Fenglei Huang

Background/Objectives: Patients with heart failure (HF) and chronic kidney disease (CKD) frequently receive complex multidrug regimens, increasing the potential for pharmacokinetic drug–drug interactions (DDIs). Methods: This study used U.S. Optum^® claims data from the period 2019–2024 to characterize medication use prevalence in adults with HF or CKD and identify potential DDI liabilities by mapping commonly used medications to established cytochrome P450 (CYP) enzyme and transporter pathways. Potential mechanistic interactions with guideline-directed medical therapy were further prioritized using a semi-quantitative framework based on perpetrator strength and substrate sensitivity, with Tier 1 and Tier 2 representing higher-priority CYP-mediated interaction candidates. Results: The analysis included 856,041 patients with HF and 1,893,827 patients with CKD, with median ages of 75 and 72 years, respectively. In CKD, atorvastatin (25.7%) and metformin (19.1%) were among the most prevalent medications, while metoprolol (33.2%) and furosemide (31.9%) were highly prevalent in HF. Many commonly used medications had established substrate or inhibitor liabilities involving CYP3A, CYP2D6, and transporters including P-gp, OATP1B, MATE, OCT2, and BCRP. CYP3A liabilities were identified in 13 of the 25 most frequently used medications in each population. Higher-priority potential interactions were pathway-specific, with Tier 1 and Tier 2 candidates primarily involving CYP3A in CKD and CYP2D6 in HF. Integrating real-world medication prevalence with established mechanistic DDI information provides a practical, hypothesis-generating framework for identifying and prioritizing potential pharmacokinetic DDI liabilities in HF and CKD. Conclusions: These findings may inform subsequent DDI evaluations and clinical pharmacology study planning to support drug development and future risk-mitigation strategies, although actual co-exposure, clinical DDIs, and associated outcomes were not assessed.

Lung cancer predominantly affects older adults who already experience comorbidities and are at increased risk of drug–drug interactions (DDIs) due to polypharmacy. Traditional DDI assessments often do not fully capture complex co‐medication patterns observed in real‐world practice due to the risk of introducing confounding factors. This study used U.S. Optum claims data from 2019 to 2024 to characterize co‐medication patterns in predominately non‐small cell lung cancer (NSCLC) patients and map enzyme‐ and transporter‐mediated liabilities across commonly used medications, chemotherapy agents, and biomarker‐driven therapies. Potential DDI liabilities were further prioritized using a semi‐quantitative clinical framework based on perpetrator strength, substrate sensitivity, narrow therapeutic index status, and available evidence. Frequently used medications included disease‐specific supportive therapies such as ondansetron and dexamethasone, as well as treatments for common U.S. comorbidities such as cardiovascular and endocrine disorders. Many of the co‐medications were substrates, inhibitors, and inducers of pathways critical for NSCLC treatments such as CYP3A, CYP2D6, P‐gp, and OATP1B and additional risks were identified when giving treatments with acid‐reducing agents or QT‐prolonging drugs. These findings demonstrate how integrating real‐world data (RWD) can strengthen DDI assessment by improving clinical trial design, informing risk‐mitigation strategies, and supporting more patient‐centric drug development.

Endogenous biomarkers have gained considerable attention as tools for assessing transporter‐mediated drug–drug interactions (tDDIs). While plasma coproporphyrin I is considered a validated biomarker for assessing hepatic organic anion transporting polypeptide (OATP) 1B‐mediated DDI risk, data supporting biomarkers for other major drug transporters, particularly renal transporters, remains limited. The International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) included endogenous biomarker‐based assessments in its recent guidance, ICH M12, as an alternative approach for evaluating tDDIs, which is anticipated to broaden the use of this strategy in early clinical development. However, methodological gaps remain in biomarker measurement, including study design, biological matrix selection, and data analysis. Importantly, consensus is still needed on the most appropriate biomarkers for renal organic cation transporter (OCT) 2 and multidrug and toxin extrusion (MATE) transporters 1 and 2‐K, such as creatinine, N ¹ ‐methylnicotinamide, and N ¹ ‐methyladenosine, given the influence of various extrinsic and intrinsic factors on their performance. This review aims to summarize the current state of clinical evidence for endogenous biomarkers of major hepatic and renal transporters, including study design considerations relevant to meaningful data interpretation. It also briefly summarizes reported Physiologically based pharmacokinetic (PBPK) models of these biomarkers, highlights key gaps in current understanding, and identifies future work needed to expand the application of a biomarker‐based tDDI strategy.

Mechanistic modeling remains a cornerstone of model-informed drug development (MIDD), enabling quantitative integration of biological mechanisms with pharmacokinetic and pharmacodynamic knowledge to support drug discovery, translational research, dose optimization, and precision medicine. The four studies in this collection illustrate how mechanistic modeling is being applied across the entire drug development continuum, from early lead optimization to clinical dose selection in special populations.Xu et al. developed a mechanistic quantitative systems pharmacology (QSP) platform integrating multiple immune checkpoint pathways with extensive preclinical datasets to evaluate bispecific antibody target combinations, characterize dose-response relationships, and explore alternative dosing strategies. Model simulations further suggested that combining checkpoint-targeting bispecific antibodies with monoclonal antibodies could enhance antitumor efficacy, demonstrating the value of QSP for optimizing antibody design and translational development.Expanding mechanistic modeling into early drug discovery, Zhang et al. integrated artificial intelligence with physiologically based pharmacokinetic (AI-PBPK) modeling to predict the pharmacokinetic and pharmacodynamic properties of aldosterone synthase inhibitors directly from their chemical structures. The framework enabled rapid prediction of PK/PD properties for multiple compounds and provides an efficient approach for lead optimization and candidate selection before clinical evaluation.Using another complementary PBPK strategy, Zhang et al. combined quantitative structureactivity relationship (QSAR) modeling with mechanistic PBPK modeling to predict the human pharmacokinetics and tissue distribution of 34 fentanyl analogs. Besides improving prediction accuracy without extensive experimental data, the framework identified analogs with greater predicted central nervous system penetration, providing valuable insights into their potential abuse risk and supporting public health and regulatory risk assessment.Moving from discovery to clinical development, Pokladyuk et al. established a PBPK model to characterize niraparib pharmacokinetics in patients with hepatic impairment. By incorporating disease-related physiological changes into the model, simulations successfully supported dose adjustments that maintained drug exposure within the desired therapeutic range, illustrating the value of mechanistic PBPK modeling for precision dosing in special populations.Together, these studies illustrate how mechanistic modeling has evolved from describing pharmacokinetics to becoming an integrated decision-making framework that supports drug discovery, translational research, precision dosing, and clinical development.Artificial intelligence (AI) and machine learning (ML) are increasingly complementing traditional pharmacometric and mechanistic modeling by improving prediction, integrating diverse datasets, and supporting data-driven decision-making throughout drug development. Rather than replacing mechanistic approaches, AI enhances their predictive capability and broadens their applications across drug discovery, precision dosing, and translational research.At the drug discovery stage, Zhang et al. demonstrated how AI can be integrated with physiologically based pharmacokinetic (AI-PBPK) modeling to predict the pharmacokinetic and pharmacodynamic properties of aldosterone synthase inhibitors directly from their chemical structures. By enabling rapid screening and optimization of lead compounds before clinical evaluation, this framework illustrates the potential of AI to accelerate candidate selection while complementing mechanistic modeling.The application of AI also extends into clinical practice. Jian et al. developed a machine learningbased model to predict teicoplanin plasma concentrations in patients with liver disease using real-world therapeutic drug monitoring data. Among ten evaluated algorithms, the LightGBM model achieved the best predictive performance by integrating clinical and laboratory variables, demonstrating how machine learning can improve individualized dosing and therapeutic drug monitoring in patients with altered pharmacokinetics.Providing a broader perspective, Baran et al., in the only review article of this Research Topic, discussed the emerging role of hybrid mechanistic-machine learning PK/PD models that integrate digital biomarkers, continuous monitoring data, imaging, and multi-omics datasets. The authors highlight that AI is most powerful when combined with mechanistic pharmacology, improving pharmacokinetic prediction, covariate identification, and cross-species translation while emphasizing the importance of rigorous validation, interpretability, and regulatory acceptance.Taken together, these advances suggest that AI will be most impactful when integrated with mechanistic pharmacology rather than used as a standalone predictive technology. By combining machine learning with mechanistic models and real-world clinical data, AI has the potential to accelerate drug discovery, optimize precision dosing, and support model-informed decision-making while maintaining biological interpretability and regulatory credibility.Population PK/PD modeling remains a cornerstone of individualized therapy and an essential component of model-informed drug development. By quantifying interindividual variability and identifying clinically relevant covariates, these approaches support dose optimization, precision medicine, and evidence-based treatment across diverse patient populations. The studies in this collection demonstrate the expanding role of population PK/PD modeling in pharmacogenomics, biosimilar development, special populations, precision dosing, and long-term safety evaluation.The integration of population PK modeling with pharmacogenomics is illustrated by Wang et al., who investigated the influence of ABCB1 polymorphisms on...

Background: Organic Anion Transporting Polypeptide (OATP) 1B1/3 inhibition can lead to clinically significant drug-drug interactions (DDIs). Coproporphyrin-I (CP-I) is an established endogenous biomarker for assessing the OATP1B inhibitory activity of new chemical entities. However, current physiologically based pharmacokinetic (PBPK) models for CP-I are restricted to proprietary software, limiting widespread adoption and collaboration. Methods: We developed and qualified an open-source PBPK model for CP-I using the Open Systems Pharmacology (OSP) Suite. The model describes the endogenous synthesis and OATP1B1/MRP2-mediated disposition of CP-I. It was integrated into a DDI network comprising the precipitants rifampicin, probenecid, and cyclosporine A, and the object drugs atorvastatin, pitavastatin, and rosuvastatin. A key methodological feature was the estimation of in vivo inhibition constants (K _i,OATP1B ) for precipitants based on clinical CP-I data rather than relying solely on in vitro measurements. Results: The CP-I model successfully captured baseline plasma concentration-time profiles with a Geometric Mean Fold Error (GMFE) of 1.15. In the DDI network analysis, using CP-I-derived in vivo K _i,OATP1B values for the strong OATP1B inhibitors rifampicin and cyclosporine yielded superior predictive accuracy compared to in vitro values, with most predicted DDI ratios for statins falling within the 2-fold acceptance range. Specifically for cyclosporine, the model required an adjustment of the endogenous CP-I synthesis rate to account for baseline differences in the specific study population; incorporating this allowed for the estimation of an in vivo K _i,OATP1B (4.7 nM) that accurately predicted interactions with pitavastatin and rosuvastatin. Notably, probenecid, a weak OATP1B inhibitor, showed an in vivo K _i,OATP1B estimate within a similar range of its in vitro measured value. Consequently, the scaling factor between in vitro and in vivo K _i,OATP1B appears to be a compound-specific parameter requiring independent identification for each precipitant. Conclusion: This study presents a qualified, freely available PBPK framework for CP-I. It demonstrates that calibrating participant models using early clinical biomarker data enables reliable prospective prediction of OATP1B-mediated DDIs. This workflow supports model-informed drug development (MIDD) strategies, potentially reducing the need for dedicated clinical DDI studies.

Faldaprevir (FDV) is a novel NS3/NS4A inhibitor used in the treatment of hepatitis C infection in an interferon-free regimen. This study evaluated the safety, tolerability, and pharmacokinetics of FDV following a single dose in healthy male subjects and assessed the effect of food on FDV bioavailability. In the placebo-controlled, randomized, single-blind, single-increasing-dose part of the study (Part 1), 64 healthy male subjects were randomized to receive FDV in PEG/TRIS/meglumine solution at one of eight dose levels (4–1,200 mg, n = 6 per dose group) or placebo (n = 2 per dose group). In Part 2, the effect of food on the relative bioavailability (rBA) of 480 mg FDV in solution was evaluated in an open-label, crossover comparison, with and without a high-fat breakfast, in an additional 10 subjects (8 FDV and 2 placebo). Following single doses of 4–1,200 mg FDV, geometric mean (gMean) C_max and AUC_0-inf were 3.57–16500 ng/mL and 254–402000 h*ng/mL, respectively, displaying more than dose-proportional increases in exposure. FDV was slowly absorbed, with gMean t_1/2 and median t_max of 15.5–39.2 h and 4.0–14.0 h, respectively; both were dose dependent. The urinary excretion of FDV was less than 0.1% of the dose. A high-fat breakfast increased systemic exposure to FDV in solution by 14%. FDV was generally well tolerated; subjects who experienced adverse events (AEs) recovered without sequelae, and no serious AEs were reported. Indirect (unconjugated) bilirubin of >3.0 mg/dL was observed in two subjects at 480 mg and five subjects at 1,200 mg. In conclusion, at single doses of 4–1,200 mg in healthy male subjects, FDV showed dose-dependent pharmacokinetics and was generally considered safe and well tolerated. Food had no clinically relevant effect on the rBA of FDV.

Faldaprevir (FDV) is an investigational NS3/NS4A protease inhibitor for chronic hepatitis C. This study evaluated the safety, tolerability, and pharmacokinetics (PK) of FDV after multiple rising doses in healthy male volunteers and subjects with Gilbert syndrome (GS). In this randomized, double‐blind, placebo‐controlled study, healthy males received once‐daily oral FDV (20, 48, 120 mg [ n = 6 per group], 240 mg [ n = 5]), or placebo ( n = 7). A single dose was given on Day 1, followed by a 72‐h washout and 21 days of dosing from day 4. Separately, 9 GS subjects received open‐label FDV 240 mg daily for 28 days. PK was assessed after the first and last doses; safety was evaluated throughout. FDV showed greater than dose‐proportional increases in exposure (gMean C _max,ss : 99–5360 ng/mL; AUC _{τ ,ss} : 1740–50,100 h·ng/mL) and time‐dependent PK with a linearity index >1. Mean t _1/2 was 20–30 h; steady state was reached in 6–7 days with an accumulation ratio of 2.8–3.1. FDV exposure in GS subjects was similar but slightly lower. Total bilirubin increased dose‐dependently, with higher indirect bilirubin in GS subjects. FDV exhibited non‐linear, time‐dependent PK and was generally well tolerated up to 240 mg/day in subjects with or without GS.

Background: Pharmacokinetic drug–drug interactions (DDIs) can be caused by the effect of a pharmaceutical compound on the activity of one or more subtypes of the Cytochrome P450 (CYP) family, UDP-glucuronosyltransferases (UGTs), and/or transporters. As the number of therapeutic areas with polypharmacy has increased, interest has grown in assessing the risk of DDIs during the early phases of drug development. Various lines of research have led to improved mathematical models to predict DDIs, culminating in the Food and Drug Administration’s (FDA) guidelines on evaluating pharmacokinetic DDI risks. However, the recommended static models are highly conservative and often result in false positive predictions. The current research aims to improve the workflow for assessing CYP-mediated DDI risk using Boehringer Ingelheim (BI) proprietary compounds. Methods: The Drug–drug Interaction Risk Calculator (PharmaPendium) was used to evaluate the mechanistic static model, and predictions were correlated with human pharmacokinetic studies from Phase I clinical trials. Results: The results demonstrated that the FDA formula performed well in predicting DDIs for BI proprietary compounds. Furthermore, the integration of either human renal excretion or preclinical species total excretion data into the mechanistic static model enhanced the predictive performance for candidate drugs as victims in DDIs. Conclusions: The basic static models (BSMs) for drug interactions should be used in early drug discovery to “rule out” DDI risks because of the minimal inputs required and the low rate of false negative predictions. Mechanistic static models (MSMs) can then be implemented for compounds that require additional evaluation.

The renal secretion of many drugs is facilitated by membrane transporters, including organic cation transporter 2, multidrug and toxin extrusion protein 1/2-K and organic anion transporters 1 and 3. Inhibition of these transporters can reduce renal excretion of drugs and thereby pose a safety risk. Assessing the risk of inhibition of these membrane transporters by investigational drugs remains a key focus in the evaluation of drug–drug interactions (DDIs). Current methods to predict DDI risk are based on generating in vitro data followed by a clinical assessment using a recommended exogenous probe substrate for the individual drug transporter. More recently, monitoring plasma-based and urine-based endogenous biomarkers to predict transporter-mediated DDIs in early phase I studies represents a promising approach to facilitate, improve and potentially avoid conventional clinical DDI studies. This perspective reviews the evidence for use of these endogenous biomarkers in the assessment of renal transporter-mediated DDI, evaluates how endogenous biomarkers may help to expand the DDI assessment toolkit and offers some potential knowledge gaps. A conceptual framework for assessment that may complement the current paradigm of predicting the potential for renal transporter-mediated DDIs is outlined.