We performed a retrospective cohort study to evaluate risk of antidepressant discontinuation as a function of CYP2C19 and CYP2D6 phenotype. Patients who participated in an elective genomic screening and had a diagnosis of depression or anxiety with a corresponding antidepressant were analyzed. This resulted in 5 808 patients comprising 8 571 antidepressant orders. Of the total antidepressant orders, 51% of the antidepressants ordered before pharmacogenetic testing were discontinued. A multivariate analysis - accounting for age, medication tenure, allergies, PHQ-9, malaise/fatigue, and medication indication - revealed a significant association between metabolizer status and discontinuation for CYP2C19 increased metabolizers (ultrarapid plus rapid metabolizers) compared to normal metabolizers (n = 4 635) (HR = 1.17 [1.08, 1.27], p < 0.001) for CPIC guideline-recommended medications (i.e., citalopram, escitalopram, sertraline). Stratifying risk by individual drug maintained significant associations for escitalopram (HR = 1.27 [1.1, 1.46], p < 0.05) and sertraline (HR = 1.15 [1.01, 1.31], p < 0.05). While the CYP2D6 decreased metabolizer group (intermediate plus poor metabolizers) did not reach statistical significance for guideline recommended medications (i.e., paroxetine, venlafaxine, vortioxetine), the individual effect for venlafaxine showed an increased risk of discontinuation (HR = 1.23 [1.01, 1.5], p < 0.05). These results suggest phenotypic variations may have variable impact on risk of discontinuation amongst different antidepressant medications, but for escitalopram, sertraline, and venlafaxine, the risk of discontinuation in particular phenotypes should be considered at the time of therapy initiation.

PURPOSE:The aim of this study was to investigate the influence of CYP2D6 functionality on metabolic abnormalities in patients prescribed aripiprazole. METHODS:A retrospective cohort review was conducted in patients who had CYP2D6 genotyping and an oral prescription for aripiprazole. Primary outcomes were changes in metabolic parameters from baseline based on CYP2D6 genotype-based phenotype. Patients concurrently taking medications that cause CYP2D6 inhibition were classified as phenoconverters and were categorized as poor metabolizers in subanalyses. Index and post-aripiprazole (after either initiation or dose modification) data were abstracted from the electronic medical record for the following endpoints: body mass index (BMI), blood glucose, glycated hemoglobin, low-density lipoprotein (LDL), triglycerides, and total cholesterol. RESULTS:In multivariate mixed-effects analysis of 1,562 patients, the genotype-based phenotype cohort had no significant differences. A marginal difference in triglyceride levels was observed between poor metabolizers and normal metabolizers (β = 23.42; range, -4.9 to 51.73; P = 0.11). Findings were similar between poor metabolizers and normal metabolizers when evaluating outcomes encompassing phenoconversion, although poor metabolizers had an increased mean change in LDL levels (β = 10.04; range, 1.46 to 18.61; P < 0.05) and had a decreased mean change in BMI (β = -0.34; range, -0.72 to 0.04; P = 0.08); however, these differences were not statistically significant. CONCLUSION:This study suggests that the incidence of metabolic abnormalities beyond LDL may not be higher in CYP2D6 poor metabolizers compared to normal metabolizers. Further research is needed to investigate clinically meaningful outcomes in relation to CYP2D6 and aripiprazole.

BACKGROUND:Integration of pharmacogenomics (PGx) into clinical care has primarily occurred within academic medical centers. Clinical decision support (CDS) is vital to the incorporation of PGx into clinical care, however, efforts have not capitalized on the integration within community pharmacy settings. OBJECTIVE:To describe the implementation efforts for community pharmacy CDS alerts and assess effectiveness. METHODS:A retrospective manual chart review was conducted to evaluate the effectiveness of deploying PGx alerts across 14 community pharmacies within a single health system. A report was created to capture all PGx CDS alerts generated during the eight-month study period (October 2024-May 2025). Alerts were categorized as actionable when documentation was sparse for medication adjustments based on PGx results from the prescribing clinician upon transmission of the prescription to the community pharmacy. RESULTS:A total of 74 alerts were generated, of which 36 were unique alerts. Thirty-eight alerts were excluded as duplicates when the pharmacist re-entered the dispensing function. Clinicians adjusted 17 chemotherapy prescriptions prior to order transmission. Additionally, five alerts for chemotherapy medications were considered not actionable as the patient had previously tolerated therapy. The remaining 13 alerts comprised of 11 for clopidogrel and 2 for tramadol. Pharmacists intervened in 8 of the 11 clopidogrel orders which resulted in 3 medication modifications. The overall alert acceptance was 23%. CONCLUSION:This study describes one of the first institutions to deploy PGx alerts through the community pharmacy module across a health system's electronic health record. PGx alerts in the medication dispensing module provide an additional safeguard for precision medicine, enabling pharmacists to make targeted interventions and provide patient education. Continuous evaluation of PGx alerts affords opportunities for alert refinement and further optimizing patient care.

This study aims to design and validate a large language model (LLM) framework for systematic extraction of medication dosing across multiple therapeutic classes using electronic health record (EHR) data. Manual dose annotations were completed for 4295 medications across nine therapeutic classes. Five publicly available LLMs (Mistral-small, Llama 3-70B, Nova lite, DeepSeek, and Claude 3.5 Sonnet) were tested through iterative prompt engineering. Discrepancies between manual and model-derived doses were assessed using R2 and classification accuracy. Among 2146 training and 2149 testing samples, Claude achieved high-performance metrics (R2 = 99.32%, accuracy = 92.36%). DeepSeek had R2 of 97.17% and accuracy of 90.31%. After prompt optimization, both models showed minor improvements (Claude R2 = 99.34%, accuracy = 92.78%; DeepSeek R2 = 98.89%, accuracy = 91.01%). Nova lite, Llama, and Mistral-small performance metrics improved after prompt re-engineering (ΔR2 = 9.23%, 12.65%, and 39.03%, respectively). Models had modest agreement with an intra-class correlation of 0.73 (CI = 0.72-0.74). In testing data, Claude and DeepSeek demonstrated high-performance metrics R2 = 99.74%, accuracy = 93.45% and R2 = 95.58%, accuracy = 91.73%, respectively. Performance improvements were observed across all models following prompt refinement; however, differences may be attributed to model-specific responses to the prompting strategy used in this study. LLMs show high accuracy for automated medication dose extraction from EHR data. This study can serve as an initial attempt to demonstrate LLMs' capacity to extract medication dosing information from real-world data. Utilization of LLMs can augment clinicians in clinical practice through medication reconciliation and enhance research endeavors.

ABSTRACT Pharmacogenomics (PGx) utilizes genetic information to optimize medication management. Barriers to PGx implementation include limited confidence and knowledge in applying results, time constraints, and financial barriers. Clinical services across health systems vary greatly with differing consultative PGx service models. Most research on clinicians and PGx has been centered around attitudes and perspectives, with limited data regarding clinician satisfaction with PGx clinical services. An internally developed survey was created to assess clinician satisfaction with PGx services across a single health system. A survey was deployed to 645 clinicians (physicians, advanced practice providers, and genetic counselors) who utilized PGx clinical services (ordered PGx testing or a referral to the PGx clinic) within the past 3 years. Surveys were distributed via secure email in June 2025. A total of 36 out of 645 clinicians participated in the survey, which is noted as a limitation. Respondents tended to be satisfied (“somewhat satisfied” or “very satisfied”) across several clinical service domains (e.g., process of ordering PGx testing, integration into the electronic health record, and return of PGx results). Pharmacist notes and clinical decision support yielded the highest satisfaction. Most clinicians reported PGx results have positively impacted patient outcomes. Nearly half of respondents noted experiencing barriers in explaining PGx results to patients. The PGx clinic may help mitigate barriers in explaining PGx results to patients. Overall, responding clinicians were satisfied with the majority of PGx clinical services.

OBJECTIVE:To develop a statistical model to capture medication dosing for proton pump inhibitors (PPIs) using structured data from electronic health records (EHR). METHODS:Medication data for PPIs was extracted from a single health care system EHR to develop a statistical model. Nearly 20 years' worth of PPI prescriptions were extracted and 25% of unique dosing regimens were manually labeled by two clinical pharmacists. Several machine learning models were trained and evaluated to predict dose. Training was applied to 70% of the unique dosing regimens. The remaining unique dosing regimens were tested and validated with standard regression metrics: root mean squared error (RMSE) and R-squared. RESULTS:A total of 17,271 distinct patients had orders for a PPI comprising 186,801 unique PPI orders. Distinct pairs built on medication descriptions and SIG combinations resulted in 10,739 unique entities. Clinical pharmacists manually labeled 2679 examples for medication entity extraction. Regression metrics (R-squared, RMSE) were chosen as metrics to evaluate model performance. A stacked ensembled model proved to have the best results with a 0.09 RMSE and an R-squared of 0.825. CONCLUSION:The development of a statistical model to capture PPI dosing for both maintenance and complex dosing strategies was highly sensitive and accurate. A supervised learning prediction model helps overcome challenges in medication dosing identification by addressing concerns related to variability and complexity. Future strategies should focus on integrating unstructured data within the algorithm to further refine medication dosing capture.

BACKGROUND:CYP2D6 affects metabolism of several opioids; however, the clinical impact of genetic variants on efficacy has limited evidence in large patient populations. OBJECTIVE:This study aims to assess the impact of CYP2D6 phenotype on pain response in an elective pharmacogenomics (PGx) screening population prescribed opioids. METHODS:A retrospective review was conducted on hospitalized patients with CYP2D6 genotyping, prescribed either codeine, tramadol, hydrocodone, or oxycodone within 24 months prior to PGx testing and through 36 months after results. Pain scores were abstracted on a 10-point analog scale and categorized into 3 cohorts (mild, moderate, and severe) based on their baseline pain score. Baseline pain score was measured within 30 min of each opioid dose administration. Percentage changes in pain scores from baseline to 6, 12, 24, and 48 hours following each respective opioid dose administration were analyzed. Morphine milligram equivalents (MME) were averaged amongst the days of opioid administration. RESULTS:A total of 8062 patients were analyzed. Oxycodone was the most administered (4856, 41%). Mild pain cohort poor metabolizers (PMs) had significant increase in pain scores compared to normal metabolizers (NMs) at all hours from baseline (P < 0.001). PMs in moderate and severe pain cohorts had significantly decreased pain score reduction than NMs at all hours from baseline (P < 0.001). PMs had significantly higher MME compared to NMs in these cohorts (15 vs. 10, P < 0.001). CONCLUSION:CYP2D6 PMs had significantly less pain score reduction. CYP2D6 genotyping can lead to effective use of opioids in pain management and may display greater impact on efficacy of oxycodone than previously studied. Full implication of PGx testing is limited by the study's retrospective nature. PMs across all pain intensity cohorts had significantly less reduction in pain scores from baseline compared to NMs (P < 0.05). These results encourage further investigation into prospective pre-emptive CYP2D6 testing regarding effective pain management by optimizing opioid administration.

The aim of this study was to investigate the influence of CYP2D6 functionality on metabolic abnormalities in patients prescribed aripiprazole. A retrospective chart review was conducted in patients who had CYP2D6 genotyping and an oral prescription for aripiprazole. Primary outcomes were changes in metabolic parameters from baseline based on CYP2D6 genotype–based phenotype. Patients concurrently taking medications that cause CYP2D6 inhibition were classified as phenoconverters and were categorized as poor metabolizers in subanalyses. Index and post-aripiprazole (after either initiation or dose modification) data were abstracted from the electronic medical record for the following endpoints: body mass index (BMI), blood glucose, glycated hemoglobin, low-density lipoprotein (LDL), triglycerides, and total cholesterol. In multivariate mixed-effects analysis of 1,562 patients, the genotype-based phenotype cohort had no significant differences. A marginal difference in triglyceride levels was observed between poor metabolizers and normal metabolizers (β = 23.42; range, –4.9 to 51.73; P = 0.11). Findings were similar between poor metabolizers and normal metabolizers when evaluating outcomes encompassing phenoconversion, although poor metabolizers had an increased mean change in LDL levels (β = 10.04; range, 1.46 to 18.61; P < 0.05) and had a decreased mean change in BMI (β = –0.34; range, –0.72 to 0.04; P = 0.08); however, these differences were not statistically significant. This study suggests that the incidence of metabolic abnormalities beyond LDL may not be higher in CYP2D6 poor metabolizers compared to normal metabolizers. Further research is needed to investigate clinically meaningful outcomes in relation to CYP2D6 and aripiprazole.

INTRODUCTION:Genetic variants can impact medication response. The study of genetic variants on medications is called pharmacogenomics (PGx). Understanding PGx results can be difficult as results are reported differently than other laboratory tests. Patients have reported a lack of understanding and satisfaction with PGx information. METHODS:Surveys were emailed to patients seen in the PGx clinic and patients who participated in an elective screening (Sanford Chip) at Sanford Health. Surveys were conducted to assess literacy, understanding and satisfaction of PGx testing. Survey responses were summarized using descriptive statistics. RESULTS:There were 121 responses that were initially collected. A total of 100 responses were included in the analysis. The median response amongst all individuals was 9 out of a possible 13 points on the PGx literacy assessment. PGx clinic patients had increased satisfaction compared to Sanford Chip patients for being able to understand results (p < 0.05), that PGx test provided information to improve my care plan (p < 0.05), that they feel confident that my medication will be effective for me based on my genetics (p < 0.05), were satisfied with communication of results (p < 0.001) and overall experience (p < 0.01). DISCUSSION:Implementation of a PGx clinic improves patient experience, confidence, and satisfaction.

Pharmacogenomic (PGx) considerations are increasingly influencing health system formularies, prompting Pharmacy and Therapeutics (P&T) committees to integrate genetic insights into their decision-making processes. This can be the case in several clinical scenarios, including (1) Prioritizing the use of lower-cost drugs, where PGx testing is applied to guide the use of lower-cost medications when appropriate, (2) Rescuing medications that are effective in the majority of individuals but were removed from the formulary because of harm in a subset of individuals with a specific phenotype, (3) Improving medication safety by utilizing PGx testing/results to guide initiation of reduced drug doses or use of alternative therapy in patients with at-risk genotypes, and (4) Restricting use of high-cost drugs to individuals most likely to respond based on genotype. The primary objective of this PRN opinion piece is to describe how PGx impacts formulary management, including the challenges and opportunities that arise from precision medicine approaches to prescribing.

Objectives:This study evaluates response rates of pharmacogenomics (PGx) nonsteroidal anti-inflammatory drugs (NSAIDs) clinical decision support (CDS) alerts at Sanford Health from May 2020 to December 2024. Materials and Methods:A retrospective analysis was conducted on PGx NSAIDs interruptive alerts. Response options were classified into five categories (1) continuation of triggering NSAID order, (2) dose modification, (3) alternative NSAID ordered without PGx implications, (4) alternative analgesic (ie, opioid) ordered, and (5) discontinuation of NSAID without alternative therapy. Results:The study analyzed 2361 alert instances from 978 patients. The most common response was discontinuing NSAID without alternative therapy (43%). Dose modifications and orders for alternative analgesics comprised 2.57% and 14.67% of responses, respectively. The initial acceptance rate was 62.6%. Prior NSAID use significantly impacted override rates (60% vs 40%, P < .001). A 409-day breaking point was observed to affect alert acceptance rates, with the highest acceptance in NSAID naïve patients (96.1%). Discussion:PGx NSAIDs CDS alert acceptance rates were higher compared to general CDS acceptance rates. This study highlights opportunities for continuous improvement including optimizing alert modality, modifying alert criteria to include look-back periods, and implementing genetically adapted ordersets. Conclusion:The initial acceptance rate of PGx NSAIDs CDS alerts was 62.6%, however, with significantly higher acceptance rates in NSAID naïve patients (62.6% vs 96.1%, P < .001). Integration of CDS is vital to the successful implementation of PGx in clinical practice.

Purpose We describe the implementation and ongoing maintenance of CYP2C19 and CYP2D6 focused pharmacogenetic (PGx) testing to guide antidepressant and antianxiety medication prescriptions in a large rural, nonprofit health system. Summary Depression and anxiety are common psychiatric conditions. Sanford Health implemented PGx testing for metabolism of cytochrome P450 (CYP) isozymes 2C19 and 2D6 in 2014 to inform prescribing for multiple medications, including antidepressant and antianxiety therapies. As guidelines, genotype to phenotype translation, panel offerings, and other resources are updated, we adapt our approach. We make educational and informational materials available to providers and patients. Pharmacogenomic clinical pharmacists review PGx results with discrete values and provide guidance documentation in the electronic medical record. A robust clinical decision support system is in place to provide interruptive alerts, noninterruptive alerts, and genomic indicators. A referral-based interdisciplinary clinic is also available to provide in-depth education to patients regarding PGx results and implications. Additionally, partnering with our health plan has expanded access to PGx testing for patients with anxiety or depression. Conclusion The implementation and maintenance of Sanford Health’s PGx program to guide antidepressant and antianxiety medication use continues to evolve and requires a multipronged approach relying on both human and informatics-based resources.

Background: Statins are commonly used medications. Variants in SLCO1B1, CYP2C9, and ABCG2 are known predictors of muscle effects when taking statins. More exploratory genes include RYR1 and CACNA1S, which can also be associated with disease conditions. Methods: Patients with pathogenic/likely pathogenic variants in RYR1 or CACNA1S were identified through an elective genomic testing program. Through chart review, patients with a history of statin use were assessed for statin-associated muscle symptoms (SAMS) along with collection of demographics and other known risk factors for SAMS. Results: Of the 23 patients who had a pathogenic or likely pathogenic RYR1 or CACNA1S variant found, 12 had previous statin use; of these, SAMS were identified in four patients. Conclusion: These data contribute to previous literature suggesting patients with RYR1 variants may have an increased SAMS risk. Additional research will be helpful in further investigating this relationship and providing recommendations.

Aim: Clopidogrel requires CYP2C19 activation to have antiplatelet effects. Pharmacogenetic testing to identify patients with impaired CYP2C19 function can be coupled with clinical decision support (CDS) alerts to guide antiplatelet prescribing. We evaluated the impact of alerts on clopidogrel prescribing. Materials & methods: We retrospectively analyzed data for 866 patients in which CYP2C19-clopidogrel CDS was deployed at a single healthcare system during 2015-2023. Results: Analyses included 2,288 alerts. CDS acceptance rates increased from 24% in 2015 to 63% in 2023 (p < 0.05). Adjusted analyses also showed higher acceptance rates when clopidogrel had been ordered for a percutaneous intervention (OR: 28.7, p < 0.001) and when cardiologists responded to alerts (OR: 2.11, p = 0.001). Conclusion: CDS for CYP2C19-clopidogrel was effective in reducing potential drug-gene interactions. Its influence varied by clinician specialty and medication indications.

Introduction: Clopidogrel remains widely utilized in patients undergoing percutaneous coronary intervention despite compelling evidence that genetic variation impacts patient response to clopidogrel. Clinical decision support (CDS) is frequently used to aid in precision medicine; however, limitations within the electronic medical record can hinder reliable CDS. Procedural areas where standard order entry is not utilized create a barrier to CDS implementation. Objectives: We aimed to evaluate the implementation of a novel alerting mechanism on genotype-guided antiplatelet prescribing within the cardiac catheterization laboratory procedural setting. Methods: A retrospective cohort study was conducted to assess the rate of antiplatelet ordering in patients with one or two loss-of-function cytochrome P450 2C19 (CYP2C19) alleles before and after alert implementation. Pharmacogenomic congruence was measured before and after the alert via chart abstraction that included the CYP2C19 genotype and antiplatelet medications ordered within that encounter. Results: A total of 236 patients were included in analyses, 127 encounters within the cohort before alert implementation and 136 encounters in the cohort after alert implementation. Prior to alert implementation, 40.9% (n = 127) were prescribed clopidogrel compared with 25.7% (n = 136) post implementation. After implementing a genotype-guided alert within the cardiac catheterization laboratory procedural setting, providers were 2.22 times more likely to prescribe an alternative antiplatelet (p = 0.024). Clopidogrel-na & iuml;ve patients were 9.75 times more likely to receive a genotype-guided antiplatelet order following alert implementation (p < 0.05). Conclusion: Providers were responsive to a novel alert within the cardiac catheterization laboratory procedural setting. Genotype-guided antiplatelet prescribing significantly increased following the alert implementation.

CYP2C19 genotyping to guide antiplatelet therapy after patients develop acute coronary syndromes (ACS) or require percutaneous coronary interventions (PCIs) reduces the likelihood of major adverse cardiovascular events (MACE). Evidence about the impact of preemptive testing, where genotyping occurs while patients are healthy, is lacking. In patients initiating antiplatelet therapy for ACS or PCI, we compared medical records data from 67 patients who received CYP2C19 genotyping preemptively (results >7 days before need), against medical records data from 67 propensity score-matched patients who received early genotyping (results within 7 days of need). We also examined data from 140 patients who received late genotyping (results >7 days after need). We compared the impact of genotyping approaches on medication selections, specialty visits, MACE and bleeding events over 1 year. Patients with CYP2C19 loss-of-function alleles were less likely to be initiated on clopidogrel if they received preemptive rather than early or late genotyping (18.2%, 66.7%, and 73.2% respectively, p = 0.001). No differences were observed by genotyping approach in the number of specialty visits or likelihood of MACE or bleeding events (all p > 0.21). Preemptive genotyping had a strong impact on initial antiplatelet selection and a comparable impact on patient outcomes and healthcare utilization, compared to genotyping ordered after a need for antiplatelet therapy had been identified.

Purpose To evaluate the impact of dose-specific logic for tricyclic antidepressant (TCA) pharmacogenomics (PGx) clinical decision support (CDS). We aimed to provide guidance in an area with limited supporting literature, ensure optimal dosing through CDS, and limit alert fatigue. The primary outcome was the reduction in alerts prescribers encountered, while the secondary outcome included an analysis across specialties.Methods A retrospective chart review was conducted to examine TCA PGx CDS before and after implementation of dosing criteria for alerts. Data were abstracted from the electronic medical record. A chi 2 test was performed to analyze the frequency of alerts in behavioral health and other specialties.Results In the cohort lacking dose criteria, most TCA orders were for indications other than depression (76%) and guidelines would not apply to the majority of these orders. Using dosing criteria to refine CDS reduced the volume of TCA alerts by 74.8%. Alert volume decreased the most in specialties other than behavioral health due to prescriptions for indications other than anxiety or depression (P = 0.035).Conclusion Dose-centric alerts may be used as a strategy to achieve optimal dosing. Alerting clinicians when dose modifications should occur contributes to getting the right dose to the right patient. Future efforts should focus on optimal dosing of medication through CDS enhancements.

Meaningful clinical decision support (CDS) recommendations are vital for implementation of pharmacogenomics (PGx) into routine clinical care. PGx CDS alerts include interruptive and noninterruptive alerts. The objective of this study was to evaluate provider ordering behavior after noninterruptive alerts are displayed. A retrospective manual chart review was conducted from the time of noninterruptive alert implementation to the time of data analysis to determine congruence with CDS recommendations. The congruence rate for noninterruptive alerts was 89.8% across all drug-gene interactions. The drug-gene interaction with the most alerts for analysis included metoclopramide (n = 138). The high rate of medication order congruence after noninterruptive alerts were deployed suggests this modality may be appropriate for PGx CDS as a method for best practice adherence.

Introduction: Pharmacogenomics (PGx) aims to maximize drug benefits while minimizing risk of toxicity. Although PGx has proven beneficial in many settings, clinical uptake lags. Lack of clinician confidence and limited availability of PGx testing can deter patients from completing PGx testing. A few novel PGx clinic models have been described as a way to incorporate PGx testing into the standard of care. Background: A PGx clinic was implemented to fill an identified gap in provider availability, confidence, and utilization of PGx across our health system. Through a joint pharmacist and Advanced Practice Provider (APP) collaborative clinic, patients received counseling and PGx medication recommendations both before and after PGx testing. The clinic serves patients both in-person and virtually across four states in the upper Midwest. Results: The majority of patients seen in the PGx clinic during the early months were clinician referred (77%, n = 102) with the remainder being self-referred. Patients were, on average, taking two medications with Clinical Pharmacogenetics Implementation Consortium guidelines. Visits were split almost equally between in-person and virtual visits. Conclusion: Herein, we describe the successful implementation of an interdisciplinary PGx clinic to further enhance our PGx program. Throughout the implementation of the PGx clinic we have learned valuable lessons that may be of interest to other implementors. Clinicians were actively engaged in clinic referrals and early adoption of telemedicine was key to the clinic's early successes.