Pharmacotherapy最新文献

Introduction: Self-care and self-medication are increasingly viewed as helpful approaches to managing minor ailments; however, patients are often not confident in making informed choices. Pharmacists have traditionally assisted patients in this domain, but the emergence of digital health technologies has transformed the way individuals seek information towards the use of artificial intelligence (AI) tools. ChatGPT-4o mini, Gemini, and Copilot are recently growing popular for health-related guidance. Despite the accessibility and ease of use that these AI tools offer, their accuracy, patient-centeredness, and reliability in supporting self-care remain insufficiently evaluated.

Aims and objectives: The primary objective of this study is to evaluate and compare the performance of ChatGPT-4o mini, Gemini, and Copilot in the context of patient self-care by assessing the accuracy, patient-centeredness, and comprehensiveness of their responses against standard recommendations.

Materials and methods: Ninety-one case scenarios representing the most common minor ailments were introduced to the three AI models to generate responses that were subsequently assessed and compared with established standard recommendations by three of the study investigators. Evaluation of the responses was conducted on their accuracy, patient-centeredness, comprehensiveness, and similarity. An inter-reliability test was also carried out to confirm the consistency between the three evaluators' assessments.

Results: The study findings indicate that ChatGPT-4o mini significantly exceeded Gemini and Copilot in terms of accuracy and presented as mean ± SD (ChatGPT-4o mini: 4.4 ± 0.6, Gemini: 4.1 ± 0.8, Copilot: 3.7 ± 0.7, p < 0.001), patient-centeredness (ChatGPT-4o mini: 4.7 ± 0.6, Gemini: 4.3 ± 1.0, Copilot: 4.2 ± 0.8, p < 0.001), and comprehensiveness (ChatGPT-4o mini: 4.6 ± 0.7, Gemini: 4.2 ± 0.8, Copilot: 3.4 ± 0.7; p < 0.001) among 91 minor ailment case scenarios. Gemini and Copilot showed moderate and low performance, respectively, particularly in complex cases, in contrast to ChatGPT-4o mini. Inter-rater reliability was excellent (Cronbach's alpha ≥ 0.9), confirming assessment consistency. Cosine similarity analysis indicated high overlap between AI and standard recommendations.

Conclusion: This study shows that AI tools are reliable and precise instruments for self-care of mild diseases. These findings highlight ChatGPT-4o mini's superior reliability and patient-centeredness for self-medication guidance, while underscoring the need for human oversight. However, there is a small chance of variation and errors in the AI-generated responses, which may prohibit complete dependence on AI for self-care recommendations.

Introduction: Prediction algorithms for prolonged mechanical ventilation (PMV) in the intensive care unit (ICU) have rarely incorporated detailed medication data, despite medications being important causal contributors to patient outcomes. The purpose of this study was to develop and validate PMV prediction models to assess the contribution of medication-related variables alongside established physiologic predictors.

Methods: In this retrospective cohort study, models were developed using data from a random sample of 318 adults admitted to ICUs within the University of North Carolina (UNC) health system who received mechanical ventilation for ≥ 24 h from October 2015 to October 2020. Validation was performed in two datasets: a temporally distinct cohort from UNC from June 2021 to June 2023, and a cohort from Oregon Health Sciences University from June 2020 to June 2023. Logistic regression and supervised, classification-based machine learning (ML) models [XGBoost, Random Forest, Support Vector Machine (SVM)] were trained on 30 demographic, clinical, laboratory, and medication-related variables. The primary outcome was area under the receiver operating characteristic (AUROC) of developed prediction models for the occurrence of PMV.

Results: The base logistic regression model with medication regimen complexity and severity of illness data added was the best-performing regression model, achieving an AUROC of 0.75. Random Forest and SVM ML models achieved AUROCs of 0.78. Model discrimination decreased modestly in external validation. Explainability analyses of ML models expectedly included severity of illness scores and respiratory indices among the most important features, but also consistently included the medication regimen complexity-intensive care unit (MRC-ICU) score and other medication metrics. Incorporation of medication data yielded modest improvements in overall discrimination and negative predictive value.

Conclusions: Medication-related variables contributed incremental value to PMV prediction. ML methods provided marginal improvements over regression models. These findings highlight the potential value of medication data in prediction modeling for patient outcomes but emphasize the need to contextualize the value of complex models over simpler alternatives.

Background: Clozapine initiation often triggers inflammatory responses that can alter metabolism via Cytochrome P450 1A2 (CYP1A2) suppression. Although C-reactive protein (CRP) is the recommended marker, it may be unavailable in community settings. Neutrophil-to-lymphocyte ratio (NLR), routinely measured, could serve as a surrogate, though its value in detecting clozapine-related inflammation and metabolic changes remains unclear.

Aims: This study aimed to assess the relationship between CRP and NLR in individuals treated with clozapine, evaluate whether NLR can act as a proxy for elevated CRP (> 5 mg/L), and determine whether NLR, like CRP, explains variability in clozapine metabolism (concentration to dose (C/D) ratios) after adjusting for covariates.

Methods: We performed a retrospective cohort study of clozapine-treated inpatients at the British Columbia Psychosis Program (2012-2021). Patients with clozapine levels and matched complete blood counts (CBCs) (±7 days) were included, with CRP added when available. Multivariate mixed models assessed associations between CRP, NLR, and clozapine C/D ratios, while receiver operating characteristic (ROC) analyses evaluated NLR as a proxy for elevated CRP.

Results: Among 150 patients, 760 clozapine serum/CBC pairs and 212 CRP measurements met eligibility criteria. NLR was modestly associated with CRP (estimate = 0.027, p < 0.001). ROC analysis indicated that NLR had limited predictive utility, with an area under the curve (AUC) of 0.640 for detecting CRP > 5 mg/L. Subsequent analyses for higher CRP thresholds (> 10 and > 20 mg/L) produced comparable NLR AUC values of 0.621 and 0.669, respectively. Neutrophil count alone demonstrated marginally better performance but remained similarly limited in predictive value. In multivariate models, CRP but not NLR, was independently associated with clozapine C/D ratios.

Conclusion: Our findings indicate that although NLR and other hematological indices are easily accessible and may provide some indication of inflammation, they cannot substitute for CRP in guiding clozapine titration decisions. Where CRP is unavailable, NLR > 3 may be cautiously informative, though CRP remains the preferred marker for early detection and dose adjustment to optimize tolerability, adherence, and safety during clozapine initiation.

The treatment of migraine is hampered by inter-individual variability, leading to an inefficient "trial and error" approach. Artificial intelligence (AI) and machine learning (ML) offer a path towards precision medicine by predicting therapeutic outcomes. This scoping review systematically evaluates the evidence for AI and ML models for predicting pharmacologic response in migraine. A systematic search of four databases (PubMed, Web of Knowledge, Cochrane Library, and OpenGrey) identified 12 eligible studies using AI/ML to predict acute or prophylactic response to migraine treatment. These studies, which date back to articles published in 2006 and have been increasingly published recently, used a wide range of methods, from classical algorithms like support vector machines to deep learning and probabilistic models. The models primarily utilized clinical phenotyping and neuroimaging data and reported high predictive accuracy for novel biologics (e.g., anti-calcitonin gene-related peptide monoclonal antibodies (CGRP mAbs)) and acute treatments (e.g., nonsteroidal anti-inflammatory drugs (NSAIDs)). However, our systematic review finds that this apparent success is undermined by critical and pervasive methodological weaknesses. The central finding is that most studies relied solely on internal validation, carrying a high risk of overfitting, with external validation being exceptionally rare. Furthermore, several publications were based on overlapping patient cohorts, and a complete lack of biomarker or genetic data was noted. Consequently, the clinical application of AI and ML is currently stalled. Future progress depends on overcoming the "crisis of generalizability" by mandating external validation, addressing the "data bottleneck" with large, diverse datasets, and expanding data modalities to include "omic" data. These measures are critical to begin to realize the potential of AI and ML to personalize migraine treatment and significantly improve patient outcomes.

Background: Gender-affirming testosterone therapy is one part of the standard of care for more than 1 million transgender adults in the United States. Testosterone therapy may influence the activities of drug-metabolizing enzymes and transporters, but knowledge about its effect on the pharmacokinetics of other medications is limited. We determined the effects of gender-affirming testosterone therapy on apparent cytochrome P450 (CYP) 3A and P-glycoprotein activities using midazolam and digoxin as model probe substrates among transgender adults.

Methods: This was a longitudinal (pre-treatment and with concomitant testosterone therapy), prospective, non-randomized, open-label, three-phase probe substrate study. Eligible participants started testosterone therapy based on clinical need. Participants received one oral dose of midazolam 2 mg and digoxin 0.25 mg (simultaneous dosing) under fasted conditions before starting gender-affirming testosterone therapy (baseline), and at 1-month and 3-months on gender-affirming testosterone therapy. Midazolam, 1'-hydroxymidazolam, 4-hydroxymidazolam, digoxin, and total testosterone concentrations were determined by liquid chromatography-tandem mass spectrometry assays. We estimated single-dose pharmacokinetic parameters of midazolam, its metabolites, and digoxin using standard noncompartmental methods. Pharmacokinetic parameters were compared with testosterone therapy at 1-month and 3-months to baseline as geometric mean ratios (90% confidence intervals) and paired t-tests after log transformation. A p < 0.025 was considered significant.

Results: Among 14 participants (mean age: 24 ± 3 years; weight: 82.9 ± 20.9 kg; race/ethnicity: 71% White, non-Hispanic, 14% Hispanic, 7% Asian, 7% mixed race), nine participants started weekly testosterone injections (20 mg to 80 mg once weekly) and five started daily transdermal testosterone applications (12.5 mg to 50 mg once daily gel or cream, 2 mg daily patch). Mean total testosterone concentrations at 3 months increased more than 20-fold from baseline concentrations (25 ± 7 ng/dL to 507 ± 263 ng/dL). Geometric mean midazolam and metabolite pharmacokinetic parameters and digoxin parameters were not significantly different at baseline and with testosterone therapy.

Conclusion: Gender-affirming testosterone therapy did not significantly affect CYP3A or P-glycoprotein activities. Gender-affirming testosterone therapy may have minimal effects on the pharmacokinetics of other medications that are substrates of CYP3A and P-glycoprotein. Caution may be warranted for medications with a narrow therapeutic index.

Background: Falls and related injuries (FRI) pose a large burden among older adults with depression. Proactively identifying individuals at high FRI risk enables timely and tailored interventions, reducing unnecessary health care resource utilization. However, prior prediction models relied on fixed time intervals and failed to capture dynamic changes in health status over time.

Objectives: To develop and validate machine-learning algorithms (i.e., elastic net, random forest, and gradient boosting machine) for predicting 3-month FRI risk among older adults with depression.

Methods: This prognostic modeling study included fee-for-service Medicare beneficiaries aged 65 years or older with a depression diagnosis in 2017. Beneficiaries were followed in 3-month episodes from the first depression diagnosis until the earliest of death, hospice services or nursing facility utilization, switching to Medicare Advantage plans, or the end of the study period (i.e., December 31, 2019). A total of 261 time-varying predictors, spanning patient-, provider-, health system- and region-related factors, were updated every 3 months to predict incident FRI risk in the subsequent 3 months. We assessed prediction performance using c-statistics and stratified patients into different risk subgroups using the best-performing model.

Results: Among 274,268 eligible beneficiaries, the mean age was 74.6 (standard deviation [SD] = 7.2) years, 32.0% were male, 85.2% were White, and 15.1% experienced at least one FRI event throughout the study period. Using the random forest model (c-statistics = 0.68), 68.9% of the actual FRI cases were captured in the top three deciles of predicted risk. Individuals in the bottom seven deciles had a minimal FRI incidence (< 1.7%). Key predictors included frailty, age, prior FRI history, and daily dose of antidepressants.

Conclusion: Using a nationally representative cohort and time-varying predictors, our model offers a practical approach for efficiently identifying older adults at high FRI risk, which can be updated over time. This approach can inform clinical decision-making and optimize the allocation of fall prevention resources.