Applications of Model-Based Meta-Analysis in Drug Development

Application and Quality of Model-Based Meta-Analysis in Pharmaceutical Research: A Systematic Cross-Sectional Analysis and Practical Considerations

Model-based meta-analysis (MBMA) can be used in assisting drug development and optimizing treatment in clinical practice, potentially reducing costs and accelerating drug approval. We aimed to assess the application and quality of MBMA studies. We searched multiple databases to identify MBMA in pharmaceutical research. Eligible MBMA should incorporate pharmacological concepts to construct mathematical models and quantitatively examine and/or predict drug effects. Relevant information was summarized to provide an overview of the application of MBMA. We used AMSTAR-2 and PRISMA 2020 checklists to evaluate the methodological and reporting quality of included MBMA, respectively. A total of 143 MBMA studies were identified. MBMA was increasingly used over time for one or more areas: drug discovery and translational research (n = 8, 5.6%), drug development decision making (n = 42, 29.4%), optimization of clinical trial design (n = 46, 32.2%), medication in special populations (n = 15, 10.5%), and rationality and safety of drug use (n = 71, 49.7%). The included MBMA covered 17 disease areas, with the top three being nervous system diseases (n = 19, 13.2%), endocrine/nutritional/metabolic diseases (n = 17, 11.8%), and neoplasms (n = 16, 11.1%). Of these MBMA studies, 138 (96.5%) were rated as very low quality. The average rate of compliance with PRISMA was only 51.4%. Our findings suggested that MBMA was mainly used to evaluate the efficacy and safety of drugs, with a focus on chronic diseases. The methodological and reporting quality of MBMA should be further improved. Given AMSTAR-2 and PRISMA checklists were not specifically designed for MBMA, adapted assessment checklists for MBMA should be warranted.

New Horizons of Model Informed Drug Development in Rare Diseases Drug Development

Model-informed approaches provide a quantitative framework to integrate all available nonclinical and clinical data, thus furnishing a totality of evidence approach to drug development and regulatory evaluation. Maximizing the use of all available data and information about the drug enables a more robust characterization of the risk-benefit profile and reduces uncertainty in both technical and regulatory success. This offers the potential to transform rare diseases drug development, where conducting large well-controlled clinical trials is impractical and/or unethical due to a small patient population, a significant portion of which could be children. Additionally, the totality of evidence generated by model-informed approaches can provide confirmatory evidence for regulatory approval without the need for additional clinical data. In the article, applications of novel quantitative approaches such as quantitative systems pharmacology, disease progression modeling, artificial intelligence, machine learning, modeling of real-world data using model-based meta-analysis and strategies such as external control and patient-reported outcomes as well as clinical trial simulations to optimize trials and sample collection are discussed. Specific case studies of these modeling approaches in rare diseases are provided to showcase applications in drug development and regulatory review. Finally, perspectives are shared on the future state of these modeling approaches in rare diseases drug development along with challenges and opportunities for incorporating such tools in the rational development of drug products.

Time-course and dose-effect of omalizumab in treating chronic idiopathic urticaria/chronic spontaneous urticaria

PURPOSE

Several studies have shown that subcutaneous injections of omalizumab can treat chronic idiopathic/spontaneous urticaria (CIU/CSU) patients by only assessing the efficacy on specific endpoints. This study aimed to quantitatively analyze different doses of omalizumab in CIU/CSU and compare it with ligelizumab.

METHODS

Literature searches were performed in PubMed, Embase, and Web of Science databases. A model-based meta-analysis (MBMA) was utilized to develop a model incorporating time since the initiation of treatment and dose for omalizumab, with the change from baseline in Urticaria Activity Score (CFB-UAS7) as the primary efficacy endpoint. The time-course and dose-effect relationship throughout the omalizumab treatment period was analyzed, and the findings were compared with those of the investigational ligelizumab.

RESULTS

The model equation for the CFB-UAS7 was established as E = -Emax × time/(ET50 + time) × (b0 + b1 × dose). The estimated values of the model parameters E max ,ET 50 , b 0 , and b 1 were -1.16, 1.26 weeks, -9.90, and -0.0361 mg-1, respectively. At week 12 after the first dose, the model-predicted CFB-UAS7 for 150 mg and 300 mg of omalizumab were -16.0 (95% CI, -17.2 to -14.8) and -21.7 (95% CI, -22.9 to -20.5), respectively. In the PEARL-1 trial, the CFB-UAS7 for 72 mg and 120 mg of ligelizumab were -19.4 (95% CI, -20.7 to -18.1) and -19.3 (95% CI, -20.6 to -18.0), respectively. In the PEARL-2 trial, these values were -19.2 (95% CI, -20.5 to -17.9) and -20.3 (95% CI, -21.6 to -19.0), respectively.

CONCLUSION

Omalizumab showed a significant dose-dependent effect in the treatment of CSU. Both 72 mg and 120 mg ligelizumab might have the potential to outperform 150 mg (but not 300 mg) omalizumab.

Factors affecting the effectiveness and safety of polymyxin B in the treatment of Gram-negative bacterial infections: A meta-analysis of 96 articles

PURPOSE

Polymyxin B, with its unique structure and mechanism of action, has emerged as a key therapeutic agent against Gram-negative bacteria. The study aims to explore potential factors to influence its effectiveness and safety.

METHODS

A model-based meta-analysis of 96 articles was conducted, focusing on factors like dosage, bacterial species, and combined antibiotic therapy. The analysis evaluated mortality rates and incidence rate of renal dysfunction, also employing parametric survival models to assess 30-d survival rates.

RESULTS

In the study involving 96 articles and 9716 patients, polymyxin B's daily dose showed minimal effect on overall mortality, with high-dose group mortality at 33.57% (95% confidence intervals [CI]: 29.15-38.00) compared to the low-dose group at 35.44% (95% CI: 28.99-41.88), P = 0.64. Mortality significantly varied by bacterial species, with Pseudomonas aeruginosa infections at 58.50% (95% CI: 55.42-63.58). Monotherapy exhibited the highest mortality at 40.25% (95% CI: 34.75-45.76), P < 0.01. Renal dysfunction was more common in high-dose patients at 29.75% (95% CI: 28.52-30.98), with no significant difference across antibiotic regimens, P = 0.54. The 30-d overall survival rate for monotherapy therapy was 63.6% (95% CI: 59.3-67.5) and 70.2% (95% CI: 64.4-76.2) for association therapy with β-lactam drugs.

CONCLUSIONS

The dosage of polymyxin B does not significantly change death rates, but its effectiveness varies based on the bacterial infection. Certain bacteria like P. aeruginosa are associated with higher mortality. Combining polymyxin B with other antibiotics, especially β-lactam drugs, improves survival rates. Side effects depend on the dose, with lower doses being safer. These findings emphasize the importance of customizing treatment to balance effectiveness and safety.

Systematic Quantitative Analysis of Fetal Dexamethasone Exposure and Fetal Lung Maturation in Pregnant Animals: Model Informed Dexamethasone Precision Dose Study

Dexamethasone (DEX) was applied in neonatal respiratory distress syndrome treatment of pregnant women. We established a pharmacokinetics (PK)/pharmacodynamics(PD)/end point model of pregnant animals based on published data and then extrapolated to simulate fetal exposure and lung maturation in pregnant women. We first established the PK/PD/end point model for DEX in pregnant sheep. We considered the competitive effect of cortisol (Cort) and DEX binding with glucocorticoid receptor and then used the indirect response model to describe disaturated-phosphatidylcholine (DSPC) dynamics. Based on that, we established a regression relationship between DSPC and fetal lung volume (V40). We then extrapolated the PD/end point model of pregnant sheep to pregnant monkeys by corrected stages of morphologic lung maturation in two species. Finally, we utilized the interspecies extrapolation strategy to simulate fetal exposure (AUC0-48h) and V40 relationship in pregnant women. The current model could well describe the maternal-fetal PK of DEX in pregnant animals. Simulated DEX AUC0-24h values of the umbilical venous to maternal plasma ratio in pregnant sheep and monkeys were 0.31 and 0.27, respectively. The simulated Cort curve and V40 in pregnant sheep closely matched the observed data within a 2-fold range. For pregnant monkeys, model-simulated V40 were well fitted with external verification data, which showed good interspecies extrapolation performance. Finally, we simulated fetal exposure-response relationship in pregnant women, which indicated that the fetal AUC0-48h of DEX should not be less than 300 and 100 ng/mL·hr at GW28 and GW34 to ensure fetal lung maturity. The current model preliminarily provided support for clinical DEX dose optimization.

Model-based meta-analysis to quantify the effects of short interfering RNA therapeutics on hepatitis B surface antigen turnover in hepatitis B-infected mice

The objective of this study was to compare the efficacy of short interfering RNA therapeutics (siRNAs) in reducing hepatitis B surface antigen (HBsAg) levels in hepatitis B-infected (HBV) mice across multiple siRNA therapeutic classes using model-based meta-analysis (MBMA) techniques. Literature data from 10 studies in HBV-infected mice were pooled, including 13 siRNAs, formulated as liposomal nanoparticles (LNPs) or conjugated to either cholesterol (chol) or N-acetylgalactosamine (GalNAc). Time course of the baseline- and placebo-corrected mean HBsAg profiles were modeled using kinetics of drug effect (KPD) model coupled to an indirect response model (IRM) within a longitudinal non-linear mixed-effects MBMA framework. Single and multiple dose simulations were performed exploring the role of dosing regimens across evaluated siRNA classes. The HBsAg degradation rate (0.72 day-1) was consistent across siRNAs but exhibited a large between-study variability of 31.4% (CV%). The siRNA biophase half-life was dependent on the siRNA class and was highest for GalNAc-siRNAs (21.06 days) and lowest for chol-siRNAs (2.89 days). ID50 estimates were compound-specific and were lowest for chol-siRNAs and highest for GalNAc-siRNAs. Multiple dose simulations suggest GalNAc-siRNAs may require between 4 and 7 times less frequent dosing at higher absolute dose levels compared to LNP-siRNAs and chol-siRNAs, respectively, to reach equipotent HBsAg-lowering effects in HBV mice. In conclusion, non-clinical HBsAg concentration-time data after siRNA administration can be described using the presented KPD-IRM MBMA framework. This framework allows to quantitatively compare the effects of siRNAs on the HBsAg time course and inform dose and regimen selection across siRNA classes. These results may support siRNA development, optimize preclinical study designs, and inform data analysis methodology of future anti-HBV siRNAs; and ultimately, support siRNA model-informed drug development (MIDD) strategies.

The role of partial area under the curve and maximum concentrations in assessing the bioequivalence of long-acting injectable formulation of exenatide_A sensitivity analysis

To ensure therapeutic equivalence between the long-acting injectable (LAI) products, additional PK metrics other than Cmax and AUC were considered necessary. However, regarding the selection of additional PK metrics for bioequivalence (BE) assessment of exenatide LAI, a discrepancy existed between EMA's and USFDA's product-specific guidance. The EMA recommends that both the maximum plasma concentration in the initial-release phase (Cmax,1) and the extended-release phase (Cmax,2) should be determined. Nevertheless, the USFDA recommends the use of the partial area under the curve (i.e., the area under the curve from week 4 to the last sampling point; pAUC4w-t). The focus of this study was to compare the sensitivity of different PK metrics, including Cmax,1, Cmax,2, pAUC4w-t, early and late pAUC metrics truncated at different time points (three, four, five, six and seven weeks), to formulation-related parameters and pharmacodynamic (PD) markers of glycemic control. A sensitivity analysis was conducted using the published PK/PD model of exenatide LAI. The results indicated that Cmax,1 and Cmax,2 exhibited comparable sensitivities. AUC4w-t was sensitive to changes in detecting the differences in formulation-related parameters and PD markers of glycemic control, but did not provide superior sensitivity performance compared to Cmax,1 and Cmax,2. Among all tested PK metrics, AUC7w-t was found to be the most sensitive. The optimal cut-off time point for the pAUC should be set at the time of maximum plasma concentration in the extended-release phase (approximately 6-7 weeks). These results may provide useful insights into the selection of appropriate PK metrics for BE determination of exenatide LAI.

Model-Based Assessment of the Liver Safety Profile of Acetaminophen to Support its Combination Use with Topical Diclofenac in Mild-to-Moderate Osteoarthritis Pain

INTRODUCTION

The use of combination therapy of oral acetaminophen and topical diclofenac, having complementary mechanisms of action, is an attractive strategy to enhance the analgesic response in osteoarthritis (OA) pain. While topical diclofenac is considered as well tolerated due to its low systemic exposure, concerns of liver toxicity with acetaminophen at standard analgesic doses remain. Thus, this study aimed to assess the liver safety profile of acetaminophen, particularly in OA management, using a model-based meta-analysis (MBMA).

METHODS

A literature review was conducted using the MEDLINE database to identify randomized clinical trials (RCTs) reporting liver toxicity on acetaminophen use. An MBMA was implemented to assess the deviation from the upper limit of normal (ULN) of alanine aminotransferase or aspartate aminotransferase, namely > 0-1 × ULN, > 1.5-2 × ULN, and > 3 × ULN representing mild, moderate, and severe risk of liver abnormality, respectively.

RESULTS

A total of 15 RCTs were included in the MBMA, encompassing over 4800 subjects and exposure to acetaminophen ranging from 2 to 26 weeks. Of the 15 included studies, eight involved patients with OA pain, four involved healthy subjects and three were in patients with conditions such as asthma, glaucoma, chronic pain, and cardiovascular disease. Acetaminophen 1500-4000 mg/day was found to exhibit 23% (95% confidence interval (CI): 17.74-29.20), 1.35% (95% CI: 0.17-2.51) and 0.01% (95% CI: 0.00-0.32) increased risk for mild, moderate, and severe liver injury, respectively, versus placebo. Moreover, at therapeutic doses, no correlation was identified between acetaminophen intake and liver abnormality risk.

CONCLUSIONS

Overall, our analysis shows that short-term (~ 8-16 weeks) acetaminophen use at therapeutically recommended doses is associated with a low risk of clinically relevant changes in liver enzymes. Given the good tolerability of topical diclofenac, the findings support the safety of the combination of acetaminophen and topical diclofenac, at least over the short term, as treatment for mild-to-moderate OA pain.

Model-Based Meta-Analysis Supporting the Combination of Acetaminophen and Topical Diclofenac in Acute Pain: A Therapy for Mild-to-Moderate Osteoarthritis Pain?

INTRODUCTION

Acetaminophen and topical diclofenac (AtopD) have complementary mechanisms of action and are therefore candidates for combination use in osteoarthritis (OA) pain. However, an evidence gap exists on their combination use in OA pain. This study aimed to assess the effects of this combination and compare its performance relative to monotherapies on pain score reduction and opioid-sparing effect by leveraging evidence from acute pain setting using a model-based meta-analysis (MBMA).

METHODS

A literature search was conducted using the MEDLINE database to identify randomized controlled trials (RCTs) studying the combination for acute pain. Subsequently, an MBMA of RCTs was implemented in conjunction with extrapolation principles to infer efficacy in the population of interest. Pain score reduction and opioid-sparing effect (OSE) were selected as the measures of efficacy.

RESULTS

A total of 11 RCTs encompassing 1396 patients were included. Exploratory evaluation revealed AtopD combination to show greater pain score reduction versus acetaminophen monotherapy. However, pain score reduction was more susceptible to confounding by opioid patient-controlled analgesia (PCA) than OSE. Therefore, a parsimonious MBMA evaluating OSE was developed from 5 of the 11 RCTs (n = 353 patients). The analysis revealed a statistically significant interaction coefficient, suggesting a reduction of 32% in opioid use with the combination versus acetaminophen monotherapy. Differences in the effect size of the combination were less conclusive versus diclofenac monotherapy.

CONCLUSION

Our results indicate greater pain reduction and opioid-sparing efficacy for the AtopD combination versus acetaminophen monotherapy. Given the similar pain pathways and mechanisms of action of the two drugs in acute and mild-to-moderate OA pain, comparable beneficial effects from the combination therapy may be anticipated following extrapolation to chronic OA pain. Prospective RCTs and real-world studies in OA pain are needed to confirm the differences in the efficacy of the combination treatment observed in our study.

Model-Informed Approaches and Innovative Clinical Trial Design for Adeno-Associated Viral Vector-Based Gene Therapy Product Development: A White Paper

The promise of viral vector-based gene therapy (GT) as a transformative paradigm for treating severely debilitating and life-threatening diseases is slowly coming to fruition with the recent approval of several drug products. However, they have a unique mechanism of action often necessitating a tortuous clinical development plan. Expertise in such complex therapeutic modality is still fairly limited in this emerging class of adeno-associated virus (AAV) vector-based gene therapies. Because of the irreversible mode of action and incomplete understanding of genotype-phenotype relationship and disease progression in rare diseases careful considerations should be given to GT product's benefit-risk profile. In particular, special attention needs to be paid to safe dose selection, reliable dose exposure response (including clinically relevant endpoints), or creative approaches in study design targeting small patient populations during clinical development. We believe that quantitative tools encompassed within model-informed drug development (MIDD) framework fits quite well in the development of such novel therapies, as they enable us to benefit from the totality of data approach in order to support dose selection as well as optimize clinical trial designs, end point selection, and patient enrichment. In this thought leadership paper, we provide our collective experiences, identify challenges, and suggest areas of improvement in applications of modeling and innovative trial design in development of AAV-based GT products and reflect on the challenges and opportunities for incorporating MIDD tools and more in rational development of these products.

Longitudinal Meta-Analysis of Historical Parkinson's Disease Trials to Inform Future Trial Design

BACKGROUND

The outcome of clinical trials in neurodegeneration can be highly uncertain due to the presence of a strong placebo effect.

OBJECTIVES

To develop a longitudinal model that can enhance the success of future Parkinson's disease trials by quantifying trial-to-trial variations in placebo and active treatment response.

METHODS

A longitudinal model-based meta-analysis was conducted on the total score of Unified Parkinson's Disease Rating Scale (UPDRS) Parts 1, 2, and 3. The analysis included aggregate data from 66 arms (observational [4], placebo [28], or investigational-drug-treated [34]) from 4 observational studies and 17 interventional trials. Inter-study variabilities in key parameters were estimated. Residual variability was weighted by the size of study arms.

RESULTS

The baseline total UPDRS was estimated to average at 24.5 points. Disease score was estimated to worsen by 3.90 points/year for the duration of the treatments; whilst notably, arms with a lower baseline progressed faster. The model captured the transient nature of the placebo response and sustained symptomatic drug effect. Both placebo and drug effects peaked within 2 months; although, 1 year was needed to observe the full treatment difference. Across these studies, the progression rate varied by 59.4%, the half-life for offset of placebo response varied by 79.4%, and the amplitude for drug effect varied by 105.3%.

CONCLUSION

The longitudinal model-based meta-analysis describes UPDRS progression rate, captures the dynamics of the placebo response, quantifies the effect size of the available therapies, and sets the expectation of uncertainty for future trials. The findings provide informative priors to enhance the rigor and success of future trials of promising agents, including potential disease modifiers. © 2023 GSK. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Joint longitudinal model-based meta-analysis of FEV1 and exacerbation rate in randomized COPD trials

Model-based meta-analysis (MBMA) is an approach that integrates relevant summary level data from heterogeneously designed randomized controlled trials (RCTs). This study not only evaluated the predictability of a published MBMA for forced expiratory volume in one second (FEV₁) and its link to annual exacerbation rate in patients with chronic obstructive pulmonary disease (COPD) but also included data from new RCTs. A comparative effectiveness analysis across all drugs was also performed. Aggregated level data were collected from RCTs published between July 2013 and November 2020 (n = 132 references comprising 156 studies) and combined with data used in the legacy MBMA (published RCTs up to July 2013 - n = 142). The augmented data (n = 298) were used to evaluate the predictive performance of the published MBMA using goodness-of-fit plots for assessment. Furthermore, the model was extended including drugs that were not available before July 2013, estimating a new set of parameters. The legacy MBMA model predicted the post-2013 FEV₁ data well, and new estimated parameters were similar to those of drugs in the same class. However, the exacerbation model overpredicted the post-2013 mean annual exacerbation rate data. Inclusion of year when the study started on the pre-treatment placebo rate improved the model predictive performance perhaps explaining potential improvements in the disease management over time. The addition of new data to the legacy COPD MBMA enabled a more robust model with increased predictability performance for both endpoints FEV₁ and mean annual exacerbation rate.

The potential of a data centred approach & knowledge graph data representation in chemical safety and drug design

Big Data pervades nearly all areas of life sciences, yet the analysis of large integrated data sets remains a major challenge. Moreover, the field of life sciences is highly fragmented and, consequently, so is its data, knowledge, and standards. This, in turn, makes integrated data analysis and knowledge gathering across sub-fields a demanding task. At the same time, the integration of various research angles and data types is crucial for modelling the complexity of organisms and biological processes in a holistic manner. This is especially valid in the context of drug development and chemical safety assessment where computational methods can provide solutions for the urgent need of fast, effective, and sustainable approaches. At the same time, such computational methods require the development of methodologies suitable for an integrated and data centred Big Data view. Here we discuss Knowledge Graphs (KG) as a solution to a data centred analysis approach for drug and chemical development and safety assessment. KGs are knowledge bases, data analysis engines, and knowledge discovery systems all in one, allowing them to be used from simple data retrieval, over meta-analysis to complex predictive and knowledge discovery systems. Therefore, KGs have immense potential to advance the data centred approach, the re-usability, and informativity of data. Furthermore, they can improve the power of analysis, and the complexity of modelled processes, all while providing knowledge in a natively human understandable network data model.

How can natural language processing help model informed drug development?: a review

Objective

To summarize applications of natural language processing (NLP) in model informed drug development (MIDD) and identify potential areas of improvement.

Materials and Methods

Publications found on PubMed and Google Scholar, websites and GitHub repositories for NLP libraries and models. Publications describing applications of NLP in MIDD were reviewed. The applications were stratified into 3 stages: drug discovery, clinical trials, and pharmacovigilance. Key NLP functionalities used for these applications were assessed. Programming libraries and open-source resources for the implementation of NLP functionalities in MIDD were identified.

Results

NLP has been utilized to aid various processes in drug development lifecycle such as gene-disease mapping, biomarker discovery, patient-trial matching, adverse drug events detection, etc. These applications commonly use NLP functionalities of named entity recognition, word embeddings, entity resolution, assertion status detection, relation extraction, and topic modeling. The current state-of-the-art for implementing these functionalities in MIDD applications are transformer models that utilize transfer learning for enhanced performance. Various libraries in python, R, and Java like huggingface, sparkNLP, and KoRpus as well as open-source platforms such as DisGeNet, DeepEnroll, and Transmol have enabled convenient implementation of NLP models to MIDD applications.

Discussion

Challenges such as reproducibility, explainability, fairness, limited data, limited language-support, and security need to be overcome to ensure wider adoption of NLP in MIDD landscape. There are opportunities to improve the performance of existing models and expand the use of NLP in newer areas of MIDD.

Conclusions

This review provides an overview of the potential and pitfalls of current NLP approaches in MIDD.