unreliability of a task and rating the conditions, including medicine labels, which most affect the successful completion of the task. RESULTS: Findings suggest that error producing conditions (EPCs), such as a shortage of time available for error detection and correction, no independent checking of output, and distractions, might increase human error probability (HEP) in administering medications. Results also showed that the assessed HEP and the relative percentage of contribution to unreliability reduced by more than 40% when the HCD label was evaluated compared to the LA label. CONCLUSION: Including labelling design based on HFE might help increase human reliability when administering medications under critical conditions.

Prescriber perceptions of boxed warnings: A qualitative study.

PURPOSE: To explore how boxed warning (BW) information fits within the context of prescribers' overall treatment decision-making and communication with patients. METHODS: In-depth interviews (N = 52) were conducted with primary care providers and specialists. Participants were presented with one of two prescribing scenarios: (1) estrogen vaginal inserts to treat vulvovaginal atrophy (VVA) associated with menopause; or (2) direct-acting antivirals (DAA) to treat chronic hepatitis C virus infection (HCV). The semi-structured interviews explored participants' treatment decision-making within the scenario, reactions to current prescribing information for a product within the FDA-approved drug class, as well as their perceptions of BWs generally. RESULTS: Across scenarios, providers described that the BW is only one of several factors that influence treatment decision-making. In the VVA scenario, symptom severity, family history, and experience with nonprescription drugs were raised as common factors that influence prescribing considerations; compared to comorbid infections, viral load, and HCV genotype in the HCV scenario. Perceptions of the DAA BW were generally positive or neutral, as many participants found the information important and appropriate. The VVA BW was viewed less favorably, with many participants stating the BW overstates the risk for this drug. CONCLUSIONS: Findings suggest that BWs are one of several factors that influence providers' treatment decisions, and BW influence largely depends on context. Providers across scenarios expressed notable differences in their perceptions of the risk information provided in the presented BWs; however, across scenarios participants expressed consideration of how patients may perceive the BW.

Use of Hydrochlorothiazide in the United States Following Label Update About Skin Cancer Risk.

PURPOSE: On August 20, 2020, the United States (U.S.) Food and Drug Administration (FDA) issued a Drug Safety Communication (DSC) along with labeling updates to inform the public about a small increased risk of non-melanoma skin cancer (NMSC) associated with hydrochlorothiazide (HCTZ) use. This study aims to assess whether the DSC impacted HCTZ use in the U.S. METHODS: We conducted a trend analysis in the Sentinel Distributed Database using national healthcare administrative data from January 2017 to November 2022. We identified two cohorts each month: An overall cohort of all enrollees and a skin cancer cohort of those with a history of NMSC. For each cohort, we plotted the monthly proportion of patients receiving HCTZ-containing products among those receiving any thiazide diuretics. We performed interrupted time series analyses to quantify the impact of the DSC on these monthly proportions. Secondary analyses were conducted on the proportion of HCTZ users among patients receiving any antihypertensives. RESULTS: In the overall cohort, the DSC was only associated with a statistically significant but clinically negligible trend change of monthly HCTZ proportion within this cohort (0.018%; 95% CI, 0.012%-0.025%). Similar results were observed in the skin cancer cohort. The secondary analysis found no significant level change or trend change in the monthly proportion of HCTZ use among antihypertensive users. CONCLUSIONS: We did not observe significant changes in HCTZ use following the DSC about its NMSC risk, among the overall population and those with a history of NMSC. Our findings were in accordance with the DSC recommendation.

Safety Evaluation and Information Provision for Appropriate Drug Usage in Elderly Patients in Japan.

The elderly Japanese population is growing rapidly due to increasing longevity and declining birth rates. These findings have implications for drug development and safety in elderly patients; however, the Japanese stakeholders have been slow to adapt. This study aimed at examining methods for providing sufficient information on safe use of pharmaceuticals in elderly patients in Japan. For new drugs recently approved in Japan for diseases with a high prevalence among the elderly, we investigated the state of safety information provision for elderly patients through the package insert and also safety data evaluation in elderly patients in clinical studies. Of the 64 targeted drugs, only 14 provided geriatric use information based on clinical study data or indication, 38 had general cautionary descriptions, and 12 did not have geriatric use information. Most drugs met the recommendation of enrolling >100 elderly patients in the clinical development program. However, a discrepancy was observed in the proportion of elderly patients in clinical trials compared to that in real-world clinical setting. Twenty-nine drugs compared the incidence of key adverse events (AEs) in elderly and younger patients, whereas 25 only reported the overall incidence of AEs. To improve healthcare outcomes, healthcare professionals need access to sufficient safety information through package inserts containing data from clinical trials. Marketing authorization holders and regulatory authorities must work together to ensure that such safety information based on sufficient data is included in package inserts.

Streamlining Considerations for Safety Measures: A Predictive Model for Addition of Clinically Significant Adverse Reactions to Japanese Drug Package Inserts.

The addition of clinically significant adverse reactions (CSARs) to Japanese package inserts (PIs) is an important safety measure that can be used to inform medical personnel of potential health risks; however, determining the necessity of their addition can be lengthy and complex. Therefore, we aimed to construct a machine learning-based model that can predict the addition of CSARs at an early stage due to the accumulation of both Japanese and overseas adverse drug reaction (ADR) cases. The target comprised CSARs added to PIs from August 2011 to March 2022. The control group consisted of drugs without the same CSARs in their PIs by March 2022. Features were generated using ADR case accumulation data obtained from the Japanese Adverse Drug Event Report and the U.S. Food and Drug Administration Adverse Event Reporting System databases. The model was constructed using DataRobot, and its performance evaluated using the Matthews correlation coefficient. The target for the addition of CSARs included 414 cases, comprising 302 due to domestic case accumulation, 22 due to both domestic and overseas case accumulation, 12 due to overseas case accumulation, and 78 due to revisions of the company core data sheet. The best model was a generalized linear model with informative features, achieving a cross-validation of 0.8754 and a holdout of 0.8995. In conclusion, the proposed model effectively predicted CSAR additions to PIs resulting from the accumulation of ADR cases using data from both Japan and the United States.

A framework enabling LLMs into regulatory environment for transparency and trustworthiness and its application to drug labeling document.

Regulatory agencies consistently deal with extensive document reviews, ranging from product submissions to both internal and external communications. Large Language Models (LLMs) like ChatGPT can be invaluable tools for these tasks, however present several challenges, particularly the proprietary information, combining customized function with specific review needs, and transparency and explainability of the model's output. Hence, a localized and customized solution is imperative. To tackle these challenges, we formulated a framework named askFDALabel on FDA drug labeling documents that is a crucial resource in the FDA drug review process. AskFDALabel operates within a secure IT environment and comprises two key modules: a semantic search and a Q&A/text-generation module. The Module S built on word embeddings to enable comprehensive semantic queries within labeling documents. The Module T utilizes a tuned LLM to generate responses based on references from Module S. As the result, our framework enabled small LLMs to perform comparably to ChatGPT with as a computationally inexpensive solution for regulatory application. To conclude, through AskFDALabel, we have showcased a pathway that harnesses LLMs to support agency operations within a secure environment, offering tailored functions for the needs of regulatory research.

Removing penicillin allergy label in a hospitalized adolescent with a remote penicillin and recent cephalosporin allergy.

Background: ß-Lactam antibiotics are widely used with increased utilization in hospitalized patients. Of this population, as high as 10-20% report an allergy to ß-lactam antibiotics but <5% are at risk of developing clinically significant immunoglobulin E- or T-lymphocyte-mediated reactions. Most of the time, these reported allergies are present during an illness with no previous inquiry of their validity, which makes investigation and possible removal of this allergy label a challenge. Methods: We report a 16-year-old boy who presented with 1 week of night sweats, chills, headaches, and fatigue, followed by 1 day of fever and right knee swelling and who was diagnosed with septic bursitis. Due to concern of a penicillin allergy label, the patient was started on a cefepime infusion. Five minutes into the infusion, the patient reported puffy eyes and itchy throat, followed by a witnessed cascading flat nonpruritic erythematous rash from head to shoulders. This rash went away in 3 minutes after stopping the infusion and the patient being given 50 mg of intravenous diphenhydramine and 10 mg of oral dexamethasone. He was subsequently diagnosed with a cefepime allergy. Results: Allergy/immunology was the speciality consulted, and, by using a screening questionnaire, the patient's reported penicillin allergy was determined to be low risk. Subsequent 1-step oral challenge was the key to providing the patient with the necessary antibiotic course to resolve his infection. Conclusion: Multiple reported antibiotic allergies lead to poor antibiotic stewardship that causes impactful health and financial burden on the patient and health-care system. It is thus important to have an evidence-based systematic approach to de-label penicillin antibiotic allergy labels to reduce these potential harms.

Acceptance of Electronic Labeling for Medicinal Product Information Among Malaysian Hospital Patients: Cross-Sectional Study.

BACKGROUND: While perceptions of electronic labeling (e-labeling) in developed countries have been generally positive, existing data primarily come from studies involving hospital pharmacists, community pharmacy customers who may not be frequent medication users, and individuals receiving COVID-19 vaccines. OBJECTIVE: This study aims to assess e-labeling acceptance, perceptions of its benefits, challenges with its implementation, and preferences among hospital ambulatory care patients in Malaysia. Additionally, the study investigates the factors influencing patients' acceptance of e-labeling. METHODS: A cross-sectional study using a 28-item questionnaire was conducted at the outpatient pharmacy department of a quaternary hospital in Kuala Lumpur, Malaysia, from May to June 2023. The questionnaire was developed based on a review of published literature related to e-labeling and was guided by the Unified Theory of Acceptance and Use of Technology, second version (UTAUT2). Patients aged 18 years and above were recruited using a stratified sampling method to ensure representative age-related medication usage. A mobile tablet was provided to patients for self-completion of the e-survey in their preferred language (English, Malay, or Mandarin). Categorical data on e-labeling acceptance, perceptions, and preferences were analyzed using descriptive statistics. Qualitative content analysis was performed to characterize participants' responses to open-ended questions. Univariate and multivariate binomial logistic regression analyses were conducted to identify predictors of e-labeling acceptance. RESULTS: Out of 462 patients approached, 387 (83.8%) participated in the survey, with 283 (73.1%) accepting e-labeling. Most participants perceived the electronic version of the package insert as beneficial, particularly for understanding their medication better through the choice of language (352/387, 91.0%). However, around half of the participants (197/387, 50.9%) expressed concerns about the potential risks of obtaining illegal medication information via e-labeling. Most participants (302/387, 78.0%) preferred to access electronic leaflets through government websites. However, 221/387 (57.1%) still wanted the option to request printed leaflets. Significant predictors of e-labeling acceptance included perceived benefits such as better understanding of medication (adjusted odds ratio [AOR] 8.02, 95% CI 2.80-22.97, P<.001), environmental protection (AOR 7.24, 95% CI 3.00-17.51, P<.001), and flexibility in information retrieval (AOR 2.66, 95% CI 1.11-6.35, P=.03). Conversely, being of Chinese ethnicity compared with Malay (AOR 0.28, 95% CI 0.13-0.60, P=.005) and perceived lack of self-efficacy in browsing electronic leaflets (AOR 0.25, 95% CI 0.11-0.56, P<.001) were associated with lower acceptance. CONCLUSIONS: The acceptance rate for e-labeling among hospital ambulatory care patients was moderately high and was significantly influenced by ethnicity as well as patients' perceived benefits and challenges related to its implementation. Future strategies to enhance e-labeling uptake should address patient concerns regarding the challenges of using the digital platform and emphasize the benefits of e-labeling.

First-line treatment with ceftriaxone for Neisseria gonorrhoeae infection less likely to be prescribed to patients with a penicillin allergy label: a retrospective audit of medical records.

Background Gonorrhoea notifications have increased substantially in Australia over the past decade. Neisseria gonorrhoeae is already highly resistant to several antibiotics and so, alternatives to first-line treatment are generally strongly discouraged. The penicillin allergy label (AL) on patient medical records has previously been shown to influence prescribing practices, to the detriment of best-practice management and antimicrobial stewardship. This study aimed to understand how the penicillin AL influences antibiotic selection for gonorrhoea treatment at Canberra Sexual Health Centre. Methods A retrospective chart audit of gonorrhoea cases treated at Canberra Sexual Health Centre between January 2020 and October 2023 (n =619 patients, n =728 cases). Antibiotic selection was assessed according to penicillin AL status. Ceftriaxone selection was assessed according to penicillin allergy severity reported in the medical records and as determined using a validated antibiotic allergy assessment tool. Results Cases with a penicillin AL were more likely to receive antibiotics other than ceftriaxone (n =7/41, 17.1%) than cases without the label (n =8/687, 1.2%, P n =28/41, 68.3%) to apply the assessment tool. Those reported as low-severity in the records were more likely to receive ceftriaxone (n =21/22, 95.5%) than those reported as moderate-high (n =7/11, 63.6%) or unreported (n =6/8, 0.75%). Conclusions Treatment of gonorrhoea in outpatient settings requires an understanding of penicillin allergy, and the ability to quickly and accurately identify penicillin-AL patients who can safely tolerate ceftriaxone. Institutionally endorsed penicillin allergy de-labelling protocols and access to easy-to-navigate prescribing advice within national sexually transmitted infection management guidelines would support this.

FDA Initiative for Drug Facts Label for Over-the-Counter Naloxone.

BACKGROUND: The opioid crisis highlights the need to increase access to naloxone, possibly through regulatory approval for over-the-counter sales. To address industry-perceived barriers to such access, the Food and Drug Administration (FDA) developed a model drug facts label for such sales to assess whether consumers understood the key statements for safe and effective use. METHODS: In this label-comprehension study, we conducted individual structured interviews with 710 adults and adolescents, including 430 adults who use opioids and their family and friends. Eight primary end points were developed to assess user comprehension of each of the key steps in the label. Each of these end points included a prespecified target threshold ranging from 80 to 90% that was evaluated through a comparison of the lower boundary of the 95% exact confidence interval. RESULTS: The results for performance on six primary end points met or exceeded thresholds, including the steps "Check for a suspected overdose" (threshold, 85%; point estimate [PE], 95.8%; 95% confidence interval [CI], 94.0 to 97.1) and "Give the first dose" (threshold, 85%; PE, 98.2%; 95% CI, 96.9 to 99.0). The lower boundaries for four other primary end points ranged from 88.8 to 94.0%. One exception was comprehension of "Call 911 immediately," but this instruction closely approximated the target of 90% (PE, 90.3%; 95% CI, 87.9 to 92.4). Another exception was comprehension of the composite step of "Check, give, and call 911 immediately" (threshold, 85%; PE, 81.1%; 95% CI, 78.0 to 83.9). CONCLUSIONS: Consumers met thresholds for sufficient understanding of six of eight components of the instructions in the drug facts label for naloxone use and came close on two others. Overall, the FDA found that the model label was adequate for use in the development of a naloxone product intended for over-the-counter sales.

Assessment of clinically actionable pharmacogenetic markers to stratify anti-seizure medications.

Epilepsy treatment is challenging due to heterogeneous syndromes, different seizure types and higher inter-individual variability. Identification of genetic variants predicting drug efficacy, tolerability and risk of adverse-effects for anti-seizure medications (ASMs) is essential. Here, we assessed the clinical actionability of known genetic variants, based on their functional and clinical significance and estimated their diagnostic predictability. We performed a systematic PubMed search to identify articles with pharmacogenomic (PGx) information for forty known ASMs. Functional annotation of the identified genetic variants was performed using different in silico tools, and their clinical significance was assessed using the American College of Medical Genetics (ACMG) guidelines for variant pathogenicity, level of evidence (LOE) from PharmGKB and the United States-Food and drug administration (US- FDA) drug labelling with PGx information. Diagnostic predictability of the replicated genetic variants was evaluated by calculating their accuracy. A total of 270 articles were retrieved with PGx evidence associated with 19 ASMs including 178 variants across 93 genes, classifying 26 genetic variants as benign/ likely benign, fourteen as drug response markers and three as risk factors for drug response. Only seventeen of these were replicated, with accuracy (up to 95%) in predicting PGx outcomes specific to six ASMs. Eight out of seventeen variants have FDA-approved PGx drug labelling for clinical implementation. Therefore, the remaining nine variants promise for potential clinical actionability and can be improvised with additional experimental evidence for clinical utility.

Availability of Safe and Effective Therapeutic Options to Pregnant and Lactating Individuals Following the United States Food and Drug Administration Pregnancy and Lactation Labeling Rule.

OBJECTIVE: To explore the extent and type of pregnancy and lactation data of newly approved prescription drugs and assess whether the presented recommendations are data-driven, as required by the US Food and Drug Administration Pregnancy and Lactation Labeling Rule implemented in 2015. STUDY DESIGN: In this descriptive analysis, we reviewed pregnancy and lactation data of all new molecular entities approved between 2001 and 2020 in their most updated labeling. Information was collected regarding the pregnancy and lactation risk statements, the source of pregnancy and lactation data, and the design and methods of pregnancy and lactation studies in the labeling. RESULTS: Of the 422 new molecular entities, the key advisory statement for use of 133 (32%) drugs in pregnancy and 194 (46%) drugs in lactation were classified as "against use." Less than 2% of all drugs had a key advisory statement that supported their use during pregnancy or lactation. The sources of data regarding use in pregnancy were studies in human and animals in 46 (11%) and 348 (82%) drugs, respectively. For use during lactation, data included studies in human and animals in 23 (5%) and 251 (59%) drugs, respectively. The key advisory recommendation was consistent with the available human information in 4 (8%) drugs in pregnancy and 3 (13%) drugs in lactation. Prescription drug labeling contains limited data to support informed decision-making for the use of prescription drugs during pregnancy/lactation. Close collaboration among stakeholders is required to enhance the availability of data in this population.

Classifying Free Texts Into Predefined Sections Using AI in Regulatory Documents: A Case Study with Drug Labeling Documents.

The US Food and Drug Administration (FDA) regulatory process often involves several reviewers who focus on sets of information related to their respective areas of review. Accordingly, manufacturers that provide submission packages to regulatory agencies are instructed to organize the contents using a structure that enables the information to be easily allocated, retrieved, and reviewed. However, this practice is not always followed correctly; as such, some documents are not well structured, with similar information spreading across different sections, hindering the efficient access and review of all of the relevant data as a whole. To improve this common situation, we evaluated an artificial intelligence (AI)-based natural language processing (NLP) methodology, called Bidirectional Encoder Representations from Transformers (BERT), to automatically classify free-text information into standardized sections, supporting a holistic review of drug safety and efficacy. Specifically, FDA labeling documents were used in this study as a proof of concept, where the labeling section structure defined by the Physician Label Rule (PLR) was used to classify labels in the development of the model. The model was subsequently evaluated on texts from both well-structured labeling documents (i.e., PLR-based labeling) and less- or differently structured documents (i.e., non-PLR and Summary of Product Characteristic [SmPC] labeling.) In the training process, the model yielded 96% and 88% accuracy for binary and multiclass tasks, respectively. The testing accuracies observed for the PLR, non-PLR, and SmPC testing data sets for the binary model were 95%, 88%, and 88%, and for the multiclass model were 82%, 73%, and 68%, respectively. Our study demonstrated that automatically classifying free texts into standardized sections with AI language models could be an advanced regulatory science approach for supporting the review process by effectively processing unformatted documents.

Inconsistencies of absolute drug-drug contraindication reports: Analysis of Summaries of Product Characteristics of commonly prescribed drugs.

AIMS: Prescribing information should follow a defined structure to help prescribers easily find required information. Often information appears in different sections of Summaries of Product Characteristics (SmPCs) in an inconsistent way. Still unknown is how this inconsistency affects absolute contraindications and how it can be improved. Thisstudy aimed to evaluate the structure of absolute contraindications in SmPCs based on absolute drug-drug contraindications (DDCI) in the section 'contraindications' and references to sections 'special warnings and precautions for use' (here as 'warnings') and 'interaction with other medicinal products and other forms of interaction' (here as 'interactions'). METHODS: SmPCs of 693 commonly prescribed drugs were analysed regarding absolute DDCI in 'contraindications' sections. References to sections on 'warnings' and 'interactions' were evaluated to characterize information provided about DDCI. RESULTS: Of 693 analysed SmPCs, 138 (19.9%) contained ≥1 absolute DDCI. Of 178 SmPCs that referred to sections on 'warnings' or 'interactions', 131 (73.6%) did not contain further information on absolute DDCI, whereas 47 (26.4%) did. Such additional information was found in sections on 'interactions' and 'warnings' in 41 (87.2%) and 9 (19.1%) SmPCs, respectively. CONCLUSIONS: Information regarding absolute DDCI was found not only in sections on 'contraindications' but also in sections on 'warnings' and 'interactions'. Information was not given with consistently straightforward phrasing and structure and so can leave uncertainty for prescribers. To improve drug safety, clear definitions and wording for absolute and relative contraindications should be provided, ideally in tables.