Comparison of KN012, a denosumab biosimilar, versus reference denosumab in Chinese postmenopausal women with osteoporosis: efficacy, safety, and population pharmacokinetics in a 12-month phase III study.

Dupilumab, a fully human monoclonal antibody that blocks key drivers of type 2 inflammation, is approved across 8 diseases, including eosinophilic esophagitis. Subcutaneous dupilumab 300 mg weekly improved outcomes vs. placebo in adults (aged ≥ 18 years) and adolescents (aged ≥ 12 to < 18 years) during the phase III LIBERTY EoE TREET study (NCT03633617). The phase III EoE KIDS study (NCT04394351) demonstrated efficacy of a weight-tiered, higher exposure dupilumab dose approximating 300 mg weekly exposure in TREET vs. placebo in children aged ≥ 1 to < 12 years. To characterize dupilumab pharmacokinetics, inform clinical trial design, and optimize posology, a population pharmacokinetic model was developed using dupilumab concentration data from 6 single-dose studies in healthy adults and 3 eosinophilic esophagitis clinical trials. Healthy volunteer data were used to construct a stable base model; incorporation of patient data allowed estimation covariate effects for the disease population. Observational data from up to 52 weeks of treatment in eosinophilic esophagitis trials validated model performance. Significant effects were identified for body weight on clearance and volume of distribution, including time-varying weight in children and adolescents, and of albumin on clearance; age was not a significant covariate. Simulations identified alternative dupilumab regimens for patients weighing ≥ 5 to <15 kg (200 mg every 3 weeks) and ≥ 40 to <60 kg (300 mg weekly) with improved exposure matching to the 300 mg weekly dose in TREET. The final model was used to inform regulatory approval and potential future clinical trial design.

Model-informed precision dosing of paroxetine to optimize individualized therapy in patients with mental disorders.

Mitiperstat is a novel myeloperoxidase inhibitor being investigated for the treatment of heart failure, metabolic dysfunction-associated steatohepatitis, and chronic obstructive pulmonary disease. Pharmacokinetic data from five clinical trials were used to develop a mitiperstat population pharmacokinetics (popPK) model. In total, 2856 plasma samples from 128 mitiperstat-treated participants were collected during the first-in-human single-ascending dose (SAD) study; multiple-ascending dose studies (MAD) in healthy volunteers, including healthy Japanese and Chinese volunteers (JCMAD); the phase 2a SATELLITE study in patients with heart failure with preserved/mildly reduced ejection fraction (HFpEF/HFmrEF); and from a study in patients with severe renal impairment. The mitiperstat concentration-time data were pooled and analyzed with the software NONMEM. The covariates considered in this analysis were baseline estimated glomerular filtration rate (eGFR), baseline body weight, baseline body mass index, sex, race (Asian or non-Asian), age, disease status (healthy volunteers or patients with HFpEF/HFmrEF), and formulation (oral suspension or tablet). The final popPK model is a two-compartment model with first-order absorption and linear elimination. Baseline body weight, disease status, Asian race, and eGFR were identified as significant covariates for apparent clearance, and age was identified as a significant covariate for apparent central volume of distribution. The final model predicts that severe renal impairment and lower body weight have the largest impact on exposure. Exposure to mitiperstat increased as eGFR declined. This popPK model constitutes an important step toward optimizing doses for efficacy and safety in the different mitiperstat development programs. Trial Registration: ClinicalTrials.gov: NCT02712372.

A revised classification of FVIII concentrates: rationale and novel metrics.

Achieving the target AUC/MIC remains a critical challenge in vancomycin therapeutic drug monitoring, with traditional empirical dosing regimens often leading to suboptimal outcomes. This study aimed to develop and validate a novel population pharmacokinetic (PopPK)-informed machine learning model to optimize TDM-guided dose adjustments. By accurately predicting the 24-h area under the curve (AUC), the model assists clinicians in rapidly individualizing regimens, thereby improving target attainment, reducing subsequent modifications and minimizing toxicity. A total of 1140 vancomycin TDM samples were analysed. Five machine learning models were developed and compared. Model performance assessed using R2, RMSE, Lin's concordance correlation coefficient (CCC), mean percentage error (MPE) and mean absolute percentage error (MAPE). The top-performing model (ANN) was subsequently validated on an independent test set. Its performance, interpretability and clinical safety were further examined using SHAP analysis, Bland-Altman plots, error grid analysis (EGA) and ROC curve evaluation. The ANN (MLP) model demonstrated the best comprehensive performance. It achieved the highest R2 (0.920 ± 0.027) and CCC (0.960 ± 0.011), coupled with the lowest RMSE (0.498 ± 0.076 L/h) and minimal systematic bias (MPE: 1.290 ± 1.774). SHAP analysis identified DV (measured concentration), Cys (mg/L) and Last dose as the three most influential predictive features. The ANN model achieved exceptional accuracy, robustness and clinical safety in predicting individualized vancomycin clearance. This model is deployable as an instantaneous, point-of-care CDSS tool, offering a powerful new pathway to solve the long-standing challenges of accessibility and accuracy in routine TDM practice.

Human biomonitoring provides direct measures of internal exposure to environmental chemicals, but translating biomarker concentrations into quantitative external exposure and risk estimates remains challenging. Chloromethylisothiazolinone (CMIT) and methylisothiazolinone (MIT) are widely used biocides, and population exposure is typically assessed via urinary biomarkers. However, a quantitative framework connecting biomonitoring data to external dose and health risk is currently lacking. This study aims to quantitatively reconstruct external CMIT/MIT exposures from human urinary biomonitoring data using a population pharmacokinetic (PopPK) model and to assess human health risk via integrating reverse dosimetry with an internal dose-based reference dose (RfD). A human-scale PopPK model capable of quantitatively describing urinary excretion of N-methylmalonamic acid and the mercapturic acid metabolite M-12 following oral CMIT/MIT exposure was developed. Model parameters were estimated using nonlinear mixed-effects modeling and evaluated via bootstrap analysis, visual predictive checks, goodness-of-fit diagnostics, and normalized prediction distribution error analysis. The validated model was then applied to adult biomonitoring data from the German Environmental Sample Bank and to pediatric and adolescent survey data. External exposure doses were reconstructed via reverse dosimetry, accounting for inter-individual PK variability. Human health risk was quantified using the margin of exposure (MOE) approach, referencing an internal dose-derived oral RfD of 0.02mg/kg/day. The final PopPK model reliably captured urinary biomarker excretion dynamics at population and individual levels. Parameter estimates were robust, with bootstrap medians closely aligned with the final model values. Reconstructed external exposures exhibited no consistent long-term increasing or decreasing trend in adults and no systematic age- or sex-related pattern in pediatric and adolescents. Most exposure scenarios yielded MOE values > 10, while only extreme upper-bound conditions yielded MOEs of approximately 2-5. Even under conservative assumptions, all MOE values remained above 1. This study demonstrates that human biomonitoring data can be quantitatively translated into external exposure and risk metrics using a PopPK-based reverse dosimetry framework. The findings indicate that current CMIT/MIT exposure levels in the general population are unlikely to pose health concerns under typical environmental conditions. The integrated biomonitoring-modeling approach offers a regulatory-relevant framework for linking internal biomarkers to external exposure and health risk assessment.

Busulfan's narrow therapeutic index and high pharmacokinetic (PK) variability warrant investigation of its exposure-toxicity relationship. We retrospectively analyzed 334 pediatric and young adult patients who underwent allogeneic hematopoietic stem cell transplantation at Seoul National University Children's Hospital between 2009 and 2020 and received once-daily intravenous busulfan over four days (3-hour infusion, n = 122; 6-hour infusion, n = 212). Factors associated with toxicities were identified using multivariable logistic regression, and overall survival (OS) was evaluated by Kaplan-Meier and Cox regression analyses. Despite comparable exposures, the 3-hour infusion group showed higher Cmax (4,260.2 vs 3,028.6 μg/L, P < 0.0001) and more frequent liver function test elevations (13.1% vs 0.5%, P < 0.0001). Receiver operating characteristic analysis identified a Cmax cut-off of 4269.6 μg/L as a factor for hepatotoxicity. The 4-year OS was superior in the 6-hour infusion group (82.3% vs 71.9%, P = 0.02) and in the low Cmax group (80.8% vs 66.5%, P = 0.01). Cox regression revealed Cmax as an independent predictor of OS. A population PK model was developed using NONMEM. A 1-compartment model incorporating enzyme turnover auto-inhibition successfully described the busulfan PK profile. Simulations of published once-daily dosing regimens showed that extending the infusion to 6 hours successfully maintained Cmax below the cut-off threshold. These results show that maintaining busulfan Cmax below 4269.6 μg/L is associated with improved safety and survival. Prolonged infusion should be considered when adopting the busulfan once-daily regimen.

Antimicrobial pharmacokinetics (PK) in patients receiving extracorporeal membrane oxygenation (ECMO) remains poorly defined. Drug sequestration within the ECMO circuit has been reported; however, existing evidence is primarily based on ex vivo studies and small clinical series, resulting in conflicting findings with limited clinical applicability. This study aimed to characterize antimicrobial sequestration within different ECMO circuit components and to describe the PK of meropenem, piperacillin/tazobactam, linezolid, and ceftaroline in critically ill patients receiving ECMO support. We conducted a prospective, single-center observational study in critically ill patients receiving ECMO (veno-venous (VV) or veno-arterial (VA)) and treated with meropenem, piperacillin/tazobactam, linezolid, or ceftaroline. Serial blood samples were collected simultaneously from the patient's arterial line and from pre- and post-membrane oxygenator sampling sites. Non-compartmental analysis was used to derive PK measures, and concentration differences across sampling sites were analyzed to estimate circuit-related drug sequestration. Antimicrobial target attainment was assessed according to established PKPD targets. A total of 237 samples from 8 patients (4 VV, 4 VA) were analyzed. Limited, component-specific sequestration was observed in the membrane oxygenator for meropenem (9.28% ± 21.6%, p = 0.046) and ceftaroline (14.6% ± 17.5%, p = 0.010), for linezolid it was also identified (10.1% ± 12.3%, p = 0.005), but in tubing and connectors. Piperacillin/tazobactam showed negligible retention. However, when considering the ECMO circuit as a whole, no statistically significant reductions in systemic concentrations were observed for any antimicrobial. PK measures were generally consistent across sampling sites. In 20% of treatments, predefined PKPD efficacy targets were not achieved. Antimicrobial sequestration within the ECMO circuit is detectable and drug- and component-specific. However, overall impact on systemic antimicrobial exposure appears limited. Inadequate PKPD target attainment in a subset of antimicrobial treatments likely reflects the marked PK variability of critical illness rather than ECMO-related drug loss alone. These exploratory results support the use of therapeutic drug monitoring to individualize antimicrobial therapy in ECMO-supported patients and the need to conduct further population PK studies to better characterize determinants of antimicrobial exposure and to clarify the clinical relevance of circuit-related drug sequestration.

Rivaroxaban, a factor Xa inhibitor, is used to prevent stroke in patients of atrial fibrillation (AF). This study aimed to establish a population pharmacokinetic (PPK) model for rivaroxaban using real-world data, apply it to concentration prediction, and investigate the association between drug exposure and clinical outcomes. Patients with AF receiving rivaroxaban were enrolled from an observational cohort (2016-2023) to measure plasma concentrations. An independent cohort was randomly selected to validate the PPK model. Clinical outcomes of interest included stroke or systemic thromboembolism, and major bleeding. A total of 226 patients contributed to 452 rivaroxaban concentration measurements. Rivaroxaban pharmacokinetics were adequately described using a one-compartment model with first-order elimination. The estimated apparent clearance (CL/F) and the volume of distribution (V/F) were 6.13 L/h and 45.57 L, respectively. CL/F was significantly influenced by creatinine clearance and concomitant use of cytochrome 3A4 or P-glycoprotein inhibitors, whereas V/F was associated with lean body weight. External validation demonstrated a good predictive performance at the individual level. Patients with low trough concentrations tended to have an increased risk of systemic thromboembolism, whereas those with high trough concentrations tended to have a higher risk of major bleeding. In an Asian population, rivaroxaban pharmacokinetics are influenced by renal function, lean body weight, and drug interactions. The developed PPK model facilitates the estimation of rivaroxaban concentrations at standardized timepoints from random samples. This provides a practical tool for standardized exposure assessment and the identification of patients at risk for adverse clinical outcomes. ClinicalTrials.gov identifier no. NCT05333666.

Mycophenolate mofetil (MMF) is widely used in lupus nephritis (LN), but interindividual variability in pharmacokinetics and pharmacogenetics may affect treatment response. This study investigated the relationship between MMF pharmacokinetics, key genetic polymorphisms, and therapeutic response in Egyptian female LN patients. In this prospective study, 53 female patients with active LN (ISN/RPS class III-IV) maintained on 2g/day MMF were enrolled at the Urology and Nephrology Center, Mansoura University, Egypt. Baseline clinical, laboratory, and histopathological data were collected. Serial blood samples were obtained over 8h for mycophenolic acid (MPA) quantification by LC/MS. Genomic DNA was extracted for UGT2B7 (rs7438135) and SLCO1B1 (rs183624077) genotyping using TaqMan real-time PCR assays. Treatment response was classified as complete or partial remission versus non-response at 6 months. Population pharmacokinetics were analyzed using Monolix 2020R1. Considerable variability in MPA exposure was observed (mean AUC₀-₈: 22.4 ± 12.9µg·h/mL; mean Cmax: 12.9 ± 5.5µg/mL). Responders exhibited higher median AUC₀-₈ than non-responders (23.1 vs. 16.2µg·h/mL; p = 0.039). SLCO1B1 and UGT2B7 genotypes did not differ significantly between responders and non-responders, though TT carriers showed lower MPA AUC₀-₈. Multivariate analysis identified biopsy class (III vs. IV), induction therapy with MMF, and higher MPA AUC₀-₈ as independent predictors of response. MMF exposure, biopsy class, and induction regimen significantly influence therapeutic response in LN. Pharmacokinetic monitoring of MPA may help optimize MMF therapy, while common UGT2B7 and SLCO1B1 variants showed limited predictive value in this cohort.

Ceftolozane/tazobactam (TOL/TAZ) and ceftazidime-avibactam (CAZ/AVI) are increasingly used to treat pneumonia caused by multidrug-resistant Gram-negative bacilli. However, data on intrapulmonary penetration to confirm dosing adequacy are limited. Therefore, the aim of this study was to compare epithelial lining fluid (ELF) penetration and pharmacokinetic/pharmacodynamic (PK/PD) target attainment of TOL/TAZ versus CAZ/AVI administered by continuous infusion (CI) in critically ill patients with nosocomial pneumonia. Single-center, open-label, randomized pharmacokinetic (PK) study. Thirty patients were randomized 1:1 to receive 6g/3g of TOL/TAZ or 6g/1.5g of CAZ/AVI administered by CI. A population PK model was constructed using plasma and ELF concentrations. Lung penetration was estimated based on the ratio AUC0-8ELF/AUC0-8plasma. Simulations were performed to estimate the probability of attaining predefined joint ELF PK/PD target defined as free concentration 100% fT> minimum inhibitory concentration (MIC) for the β-lactam component and 100% fT> concentration threshold (CT) for the β-lactamase inhibitors, among relevant MIC scenarios. Three different dosing regimens (low, standard and high) were evaluated for each combination. A total of 298 plasma and 58 ELF samples were analyzed. The median [IQR] age, body mass index, and creatinine clearance were 77 [9.8] years, 26.48 [5.5] kg/m², and 76.0 [96.0] mL/min, respectively. Median intrapulmonary penetration was 0.66 [0.32] for ceftolozane, 0.41 [0.30] for ceftazidime 0.44 [0.05] for tazobactam and 0.44 [0.46] for avibactam. Under the prespecified ELF PK/PD target, standard dosing achieved adequate target attainment for both combinations. Simulations showed that all CAZ/AVI regimens achieved high ELF target attainment under conservative inhibitor threshold assumptions (assuming ceftazidime MIC = 8mg/L and avibactam CT =1mg/L), whereas for TOL/TAZ (assuming ceftolozane MIC = 4mg/L and tazobactam CT 2mg/L) at least the standard-dose regimen was required. More aggressive target scenarios reduced the probability of target attainment, particularly for inhibitor thresholds of 4mg/L and for Enterobacterales-oriented joint targets. In critically ill adults with nosocomial pneumonia receiving TOL/TAZ or CAZ/AVI by CI, standard doses achieved ELF exposures consistent with dual PK/PD target attainment against all susceptible isolates. Interindividual variability and the risk of plasma overexposure support individualized dosing and consideration of therapeutic drug monitoring. The trial was registered in the European Union Drug Regulating Authorities Clinical Trials Database (EudraCT No. 2021-006908-32). Registered 10 February 2022; https://www.clinicaltrialsregister.eu/ctr-search/trial/2021-006908-32/ES.

BackgroundRemimazolam, an ultra-short-acting benzodiazepine, is characterised by rapid onset and recovery, metabolism independent of hepatic or renal function and haemostability. Despite the above pharmacokinetic (PK) advantages and potential benefits for renal- or hepatic-insufficient patients and the elderly patients, the effects and population pharmacokinetics–pharmacodynamics (popPK/PD) of remimazolam for general anaesthesia in elderly, hepatic or renal insufficient patients from large, randomised controlled trials remain unknown.Methods and analysisThis is a randomised controlled single-blind multicentre study. Patients undergoing selective surgeries (eg, abdominal, orthopaedic or urologic procedures) will be screened and randomised into the elderly (age ≥65 years), hepatic insufficiency (Child-Pugh class B or C) or renal insufficiency (estimated glomerular filtration rate <60 mL/min/1.73 m²) groups, with a target of 752 subjects per group. Patients in each group will randomly receive remimazolam or propofol for general anaesthesia induction and maintenance. The primary outcome is the quality of induction and maintenance of general anaesthesia (the time of maintaining bispectral index (BIS) 35–65 divided by the time for test-drug administration). Secondary outcomes are PK and PD characteristics of remimazolam in the target populations, which include: the incidence of hypotension; incidence and amount of vasoconstrictor agents; area under the curves of blood pressure over time; pain on injection; eye opening time; extubation time; time to reach modified Aldrete scores 9; nursing delirium screening scores and incidence of intraoperative awareness. Exploring outcomes are remimazolam besylate plasma concentrations; the ideal target remimazolam besylate plasma concentrations (plasma concentrations corresponding to a BIS value between 35 and 65) and the optimal dosage regimen based on the popPK/PD model.Ethics and disseminationThis study was approved by the Ethics Committee of West China Hospital, Sichuan University, on 30 August 2024, and registered at the Chinese Clinical Trial Registry. The findings of this study will be disseminated through presentations at relevant conferences and published in peer-reviewed journals.Trial registration numberChiCTR2400089033.

Sirolimus is currently used off-label for paediatric patients with vascular anomalies. However, the optimal dosage regimen for paediatric patients remains controversial. This study aimed to determine the optimal dosing regimen of sirolimus in these patients using a population pharmacokinetic (PK) model. Prospective study data (seven subjects) were used as the structural model without covariates, while the retrospective study data (21 subjects) were used as the structural model with size and maturation functions for evaluation. The model was assessed using general model-building criteria, and a bootstrap stepwise covariate modelling method was used to identify relevant covariates. Simulations were performed for various scenarios to determine the optimal dosing regimen for maintaining sirolimus trough concentrations within a specific range. NONMEM software (Version 7.5) with the first-order conditional estimation method with interaction was used. The final sirolimus model was developed using a two-compartment model, which was implemented by a maturation function using postmenstrual age (PMA), allometrically scaled body weight to account for size differences, and haemoglobin as a covariate on clearance. Simulations were conducted to propose an optimal dosing regimen. A once-daily dose of 0.01-0.1 mg/kg over 1 month maintained the trough concentration within the range of 5-15 ng/mL, taking into account patient-specific weight and the influence of haemoglobin constrained on PMA. The optimal dose varied by group, within the range of 0.06-0.09 mg/kg. We developed a robust and reliable PK model of sirolimus for paediatric patients with vascular anomalies. Paediatric optimal dosing regimens should be based on individual growth patients.

To determine midostaurin and posaconazole plasma concentrations and investigate adverse events (AEs) resembling drug-drug interactions (DDI) when both drugs were administered concomitantly during induction chemotherapy for acute myeloid leukemia (AML). Patients with FLT3-mutated AML who received midostaurin and posaconazole concomitantly between May 2019 and December 2022 were included and followed up to March 2023. Twice-weekly trough levels for midostaurin and posaconazole were measured with validated liquid chromatography-tandem mass spectrometry methods. Potential DDIs were independently reviewed by two physicians and attributed using the Drug Interaction Probability Scale (DIPS). Population pharmacokinetics analysis was done via nonlinear mixed-effect modeling. In 29 patients, concentrations ranged from 0.6 to 24.5 mg/L for midostaurin and from <30 to 2,572 µg/L for posaconazole. A total of 375 AEs in 66 midostaurin cycles, with 280 AEs classified as grade ≥3, were recorded. Probable DDI with a DIPS score of ≥5 was attributed in 14/375 AEs; no highly probable AEs were registered. Eight AEs led to dose modification or discontinuation of midostaurin in seven patients. Clearance for midostaurin during co-administration with posaconazole was 0.52 L/h (95% CI, 0.42-0.62 L/h). A breakthrough fungal infection was recorded in eight patients (27.5%). DDI of midostaurin and posaconazole is clinically meaningful but infrequent. High inter- and intra-individual variabilities of midostaurin and posaconazole plasma exposure were observed. Midostaurin clearance was delayed during co-administration. Midostaurin therapeutic drug monitoring may serve for decision-making when DDI with CYP3A4 inhibitors is suspected.

Firsekibart (formerly GenSci048 or genakumab) is a humanized anti-IL-1β monoclonal antibody developed for acute gouty flares in patients unsuitable for standard therapies. This study characterized its population pharmacokinetics (PopPK), assessing covariate effects, and established exposure-response (E-R) relationships to guide dosing. Data from four clinical trials involving 296 subjects (2287 concentrations) were pooled to develop a PopPK model using NONMEM with stepwise covariate screening. A one-compartment model with sequential zero- and first-order absorption was utilized. E-R analyses were conducted to assess efficacy (dVAS72h, defined as achieving ≥50% reduction in Visual Analog Scale [VAS] scores from baseline measured at 72 h) and safety (dyslipidaemia/hepatic dysfunction/infectious and invasive diseases) using logistic regression. Body weight significantly affected clearance and distribution volume. Hepatic or renal impairment caused exposure differences of less than 27%. No exposure-dependent safety trends were identified across the evaluated exposure range. Although exposure-response trends were observed across the broader dose range, the gradient within the exposure range achieved by 195-200 mg was shallow, consistent with a plateau. Simulations indicated that at the 200-mg dose, more than 95% of patients attained exposures associated with at least 85% probability of achieving dVAS72h. Firsekibart also markedly reduced flare recurrence over 12 weeks compared with the control group. Firsekibart demonstrated predictable pharmacokinetics, a favourable safety profile, and robust efficacy coverage at the 200-mg fixed dose. These findings support fixed-dosing for patients with acute gout flares who are intolerant of, unsuitable for, or inadequately responsive to standard-of-care therapies.

With the rising use of psychotropic drugs, reports of adverse events have increased accordingly. Therapeutic Drug Monitoring (TDM) can improve the clinical efficacy of psychotropic drugs. Nevertheless, bibliometric research in this field is scarce, and conducting such research would contribute to future work. The literature on TDM of psychotropic drugs published during 2005-2024 was obtained from the Web of Science Core Collection database. Bibliometric analysis and visualization were conducted by VOSviewer, CiteSpace, and Scimago Graphica. A total of 1615 articles were retrieved, showing an upward trend over the past two decades. The most influential country and institution were the United States and Johannes Gutenberg Univer-sity of Mainz. Keyword analysis showed that clozapine, carbamazepine, lamotrigine, risperidone, olanzapine, and valproic acid are the key psychotropic drugs in current TDM research. Meanwhile, mass spectrometry technology, genetic testing, and population pharmacokinetics have promoted the development of TDM. This study revealed that TDM is widely applied in psychotropic drugs, and TDM is necessary to achieve precision medicine. The emergence of mass spectrometry has greatly promoted the development of TDM, making its clinical application more efficient and accurate. Although this study only used a single database from Web of Science, our results could still provide valuable ref-erences for clinical individualized treatment. This study systematically analyzed the literature on the TDM of psychotropic drugs. These findings would provide valuable references for further development in this field.

PurposeVoriconazole (VCZ) exhibits nonlinear pharmacokinetics, a narrow therapeutic window, and substantial interindividual variability. Inaccurate dosing may lead to underexposure or overexposure, causing treatment failure or toxicity. Existing population pharmacokinetic (PPK)-machine learning (ML) models either lack mechanistic interpretability or inadequately characterize VCZ exposure. Therefore, we propose a hybrid model embedding ML within a PPK framework to associate clinical covariates with VCZ exposure.Patients and MethodsA total of 489 inpatients receiving VCZ at the Third Affiliated Hospital of Soochow University between March 2020 and May 2024 were included. We identified candidate predictors of CL/F using dual-feature selection with Boruta and LASSO. Overlapping features were used to train four ML algorithms to estimate CL/F. The predicted CL/F values were incorporated into a steady-state PPK equation to back-calculate VCZ concentrations, followed by quadratic calibration to reduce bias. Causal mediation analysis assessed pathways from key covariates to VCZ concentration via CL/F, and Shapley Additive exPlanations (SHAP) values were used to quantify feature contributions.ResultsUnder the PK-informed hybrid strategy, XGBoost achieved the best concentration prediction (R2 = 0.739, MAE = 0.357, RMSE = 0.526, MAPE = 7.78%), outperforming a direct ML approach treating PK-related variables as inputs (CatBoost: R2 = 0.459). The temporal external validation performance of the hybrid model remained stable (R2 = 0.661, MAE = 0.473, RMSE = 0.651, MAPE = 14.71%). Mediation analysis demonstrated that CRP affected VCZ exposure primarily through CL/F, whereas albumin and age acted as modifiers. A web-based calculator was developed for real-time individualized prediction and assistance with clinical-dose adjustment.ConclusionThe hybrid model improved VCZ concentration prediction versus direct ML modeling while preserving CL/F-centered mechanistic interpretability. It may help guide dose adjustment and reduce clinically relevant misdosing. This framework may be generalizable to other narrow-therapeutic-window drugs.

Population pharmacokinetic modeling-based human health risk assessment of di-isodecyl phthalate using biomonitoring-driven reverse dosimetry.

Recent studies have applied machine learning (ML)-based limited sampling strategies (LSS) to predict drug exposure (AUC), achieving low prediction error and performance comparable to or better than multiple linear regression and population pharmacokinetics LSS. This study aimed to develop and validate a machine learning-based limited sampling strategy capable of predicting raltegravir (RAL) exposure. Four machine learning algorithms (XGBoost, Random Forest, GLMNet, and SVM) were trained using pharmacokinetic profiles generated via Monte Carlo simulation from a population pharmacokinetic (POPPK) model. Data were divided into training (75%) and test (25%) sets. All possible combinations of sampling times, pairs and triplets, in steady-state, up to 12 h post-dose were evaluated. Model performance was assessed by the lowest root mean square error (RMSE) in the cross-validation, and the best performing model was evaluated in the test set and externally validated using simulated PK profiles from an independent POPPK model and patient data from a clinical study. XGBoost trained with concentrations at 0.5, 2, and 4 h showed the best predictive performance. The model achieved excellent accuracy in the test set (bias/RMSE: 0.8%/8.7%) and in the independent simulation (1.9%/14.3%). Performance decreased in real patient data (5.0%/24.1%), highlighting the need for caution when extrapolating predictions to populations whose characteristics differ from those represented in the training datasets. A machine learning model using only three sampling timepoints has been developed and validated in different datasets, enabling accurate estimation of RAL AUC₀-₁₂. This approach provides a tool for pharmacokinetic and PK/PD studies and reduces intensive sampling need in clinical settings.