Evidence on the efficacy and safety of medicines supporting marketing authorization is largely derived from randomized controlled trials, yet these trials often exclude populations likely to use the medications in clinical practice. This study assessed eligibility of special populations, defined as pregnant and breast-feeding individuals, elderly individuals, and individuals with comorbidities, in pivotal clinical trials supporting centrally authorized products in the European Union between 2022 and 2024. It also examined whether exclusions of the special populations were supported by scientific justifications in the form of prior clinical or nonclinical evidence documented in regulatory sources. Among 24 products supported by 44 pivotal trials, pregnant and breast-feeding individuals were most frequently excluded (16/24 products), followed by individuals with cardiovascular disease or hepatic impairment (13/24 products) and renal impairment (10/24 products). Exclusion of elderly individuals based on age alone was uncommon (1/24 products). Of the 24 products, 15 excluded more than one special population. While exclusions of pregnant and breast-feeding individuals were largely supported by preclinical evidence suggesting safety concerns, those related to comorbidities were often not accompanied by prior evidence justifying exclusions. Broad exclusion criteria based on investigator judgment were also common. These findings reveal variation in how exclusion criteria are justified across special populations in regulatory documentation and suggest an opportunity for greater transparency and consistency in documenting the scientific and ethical justification for these criteria in clinical trials.
Zavegepant nasal spray, a calcitonin gene-related peptide receptor antagonist indicated for acute treatment of migraine with or without aura in adults, is a substrate for organic anion transporting polypeptide 1B3 (OATP1B3), Na+/taurocholate cotransporting polypeptide (NTCP), P-glycoprotein (P-gp), and cytochrome P450 3A4 (CYP3A4). In clinical drug-drug interaction (DDI) studies, oral zavegepant exposure increased in the presence of multiple-dose (MD) rifampin or itraconazole; however, no interaction was observed between intranasal (IN) zavegepant and itraconazole. A physiologically based pharmacokinetic (PBPK) model was developed and validated using in vitro, human mass balance, and observed clinical zavegepant pharmacokinetic data and was employed to assess clinical DDI mechanisms and further support zavegepant nasal spray labeling recommendations. PBPK modeling indicated that DDI observed between itraconazole and oral zavegepant is primarily driven by intestinal P-gp inhibition, and observed DDI with MD rifampin primarily results from hepatic uptake inhibition. Modeling predicted an increase in plasma exposure (AUC ratio [AUCR] = 2.39) of IN zavegepant with single-dose (SD) rifampin; IN zavegepant plasma exposure was also predicted to increase with MD rifampin (AUCR = 2.11). Predicted IN zavegepant exposure change with SD cyclosporine A was 1.58-fold (AUCR) when assuming 90:10 proportion for OATP1B3:NTCP contribution to hepatic uptake. CYP3A inducers, carbamazepine and efavirenz, were predicted to have minimal effect on IN zavegepant PK. Overall, PBPK modeling and observed PK data support that CYP3A/P-gp inhibitors and inducers do not produce clinically relevant modulation in IN zavegepant plasma exposure; however, OATP1B3/NTCP inhibitors can significantly increase IN zavegepant exposure, and coadministration should be avoided.
Machine learning (ML) is increasingly utilized in medical research. However, underrecognized methodological pitfalls can result in misleading conclusions, which if acted upon can result in patient mismanagement and harm. As a result, we have provided a compilation of common pitfalls and practical safeguard considerations for ML-enabled medical research. Examples emphasize the importance of integrating clinical/biologic knowledge and statistical/epidemiologic principles, illustrated with key case examples. Topics include (1) Data drift, which can create artificial trends when coding systems or clinical practices change over time; (2) Feature construction that should be clinically grounded, as models may otherwise learn operational proxies rather than true biology; (3) Accidental data leakage, which can inflate apparent model performance and collapse on real-world deployment; (4) Competing risks and differential time-at-risk that can generate spurious associations in safety analyses if not modeled explicitly; (5) Confounding and collider bias, which can distort findings, particularly in EHR-based studies; (6) For rare outcomes, inappropriate evaluation measures can overstate performance and mislead explainability analyses; (7) Finally, pooled multi-trial analyses require careful handling of the intrinsic heterogeneity to avoid extracting study artifacts rather than generalizable signals. By outlining many of the common potential issues, with lessons learned, we aim to provide the medical and research community with practical insights to help foster trustworthy ML-enabled medical research.
Advances in biologics and model-informed drug discovery and development (MID3) are transforming the treatment of immune-mediated inflammatory diseases (IMIDs). By integrating pharmacokinetics and pharmacodynamics, MID3 enables mechanism-based, patient-centered strategies to optimize therapies, strengthen benefit-risk assessment, and advance precision medicine. Innovations such as cytokine inhibitors, bispecific antibodies, antibody-drug conjugates, and mRNA therapies have expanded treatment options across dermatology, rheumatology, gastroenterology, and respiratory medicine. The rapid development of JAK- and interleukin-targeted agents underscores the shift toward individualized, mechanistically driven therapies. At ASCPT 2025, the session "Transformative Advancements in the Treatment of IMIDs: Integrating Patient-Centric Clinical Pharmacology with Translational MID3 for Biologics" brought together experts from academia, industry, and regulatory agencies and emphasized how translational MID3 and precision pharmacology are accelerating the next generation of patient-tailored biologic therapies, with case studies in inflammatory bowel disease, systemic lupus erythematosus, and rheumatology.
Drug safety in pediatric patients has always been a concern. Major tragedies with drugs in pediatric patients created an environment of caution, but the experience with pediatric legislation and drug development over the past 20 years has eased the cautionary approach to pediatric drug safety. The advent of new methods of safety testing, driven by the movement away from animal use in research, provides new opportunities to screen for and prevent adverse developmental effects of drugs on pediatric patients. One approach that can be explored for use in pediatric developmental safety testing is secondary pharmacology, or screening for off-target drug effects. Secondary pharmacology, which is an established screening mechanism described in ICH S7A, could provide the opportunity to answer important off-target questions. Examples of 12 specific secondary pharmacology targets related to adverse effects on pediatric development are provided. Clinical pharmacology plays a critical role in the interpretation of secondary pharmacology through quantitative studies necessary to validate positive findings and through the final assessment of the risk:benefit associated with these findings. In conclusion, secondary pharmacology screening has the potential to reduce animal testing while addressing two critical objectives: screening drugs for off-target effects and providing information on off-target effects relevant to pediatric developmental safety.
Bemnifosbuvir (BEM) and ruzasvir (RZR) are pan-genotypic HCV polymerase (NS5B) and NS5A inhibitors, respectively. In a single-arm study (NCT05904470), 215 treatment-naïve, adherent participants with chronic HCV (with or without cirrhosis) received BEM 550 mg QD plus RZR 180 mg QD for 8 weeks. We employed a multiscale mathematical model fitted to the HCV RNA and ALT dynamics to quantify the antiviral activity and evaluate the modes of action of BEM and RZR. We found that after treatment initiation, the viral load (VL) decreased in a triphasic manner, with a very rapid first phase lasting <12 h. The population estimate of the time to reach the LLOQ (VL ≤ 15 IU/mL) was 11 days for individuals in fibrosis stages F0-F3 and 15.6 days for those in F4, while the estimate of the time to reach a hypothetical cure boundary (VL ≤6.7 × 10-4 IU/mL) was 6.5 weeks for F0-F3 and 6.9 weeks for F4, with some variability between individuals. The population estimate of the effectiveness of the 8 weeks of treatment in blocking viral replication was 98.6% for F0-F3 and 99.4% for F4, while the estimate of the effectiveness in blocking viral assembly/secretion was 99.8% for F0-F3 and 99.0% for F4. We also found a treatment-induced 70% enhancement of the rate of intracellular HCV RNA degradation. The estimated efficacy of therapy was similar across genotypes 1 to 4. Thus, our modeling supports the conclusions that BEM/RZR is pan-genotypic, has high antiviral effectiveness and leads to rapid viral clearance in both cirrhotic and non-cirrhotic patients.
New Approach Methodologies (NAMs) represent a paradigm shift in drug development and regulatory science, offering human-relevant alternatives to traditional preclinical models. This mini-review highlights recent advances in NAM development, validation, and regulatory application from FDA's Division of Applied Regulatory Science (DARS). We describe in silico NAMs including quantitative systems pharmacology (QSP) modeling for opioid overdose scenarios, (quantitative) structure-activity relationship ((Q)SAR) modeling for toxicity prediction and drug-drug interaction assessment, and multi-omics approaches for pharmacodynamic biomarker discovery. Additionally, we present in vitro NAMs including human-induced pluripotent stem cell (hiPSC)-derived neural networks for opioid pharmacology assessment, microphysiological systems (MPS) for pulmonary drug permeability, gastrointestinal models for intestinal transport, and standardized cardiac ion channel and hiPSC-CM MEA assays for proarrhythmia risk prediction. These case studies demonstrate how NAMs can reduce reliance on animal studies while improving translation to clinical outcomes. The NAMs described here range from exploratory research tools to approaches actively informing regulatory submissions review. The integration of in silico and in vitro NAMs into regulatory decision-making frameworks represents a critical step toward more efficient, human-relevant drug development and safety assessments.
Solute carriers (SLC) and ATP-binding cassette (ABC) transporters are essential for placental solute exchange and fetal protection, yet their transcriptomic profiles in the human placenta remain poorly characterized. Although fetal sex influences placental development and function, its impact on transporter expression is unclear. Using RNA sequencing, we profiled SLC and ABC transporter expression in two anatomical regions of term human placentas (N = 10) with balanced fetal sex distribution and controlled clinical features. In the intervillous region, 276 SLCs were detected, with 60% expressed at low levels (CPM < 25) and 110 (40%) expressed at higher levels (CPM 25-1650). Four transporters-SLC2A1 (GLUT1), SLC44A2 (CTL2), SLC38A2 (SNAT2), and SLC38A1 (SNAT1)-showed the highest expression (CPM > 500). Functional annotation of 110 SLCs revealed transporters for amino acids, metals, and inorganic ions each accounted for ~9%. Other well-represented SLCs included xenobiotic and vitamin transporters. Mitochondrial, lysosomal, and endoplasmic reticulum and Golgi transporters together accounted for ~37%, while 11% were orphan transporters. Thirty-seven ABC transporter transcripts were detected, primarily associated with xenobiotic and lipid transport. Compared with the intervillous region, 1676 genes-including 32 SLC and 2 ABC transporters-were differentially expressed in the decidual region. Although 117 genes in the intervillous region and 79 genes in the decidual region exhibited sexually dimorphic expression, none were SLCs. Notably, male placentas showed higher expression of ABCB1 (P-gp), confirmed by qPCR and western blot. Our findings provide a transcriptomic map of placental transporters and highlight a potential sex-specific difference on ABCB1 expression and xenobiotic protection.
The extent of pharmacogenetic (PGx) drug dispensing among Dutch adults receiving medications for cardiovascular disease (CVD) is unknown. Using the University of Groningen IADB.nl pharmacy database, we performed a serial cross-sectional study (2019-2023) to estimate the annual prevalence of PGx drug dispensing and annual rates of initiation. We also identified the most dispensed PGx drug classes, most frequently associated genes, and the proportion of individuals with potential opportunities for reuse of PGx testing results. Adults on CVD medication treatment were defined as those with ≥2 prescriptions for the same CVD medication (ATC classes B01, C01, C03, C07-C10) within 180 days in a calendar year. PGx drugs were defined as drugs with actionable drug-gene interactions according to international CPIC or national DPWG guidelines. The annual cohort size ranged from 47,602 in 2019 to 41,846 in 2023 (mean age ~70 years; ~48% male). The prevalence of ≥1 PGx drug dispensing (~90%) and multiple (≥2) PGx drug dispensing (~68%) remained stable across 5 years, both increasing significantly with age. CVD-related PGx drugs were most common (75%), followed by proton pump inhibitors (55%). In 2023, CYP2C19, CYP2D6, and SLCO1B1-associated drugs had most users (60%, 53%, and 51%). Among those receiving drugs associated with CYP2C19, CYP2D6, and SLCO1B1, 24%, 23%, and 4% would have opportunities for reuse of testing results with single-gene testing, and 66%, 65%, and 83% with panel testing. In conclusion, PGx drug dispensing was highly prevalent among people receiving CVD medications, warranting further research into the (cost-)effectiveness of PGx testing.
Evidence generation for antihyperglycemic therapies in pediatric type 2 diabetes mellitus (T2DM) remains challenging, with most randomized trials failing to demonstrate statistically significant reductions in glycated hemoglobin (HbA1c), partly due to substantial between-subject variability. To better understand the sources and implications of this variability, we conducted a model-based longitudinal meta-analysis of individual participant data from seven pediatric T2DM clinical trials submitted to the European Medicines Agency in support of marketing authorization applications. The analysis incorporated 3295 HbA1c observations from 809 participants and quantified the contributions of baseline disease severity, disease progression, placebo response, and treatment effects to longitudinal HbA1c dynamics. Insulin use and longer duration of diabetes were associated with higher baseline HbA1c and/or faster disease progression, while the placebo effect was substantial and highly variable across participants. Clinical trial simulations using the final model demonstrated that trials with fewer than 200 participants are unlikely to achieve 80% power to detect placebo-corrected treatment effects smaller than -0.70%-points over 26 weeks. Enrichment strategies targeting patients with lower variability, as well as model-based estimators-including Bayesian approaches leveraging historical information-substantially reduced the required sample size. These findings indicate that current pediatric T2DM trials are typically underpowered due to underestimation of HbA1c variability and highlight opportunities for more efficient trial designs. By quantifying disease- and treatment-related drivers of HbA1c trajectories and demonstrating the potential of model-informed strategies to improve power, this work provides a framework to enhance pediatric T2DM drug development and support regulatory decision-making.
Agranulocytosis is a rare but serious side effect of antithyroid drugs (ATDs), defined by an absolute neutrophil count (ANC) below 500/μL; some patients develop a more severe form with ANC below 100/μL. We retrospectively compared the clinical, biochemical, and HLA-B allele profiles between the two presentations in 24 cases of ATD-induced agranulocytosis. Based on nadir ANC, 15 participants with ANC < 100/μL formed the severe group and 9 with ANC 100-500/μL comprised the typical group. The severe group had a significantly longer neutrophil recovery time (5 [5-7] vs. 3 [2.5-3.5] days; P = 0.003), higher median levels of total bilirubin (21.4 [13.3-30.4] vs. 7.1 [6.2-18.3] μmol/L; P = 0.003) and indirect bilirubin (9.9 [7.6-16.7] vs. 6.1 [3.6-9.8] μmol/L; P = 0.018), and fewer asymptomatic patients at presentation (0/15 vs. 5/9, P = 0.003). HLA typing confirmed that HLA-B*38:02 was associated with severe ATD-induced agranulocytosis (OR = 48.6; 95% CI = 2.3-1020.7; P = 5.9 × 10-4). Sensitivity analysis excluding the 2 PTU-treated patients confirmed a robust and significant association (OR = 38.3; 95% CI = 1.8-820.2; P = 0.0039). No significant differences were observed in age, sex, drug type or dosage, or other biochemical parameters at agranulocytosis onset. Severe agranulocytosis accounted for 62.5% (15/24) of cases, highlighting it as a common clinical phenotype of ATD-induced agranulocytosis. HLA-B*38:02 is associated with susceptibility to ATD-induced severe agranulocytosis, an association predominantly driven by methimazole exposure.
Host metabolic changes in response to drug treatment may shape individual pharmacological outcomes. This study aims to identify serum metabolomics signatures potentially associated with individual responses to enzalutamide treatment in metastatic castrate-resistant prostate cancer (mCRPC) patients. Targeted metabolomic profiles in serum collected from 18 mCRPC patients treated with 160 mg/day of enzalutamide were measured at baseline, after 3 months of treatment, and until disease progression by a validated liquid chromatography tandem mass spectrometry analytical platform. The difference in the serum metabolomic profiles during enzalutamide treatment between good and poor responders was investigated by both multivariate and univariate statistical analysis. Primary mouse hepatocytes and Caco-2 cell models were employed to gain mechanistic insights into the most relevant changes detected. Enzalutamide induced a significant and sustained increase in several amino acids, with taurine displaying the most significant change. The in vitro data suggest that enzalutamide-induced elevation of systemic taurine results from both increased intestinal absorption and decreased hepatic extraction. Finally, serum metabolomic changes observed after 3 months of treatment were found to be significantly associated with the pharmacological response. This hypothesis-generating study reveals that enzalutamide significantly alters the host serum metabolome, inducing changes that correlate with therapeutic response over time. These results highlight the importance of host-drug metabolic interactions during androgen receptor inhibitor treatment and, although further validation is required, they support the hypothesis that metabolic profiling could eventually aid in the personalized management of enzalutamide therapy.
The comprehensive understanding of triplet repeat expansion (TRE) disorders as monogenic neurodegenerative diseases presents a fundamental challenge for developing treatments capable of modifying the disease course, as neurogenetic research indicates. The past 30 years reveal that preclinical results demonstrating strong effectiveness have not translated into success in human clinical trials, with multiple phase II and III studies for polyglutamine and non-coding, loss-of-function repeat disorders yielding unfavorable outcomes. This review emphasizes that clinical systems, which measure target engagement do not generate effective outcomes for these TRE neurodegenerative conditions. We highlight that these failures arise not from a fundamental misunderstanding of the underlying genetic mutations, but from recurring methodological and pharmacological limitations across the translational pipeline. The main reasons for these unsuccessful clinical trials include two factors: the necessity for clinicians to monitor pharmacodynamic biomarkers indicating central nervous system target engagement, and the reliance on traditional clinical rating scales that linearly assess disease progression in patients with complex medical conditions. The use of non-selective gene silencing methods can lead to unexpected toxic effects, especially when treatments begin after patients reach an advanced disease stage, causing irreversible brain cell damage. The field of precision neurogenetics must advocate for future trials to employ objective fluid and digital biomarkers, utilize allele-specific treatments, and prioritize patients who have not yet experienced severe brain cell loss. We further propose evidence-based design principles to improve the probability of clinical success in future TRE disease trials.
Tramadol is frequently co-prescribed with oxycodone or hydrocodone after surgery under the assumption that its dual mechanism of action provides an opioid-sparing benefit. However, evidence supporting this practice is limited and inconsistent. We conducted a secondary analysis of the multicenter IGNITE ADOPT-PGx pragmatic trial of patients undergoing elective surgery who were prescribed oxycodone alone (n = 317) versus tramadol + oxycodone (n = 96), or hydrocodone alone (n = 563) versus tramadol + hydrocodone (n = 42). Co-primary outcomes of this study were cumulative morphine milligram equivalents (MME) consumed during the 10 days following surgery and composite pain scores at day 10 defined as the sum of current pain and average and worst pain in the past 7 days (score ranges from 3 to 15). Inverse probability of treatment weighting (IPTW) was used to adjust for demographic characteristics, comorbidities, and concomitant analgesics. Sensitivity analyses restricted to CYP2D6 normal metabolizers and sites prescribing both monotherapy and combination therapy and subgroup analysis including total knee arthroplasty patients were conducted. Compared to oxycodone alone, the combination of oxycodone and tramadol was associated with higher MME (adjusted mean 205.2 vs 122.5; P < 0.0001) and lower mobility, without meaningful differences in pain scores (adjusted mean 9.6 vs 9.4; P = 0.41). Similar findings were observed for hydrocodone-only vs hydrocodone + tramadol groups. Results were consistent in sensitivity and subgroup analyses. Co-administration of tramadol with oxycodone or hydrocodone was associated with substantially higher total MME and reduced mobility without any demonstrable improvement in pain control.
Low-dose aspirin is no longer routinely recommended for the primary prevention of cardiovascular disease in older adults due to a lack of net benefit over bleeding risk. We hypothesized that genetic subgroups may experience differential harm or benefit from aspirin therapy. To investigate this, we screened 572 polygenic scores (PGSs) for modification of aspirin's effect on major bleeding and major adverse cardiovascular events (MACE) in the Aspirin in Reducing Events in the Elderly (ASPREE) randomized, placebo-controlled trial of daily 100 mg aspirin. Participants were aged ≥70 years (≥65 years for US minorities) and free of cardiovascular disease, dementia, or physical disability at enrolment. Among participants with high-quality genotyping data (n = 13,571), PGS-aspirin interactions were tested using Cox proportional hazards models with Bonferroni correction for multiple testing. During a median follow-up of 4.6 years, 414 major bleeding events and 464 MACE occurred. A triglyceride-related PGS (PGS003144) significantly modified aspirin-associated bleeding (interaction P = 5.9 × 10-5; Bonferroni-adjusted P = 0.034). In the lowest quintile of the PGS distribution, aspirin increased major bleeding risk (HR 2.28; 95% CI: 1.45-3.58; P = 0.00036), including separate bleeding subgroups gastrointestinal (HR 3.17; 95% CI: 1.48-6.80; P = 0.0029), and intracranial bleeding (HR 4.10; 95% CI: 1.54-11.0; P = 0.0049). In contrast, in the highest quintile, aspirin was associated with lower risk of bleeding (HR 0.62; 95% CI 0.38-0.97) and reduced MACE (HR 0.66; 95% CI 0.44-0.99). Baseline serum triglycerides showed similar effect modification. These hypothesis-generating findings suggest triglyceride-related genetic variation may identify individuals with differential responses to aspirin.
Rifampicin is a prototypical clinical inducer used in drug-drug interaction (DDI) studies. However, the clinical relevance and predictability of rifampicin-mediated hepatic transporter induction remain poorly defined. Conventional hepatocyte culture models fail to capture clinically relevant regulation of transporters and drug-metabolizing enzymes (DMEs), particularly at the protein level. In this study, we characterized a liver tissue chip (LTC) as a new approach methodology (NAM) for long-term transporter- and DME-mediated induction risk assessment. Transporter-certified® cryopreserved primary human hepatocytes from three donors were cultured in the LTC under recirculating flow with or without 10 μM rifampicin for 7 days. The LTC facilitated simultaneous mRNA-protein quantification for transporters and mRNA-protein-activity profiling for cytochrome P450s (CYPs) within individual chips across treatment and control conditions. In addition, extracellular bile acid profiling and intracellular rifampicin concentrations were measured by LC-MS/MS. Rifampicin induced multiple transporters and enzymes at the protein level, including organic anion transporting polypeptide (OATP)2B1, P-glycoprotein, CYP3A4, CYP2C19, and uridine 5'-diphospho-glucuronosyltransferase (UGT)1A4. Induction patterns differed between transporter mRNA and protein in many cases, suggesting regulatory mechanisms beyond PXR-mediated activation. Matched mRNA-protein-activity measurements of five CYP enzymes also revealed discrepancies in rifampicin-mediated induction magnitude. Rifampicin altered bile acid composition, increasing taurine-conjugated bile acids and inducing bile acid-related genes. Measured intracellular rifampicin concentrations differed substantially from experimentally employed media concentrations. Overall, these findings demonstrate that the LTC enables integrated assessment of mRNA, protein, and activity measurements in addition to bile acid profiles and intracellular drug exposure, which may improve mechanistic understanding of induction processes.
Cofetuzumab pelidotin (Cofe-P) is an anti-protein tyrosine kinase 7 (PTK7) antibody-drug conjugate (ADC) being investigated for the treatment of adults with PTK7-expressing, recurrent non-small cell lung cancer (NSCLC). The first-in-human (FIH) trial demonstrated tolerability of dosages ≤ 2.8 mg/kg once every 3 weeks (Q3W) and ≤ 3.2 mg/kg Q2W. Population pharmacokinetics (popPK) and exposure-response analyses were conducted to characterize Cofe-P and unconjugated auristatin (Aur0101) payload pharmacokinetics and exposure-efficacy (objective response rate (ORR)) and exposure-safety (Grade ≥ 3 neutropenia, Grade ≥ 2 peripheral neuropathy or rash) endpoints to inform the selection of dosages for optimization. Data from the FIH (0.2-3.7 mg/kg Q3W, 2.1-3.2 mg/kg Q2W) and Phase 1b (2.8 mg/kg Q3W) studies were used for popPK and exposure-response analyses. PopPK models adequately described pharmacokinetic data. Exposure-response analyses showed that higher Cofe-P (ADC) exposures increased the probability of achieving an objective response but also the occurrence of examined adverse events, particularly Grade ≥ 3 neutropenia. In non-squamous epidermal growth factor receptor wild-type NSCLC patients (≥ 90% with PTK7 2+ staining), analyses predicted that 3.2 mg/kg vs. 2.8 mg/kg Q3W dosing led to a higher ORR (35% vs. 21%) while maintaining a manageable safety profile (Grade ≥ 3 neutropenia: ≤ 39%). Predicted ORR and Grade ≥ 3 neutropenia are improvements over the standard-of-care, docetaxel. Dosing regimens < 2.4 mg/kg were not predicted to improve efficacy. Mean relative dose intensity was ≥ 86% across all Cofe-P doses and schedules tested. Together, popPK and exposure-response analyses support optimal dosing regimens of 2.4, 2.8, and 3.2 mg/kg Cofe-P Q3W in patients with NSCLC.
Regulatory bodies play a central role in providing guidance that enables safe and effective use of artificial intelligence tools in medicine development and evaluation. Regulators can also act as catalysts for regulatory science research. To inform these efforts, a European-wide survey was conducted to solicit stakeholder perspectives on the priority areas for regulatory science research related to the use of artificial intelligence in the medicine lifecycle. Twenty-eight regulatory science research questions were developed covering seven thematic domains: (1) research integrity and intellectual property; (2) accuracy and reliability of AI tools; (3) data governance, confidentiality, and consent; (4) regulation and oversight; (5) ethics, fairness, and bias prevention; (6) resources and support for AI use; and (7) impact on jobs and skills. Within each domain, stakeholders ranked four research challenges. A total of 273 responses were collected from regulators, pharmaceutical industry professionals, patients and consumers, academics, and healthcare professionals. Rankings of research challenges frequently converged across stakeholder groups and levels of AI experience. The top-ranked research questions within each domain were weighted according to the overall importance ranking of each domain to identify a list of 10 priority areas of research. The majority of the 10 priority areas fell within the domains of "Accuracy and reliability of AI tools," "Data governance, confidentiality, and consent," and "Ethics, fairness, and bias prevention." This list aims to support researchers and research funding bodies in addressing knowledge gaps on artificial intelligence in the medicines lifecycle.
Exposure-response (E-R) analyses of sunitinib in metastatic renal cell carcinoma (mRCC) have supported concentration-guided dose escalation. However, such analyses may be biased when exposure metrics incorporate post-baseline dose modifications, time-varying apparent clearance, or baseline factors that influence both pharmacokinetics and clinical outcomes. We re-evaluated the E-R relationship of sunitinib in mRCC with explicit assessment of these biases. A population pharmacokinetic analysis of 294 patients from four clinical trials identified time-dependent declines in apparent clearance for sunitinib and SU12662. E-R analyses were performed in 165 cytokine-refractory patients treated with the approved 50 mg 4-weeks-on/2-weeks-off regimen, using time to progression (TTP), overall survival (OS), and tolerability-related outcomes. Baseline albumin and age were identified as shared covariates of exposure and outcome and incorporated as confounders in multivariable multistate survival models. The apparent positive association between higher exposure and improved TTP and OS was attenuated when exposure was defined using early, baseline-anchored metrics rather than time-averaged apparent clearance. It was no longer evident after adjustment for albumin and age. Consistently, simulations with confounders fixed at cohort medians showed overlapping TTP and OS profiles across exposure groups. In contrast, higher exposure remained associated with poorer tolerability, including increased risks of dose reduction and treatment discontinuation driven by adverse-events. These findings indicate that previously reported positive E-R relationships for sunitinib were largely due to time-dependent bias and confounding. The results do not support exposure-guided dose escalation based on total plasma concentrations, but instead suggest the use of therapeutic drug monitoring to identify patients at risk of excessive exposure and toxicity.
