Algorithmic Identification of Treatment-Emergent Adverse Events From Clinical Notes Using Large Language Models: A Pilot Study in Inflammatory Bowel Disease.

Outpatient clinical notes are a rich source of information regarding drug safety. However, data in these notes are currently underutilized for pharmacovigilance due to methodological limitations in text mining. Large language models (LLMs) like Bidirectional Encoder Representations from Transformers (BERT) have shown progress in a range of natural language processing tasks but have not yet been evaluated on adverse event (AE) detection. We adapted a new clinical LLM, University of California - San Francisco (UCSF)-BERT, to identify serious AEs (SAEs) occurring after treatment with a non-steroid immunosuppressant for inflammatory bowel disease (IBD). We compared this model to other language models that have previously been applied to AE detection. We annotated 928 outpatient IBD notes corresponding to 928 individual patients with IBD for all SAE-associated hospitalizations occurring after treatment with a non-steroid immunosuppressant. These notes contained 703 SAEs in total, the most common of which was failure of intended efficacy. Out of eight candidate models, UCSF-BERT achieved the highest numerical performance on identifying drug-SAE pairs from this corpus (accuracy 88-92%, macro F1 61-68%), with 5-10% greater accuracy than previously published models. UCSF-BERT was significantly superior at identifying hospitalization events emergent to medication use (P < 0.01). LLMs like UCSF-BERT achieve numerically superior accuracy on the challenging task of SAE detection from clinical notes compared with prior methods. Future work is needed to adapt this methodology to improve model performance and evaluation using multicenter data and newer architectures like Generative pre-trained transformer (GPT). Our findings support the potential value of using large language models to enhance pharmacovigilance.

Miniaturization of an enclosed electrospinning process to enhance reproducibility in the fabrication of rapidly dissolving cell-based biosensors.

There is broad interest in producing electrospun films embedded with biological materials. It is well known that electrospinning requires careful control of the process conditions, especially the environmental conditions such as relative humidity (RH). Given that commercial electrospinning systems are expensive (> $10,000) and are typically too large to be used in standard biological safety cabinets (BSC), we designed and built a miniaturized electrospinning box (E-Box) that will fit inside a BSC, and the RH can be easily controlled using simple instrumentation (gas cylinder, regulator, needle valve, rotameter). It uses an inexpensive computerized numerical control machine to control the spinneret positioning and collector rotational speed-all the parts for the device (except the syringe pump and voltage supply) can be purchased for approximately $1000. We demonstrate the usefulness of our design in optimizing the production of Escherichia coli-embedded pullulan-trehalose films to be used as rapidly dissolving biosensors for environmental monitoring. At a fixed electrospinning recipe, we showed that decreasing the RH from approximately 48% to 22% resulted in the average fiber diameter increasing from 240 (± 11) nm to 314 (± 8) nm. We also demonstrate the usefulness of our design in performing sequential electrospinning experiments to evaluate process performance reproducibility. For example, from just 1 mL of a polymer solution, we produced 16 electrospun films (approximately 3 cm by 8 cm each)-from those films we hole-punched approximately 80 biosensor discs which were then used in subsequent experiments to determine the amount of two different biocides (Grotan BK and triclosan) in aqueous samples. The technique developed in this study is ideal for creating electrospun materials in high quantities that are highly reproducible through the precise control of RH.

A descriptive analysis of postmarketing requirement studies and clinical trials.

PURPOSE: Under the Food and Drug Administration Amendments Act of 2007 (FDAAA), the FDA has the authority to require applicants to conduct postmarketing studies or clinical trials. These postmarketing requirements (PMRs) provide additional data on the safety of the drug product. The purpose of the study was to conduct a descriptive analysis of FDAAA PMRs and the resulting regulatory actions. METHODS: This study evaluated FDAAA PMRs established between 2013 and 2019. We used the Medical Dictionary for Regulatory Activities (MedDRA) to map preferred terms (PTs) for serious risks associated with the PMRs. Relevant documents available in the FDA's document archiving, reporting, and regulatory tracking system (DARRTS), including but not limited to internal letters and reviews, documents submitted by applicants, and publicly available data sources were assessed for data collection of study elements. RESULTS: Of the 1079 new FDAAA PMRs established between January 01, 2013, and December 31, 2019, 82% (n = 884) were associated with new drug applications (NDAs) and 18% (n = 195) with biologic license applications (BLAs). Most PMRs were established at the time of drug approval (73%, n = 789) compared to post-approval (27%, n = 290). The majority of PMRs had an open status (59%, n = 639) and 41% (n = 440) were closed. The median time from the PMR establishment date to submission of the results to the FDA was 690 days (interquartile range [IQR]: 748 days) for 167 completed clinical trials and 483 days (IQR: 603 days) for 241 completed studies. Approximately 53% (180/339) of fulfilled FDAAA PMRs resulted in labeling changes. CONCLUSIONS: FDAAA PMRs are useful in informing postmarket safety of drugs. Most FDAAA PMRs were established at the time of drug approval, reflecting safety signals identified during the review of the marketing application, and over half of fulfilled FDAAA PMRs resulted in regulatory action.

Algorithmic identification of treatment-emergent adverse events from clinical notes using large language models: a pilot study in inflammatory bowel disease.

Background and Aims: Outpatient clinical notes are a rich source of information regarding drug safety. However, data in these notes are currently underutilized for pharmacovigilance due to methodological limitations in text mining. Large language models (LLM) like BERT have shown progress in a range of natural language processing tasks but have not yet been evaluated on adverse event detection. Methods: We adapted a new clinical LLM, UCSF BERT, to identify serious adverse events (SAEs) occurring after treatment with a non-steroid immunosuppressant for inflammatory bowel disease (IBD). We compared this model to other language models that have previously been applied to AE detection. Results: We annotated 928 outpatient IBD notes corresponding to 928 individual IBD patients for all SAE-associated hospitalizations occurring after treatment with a non-steroid immunosuppressant. These notes contained 703 SAEs in total, the most common of which was failure of intended efficacy. Out of 8 candidate models, UCSF BERT achieved the highest numerical performance on identifying drug-SAE pairs from this corpus (accuracy 88-92%, macro F1 61-68%), with 5-10% greater accuracy than previously published models. UCSF BERT was significantly superior at identifying hospitalization events emergent to medication use (p < 0.01). Conclusions: LLMs like UCSF BERT achieve numerically superior accuracy on the challenging task of SAE detection from clinical notes compared to prior methods. Future work is needed to adapt this methodology to improve model performance and evaluation using multi-center data and newer architectures like GPT. Our findings support the potential value of using large language models to enhance pharmacovigilance.

Radiofrequency therapy improves exercise capacity of mice with emphysema.

Emphysema is a common phenotype of chronic obstructive pulmonary disease (COPD). Although resection of emphysematous tissue can improve lung mechanics, it is invasive and fraught with adverse effects. Meanwhile, radiofrequency (RF) treatment is an extracorporeal method that leads to tissue destruction and remodeling, resulting in "volume reduction" and overall improvement in lung compliance of emphysematous lungs. Whether these changes lead to improved exercise tolerance is unknown. Here, we investigated the effectiveness of RF treatment to improve the exercise capacity of mice with emphysema. Fifty-two mice (7 weeks of age) were used in this experiment. A bilateral emphysema model was created by intratracheally instilling porcine pancreatic elastase (PPE) (1.5U/100 g body weight). RF treatment (0.5 W/ g body weight) was administered extracorporeally 14 days later and mice were sacrificed after another 21 days. The exercise capacity of mice was measured using a treadmill. Treadmill runs were performed just before PPE instillation (baseline), before RF treatment and before sacrifice. Following sacrifice, lung compliance and mean linear intercept (Lm) were measured and fibrosis was assessed using a modified Ashcroft score. There were 3 experimental groups: controls (instilled with saline, n = 12), emphysema (instilled with porcine pancreatic elastase, PPE, n = 11) and emphysema + treatment (instilled with PPE and given RF, n = 9). At endpoint, the maximum velocity of the emphysema + treatment group was significantly higher than that of the emphysema group, indicating improved exercise tolerance (86.29% of baseline vs 61.69% of baseline, p = 0.01). Histological analysis revealed a significant reduction in emphysema as denoted by Lm between the two groups (median 29.60 µm vs 35.68 µm, p = 0.03). The emphysema + treatment group also demonstrated a higher prevalence of lung fibrosis (≧Grade 3) compared with the emphysema group (11.7% vs 5.4%, p < 0.01). No severe adverse events from RF were observed. RF treatment improved the exercise capacity of mice with emphysema. These data highlight the therapeutic potential of RF treatment in improving the functional status of patients with COPD.

COPD and the risk of poor outcomes in COVID-19: A systematic review and meta-analysis.

BACKGROUND: Patients with chronic obstructive pulmonary disease (COPD) are highly susceptible from respiratory exacerbations from viral respiratory tract infections. However, it is unclear whether they are at increased risk of COVID-19 pneumonia or COVID-19-related mortality. We aimed to determine whether COPD is a risk factor for adverse COVID-19 outcomes including hospitalization, severe COVID-19, or death. METHODS: Following the PRISMA guidelines, we performed a systematic review of COVID-19 clinical studies published between November 1st, 2019 and January 28th, 2021 (PROSPERO ID: CRD42020191491). We included studies that quantified the number of COPD patients, and reported at least one of the following outcomes stratified by COPD status: hospitalization; severe COVID-19; ICU admission; mechanical ventilation; acute respiratory distress syndrome; or mortality. We meta-analyzed the results of individual studies to determine the odds ratio (OR) of these outcomes in patients with COPD compared to those without COPD. FINDINGS: Fifty-nine studies met the inclusion criteria, and underwent data extraction. Most studies were retrospective cohort studies/case series of hospitalized patients. Only four studies examined the effects of COPD on COVID-19 outcomes as their primary endpoint. In aggregate, COPD was associated with increased odds of hospitalization (OR 4.23, 95% confidence interval [CI] 3.65-4.90), ICU admission (OR 1.35, 95% CI 1.02-1.78), and mortality (OR 2.47, 95% CI 2.18-2.79). INTERPRETATION: Having a clinical diagnosis of COPD significantly increases the odds of poor clinical outcomes in patients with COVID-19. COPD patients should thus be considered a high-risk group, and targeted for preventative measures and aggressive treatment for COVID-19 including vaccination.

Beta-blockers in chronic obstructive pulmonary disease: the good, the bad and the ugly.

PURPOSE OF REVIEW: Several observational studies have suggested that ß-blockers, especially cardioselective ones, are well tolerated and associated with a lower risk of acute exacerbations and death in patients with chronic obstructive pulmonary disease (COPD). However, there are dissenting studies. This review provides an update on the use of ß-blockers in COPD, focusing on results of recent prospective studies and randomized controlled trials. RECENT FINDINGS: In totality, cohort studies indicate that ß-blockers are generally well tolerated and effective in COPD patients who also have a clear cardiovascular indication for these medications. Although ß-blockers on average reduce lung function acutely in COPD patients, the absolute decrease is relatively small, especially if cardioselective ß-blockers are used. The results of two large randomized controlled trials suggest that ß-blocker use does not reduce the therapeutic benefits of inhaled bronchodilators in COPD patients. The use of ß-blockers in COPD patients, who do not have overt cardiovascular disease, does not prevent COPD exacerbations and may paradoxically increase the risk of COPD-related hospitalization and mortality. SUMMARY: The use of ß-blockers is generally well tolerated and effective in COPD patients, who also have a clear cardiovascular indication for these drugs. However, they should not be used in patients who do not have overt cardiovascular disease as ß-blockers can reduce lung function, worsen health status and increase the risk of COPD-related hospitalization.

Generic and Brand-Name Thyroid Hormone Drug Use Among Commercially Insured and Medicare Beneficiaries, 2007 Through 2016.

CONTEXT: Generic drugs account for 9 out of 10 prescriptions dispensed in the United States but for a lower proportion of commonly prescribed thyroid hormone replacement therapies. OBJECTIVE: Characterize temporal patterns of generic and brand-name thyroid hormone drug use, including patient and prescriber characteristics associated with brand-name use. DESIGN AND SETTING: Cross-sectional longitudinal analysis of national data from a large administrative claims database from January 2007 through December 2016. PATIENTS: Adults with insurance coverage through commercial, Medicare Advantage, and Medicare Part D health plans. MAIN OUTCOME MEASURES: Generic and brand-name thyroid hormone drug use. RESULTS: From 2007 to 2016, the annual number of thyroid hormone treatment pharmacy fills increased from 8,905,836 in 2007 to 11,613,923 in 2016, 73.6% of which were for generic levothyroxine, 23.4% for brand-name levothyroxine, and the remaining for other formulations. Dispensing of generic thyroid hormone drugs increased from 59.8% in 2007 to 84.9% in 2016 and was consistently higher among Medicare Advantage and Medicare Part D when compared with the commercial beneficiary population. For all three beneficiary populations, use of brand-name products was less common among older adults and more common among women and those receiving prescriptions from endocrinologists and was more common among those of white race and with greater household income for the Medicare Advantage and commercial beneficiary populations (P < 0.001). CONCLUSIONS: Brand-name thyroid hormone product use declined from 2007 to 2016 among three large, national insurer beneficiary populations. Although certain patient characteristics were associated with brand-name use, prescriber specialty was the strongest predictor.

Histomorphological and Molecular Assessments of the Fixation Times Comparing Formalin and Ethanol-Based Fixatives.

The lack of standardization of tissue handling and processing hinders the development and validation of new biomarkers in research and clinical settings. We compared the histomorphology and the quality and quantity of biomolecules in paraffin-embedded mouse tissues, followed by fixation with neutral buffered formalin (NBF), 70% ethanol, and buffered ethanol (BE70) fixative. The quality of the histomorphology and immunohistochemistry in BE70 was relatively time-independent, whereas those in NBF rapidly decreased after 1 week of fixation. Protein recovered from tissue fixed in 70% ethanol and BE70 was compatible with Western blot and protein array using AKT and GAPDH antibodies, regardless of the fixation time. In addition, the quality and quantity of RNA extracted from tissue in ethanol-based fixative showed minimal changes from 4 hr to 6 months, whereas NBF had a dramatic detrimental change in RNA quality after 1 week of fixation. Furthermore, ethanol-based fixative offers a superior DNA template for PCR amplification-based molecular assays than NBF. In conclusion, coagulative, ethanol-based fixatives show a broader time spectrum than the aldehyde crosslinking fixative NBF in their histomorphological features and the quantity and quality of the biomolecules from paraffin-embedded tissue, and they may facilitate the use of fixative-fixed paraffin-embedded tissues in research and clinical laboratories, avoiding overfixation.

Tumor necrosis factor alpha blockers and malignancy in children: forty-eight cases reported to the Food and Drug Administration.

OBJECTIVE: Malignancies reported in children using tumor necrosis factor alpha (TNFalpha) blockers have raised concerns of a potential increased risk. This study was undertaken to investigate postmarketing reports of malignancy in children treated with TNF blockers. METHODS: The FDA's Adverse Event Reporting System was searched to identify malignancies associated with the use of infliximab, etanercept, and adalimumab in children in whom therapy was initiated between the ages of 0 and 18 years. The reporting rates for infliximab and etanercept were compared with the background rate of malignancy in the general pediatric population. RESULTS: Forty-eight reports of malignancy in children were identified: 31 following infliximab use, 15 following etanercept use, and 2 following adalimumab use. Half of the malignancies reported were lymphomas and included both Hodgkin's and non-Hodgkin's lymphoma. The remaining reported cases involved a variety of different malignancies including leukemia, melanoma, and solid organ cancers. The majority of the reported cases (88%) involved the concomitant use of other immunosuppressants. Reporting rates for malignancy showed that infliximab had a consistently higher reporting rate when compared with background rates in the general pediatric population for lymphomas and all malignancies. The reporting rates for etanercept were elevated above background for lymphomas and were on par with background for all malignancies. CONCLUSION: There is evidence that treatment with TNF blockers in children may increase the risk of malignancy. However, the cases were confounded by the potential risk of malignancy associated with underlying illnesses and the use of concomitant immunosuppressants; therefore, a clear causal relationship could not be established.

Chloroplast His-to-Asp signal transduction: a potential mechanism for plastid gene regulation in Heterosigma akashiwo (Raphidophyceae).

BACKGROUND: Maintenance of homeostasis requires that an organism perceive selected physical and chemical signals within an informationally dense environment. Functionally, an organism uses a variety of signal transduction arrays to amplify and convert these perceived signals into appropriate gene transcriptional responses. These changes in gene expression serve to modify selective metabolic processes and thus optimize reproductive success. Here we analyze a chloroplast-encoded His-to-Asp signal transduction circuit in the stramenopile Heterosigma akashiwo (Hada) Hada ex Y. Hara et Chihara [syn. H. carterae (Hulburt) F.J.R. Taylor]. The presence, structure and putative function of this protein pair are discussed in the context of their evolutionary homologues. RESULTS: Bioinformatic analysis of the Heterosigma akashiwo chloroplast genome sequence revealed the presence of a single two-component His-to-Asp (designated Tsg1/Trg1) pair in this stramenopile (golden-brown alga). These data represent the first documentation of a His-to-Asp array in stramenopiles and counter previous reports suggesting that such regulatory proteins are lacking in this taxonomic cluster. Comparison of the 43 kDa H. akashiwo Tsg1 with bacterial sensor kinases showed that the algal protein exhibits a moderately maintained PAS motif in the sensor kinase domain as well as highly conserved H, N, G1 and F motifs within the histidine kinase ATP binding site. Molecular modelling of the 27 kDa H. akashiwo Trg1 regulator protein was consistent with a winged helix-turn-helix identity - a class of proteins that is known to impact gene expression at the level of transcription. The occurrence of Trg1 protein in actively growing H. akashiwo cells was verified by Western analysis. The presence of a PhoB-like RNA polymerase loop in Trg1 and its homologues in the red-algal lineage support the hypothesis that Trg1 and its homologues interact with a sigma 70 (sigma70) subunit (encoded by rpoD) of a eubacterial type polymerase. Sequence analysis of H. akashiwo rpoD showed this nuclear-encoded gene has a well-defined 4.2 domain, a region that augments RNA polymerase interaction with transcriptional regulatory proteins and also serves in -35 promoter recognition. The presence/loss of the His-to-Asp pairs in primary and secondary chloroplast lineages is assessed. CONCLUSION: His-to-Asp signal transduction components are found in most rhodophytic chloroplasts, as well as in their putative cyanobacterial progenitors. The evolutionary conservation of these proteins argues that they are important for the maintenance of chloroplast homeostasis. Our data suggest that chloroplast gene transcription may be impacted by the interaction of the His-to-Asp regulator protein (which is less frequently lost than the sensor protein) with the RNA polymerase sigma70 subunit.