Developing and optimizing drug formulations― a key component of a product development―is a very lengthy and capital intensive process. Today, most drug candidates are poorly water-soluble; this has led to greater emphasis on screening more complex formulation technologies. Formulation development is still largely an empirical process ― based on trial and error and formulation scientists’ experience!
Physiologically-based pharmacokinetic (PBPK) modeling has emerged as a valuable resource to support decisions throughout the drug development process. Utilizing PBPK models in discovery programs can support ‘rational’ product development, thereby expediting the process of moving potential active pharmaceutical ingredients (APIs) from discovery to the clinic and subsequent commercialization.
This systematic modeling approach applies to several areas of drug product development such as predicting formulation effects, forecasting food-drug interactions, developing IVIVCs, predicting virtual bioequivalence, justifying biowaivers, and more. In fact, the mechanistic and predictive ability of PBPK models enables exploring the product design spaces more effectively and can facilitate implementing ‘Quality by Design’ (QbD) in a more meaningful way!
Moreover, the interest of regulatory agencies in the diverse applications of PBPK modeling is reflected in their frequent references in recently approved drug labels, regulatory guidances, and peer-reviewed papers.
The mechanistic, physiologically-based Advanced Dissolution, Absorption and Metabolism (ADAM) Model within the Certara’s Simcyp® population-based Simulator helps formulation scientists predict the variability in human oral drug absorption from physiochemical and in vitro drug data. The ADAM model can simulate a variety of formulations: solutions, suspensions and immediate release (IR) tablets through to single unit (monoliths) and dispersible dosage forms (viz. gastro-retentive, enteric coated tablets and granules, controlled release (CR) monoliths, and CR dispersions) that release the API over time with or without lag time.
Throughout the years, biopharmaceutical experts from industrial, academic and regulatory organizations have demonstrated how absorption modeling using ADAM can inform formulation development and help generate insights into the product performance in vivo. Here we review several such case studies covering different aspects of biopharmaceutics or formulation questions.
Predicting food-drug interactions
Predicting the effect of food on drug exposure, and thereby its safety and efficacy, early in drug development is pivotal to clinical success and to optimal formulation strategy. Current ‘empirical’ methods (such as BCS, BDDCS and QSAR-based methods) often cannot quantify the magnitude of food effects; this has spurred developing physiologically-based modeling approaches.
With appropriate in vitro data, population-based PBPK models can integrate all available physiological (or system) data and drug/ formulation-specific information to predict food effects. A range of food-induced physiological changes are incorporated into the ADAM model to simulate the clinically observed phenomena, viz. increased splanchnic blood flow, delayed gastric residence time, dynamic change in the gastric pH, bile salt concentrations, viscosity, and dynamic fluid volumes.
Recently, we successfully predicted the differential food effects on absorption of nifedipine from oral IR and CR formulations using the ADAM model where established rule-based approaches are inapplicable. The study used mechanistic PBPK models with in vitro data to predict variations in the PK of the same formulation in the fasted and fed states as well as between different formulations. Anticipating the ‘formulation specific’ food effects in early stages of drug development is of great significance. Applying validated PBPK models, as described in this work, may help formulation scientists in guiding systematic formulation development, reducing undesirable food effects and avoiding relabeling and safety issues in later stages of product development.
Quantitative prediction of food effect for weakly basic drug compounds is challenging due to their variable dissolution and precipitation in the dynamically changing GI environment. PBPK models can account for these food-induced changes in GI tract and can help predict food-drug interactions. In another such study, researchers leveraged the ADAM model to explore the mechanism(s) behind the differences observed in the duodenal concentration–time profiles and in the magnitude of food effect for two weakly basic, structurally related drugs- ketoconazole and posaconazole.
Food, among a range of other effects, can also alter the viscosity of the GI tract fluids to delay tablet disintegration and potentially reduce drug absorption. In another successful case study, a dynamic viscosity-disintegration model was combined with ADAM and in vitro data to anticipate negative food effects upon drug absorption. Dynamic changes to the in vivo disintegration rate of an IR formulation of a BCS Class III drug, trospium chloride, was linked to dilutive, time-dependent viscosity changes after food intake.
Using in vitro data alone, the ADAM model has also been used to understand the mechanisms underpinning the effect of proton pump inhibitors (PPIs) and acidic carbonated beverages on the oral absorption of drugs. Proton pump inhibitors (PPIs) are OTC products routinely used to treat certain gastrointestinal disorders. PPIs work by reducing the amount of acid in the stomach. As patients are required to administer these medicines for a long period, PPIs may affect the absorption of co-medications. Many people drink soda on a daily basis. Because of their acidic nature, these drinks may alter the gastric environment and thereby may affect the PK of drug compounds. Regulatory agencies require such detrimental changes in drug exposure, if any, needs to be tested in lengthy and costly clinical trials. Validated PBPK model can help examine such interactions and may support justifying biowaivers.
These case studies demonstrate that mechanistic model-based approaches integrating both drug and system data have numerous applications including quantitative food effect predictions, rational formulation design, aiding regulatory approvals by supporting biowaivers, reducing the number of clinical studies, and thus informing better decisions.
Developing mechanistic IVIVCs
The ADAM model can also be used to establish physiologically based in vitro–in vivo correlations (PB-IVIVCs). In cases of significant gut wall and/or hepatic first-pass metabolism of a drug, establishing robust relationships between in vitro and deconvoluted in vivo dissolution profiles can become difficult, perhaps requiring complex non-linear functions. PBPK-based deconvolution can disentangle these complex processes and estimate in vivo dissolution rather than absorption allowing more robust and simpler IVIVC models compared to the conventional IVIVC methods. Such simplified and usually linear IVIVCs can accelerate formulation development while supporting safety and overall product quality.
This approach has thus far been successfully applied to the development and validation of IVIVC for CR formulations of metoprolol, diltiazem, tramadol and topiramate. The approach has also been leveraged to CR formulations of BCS II drugs, e.g. azithromycin, where absorption is governed by the complex interplay of release, transit/gastro-retention, and permeability rather than just release characteristics.
Recently FDA scientists demonstrated the importance of factoring population variability into metoprolol IVIVC estimation and profile reconvolution. They demonstrated that in addition to permeation (Peff) and disposition characteristics (Vss/CL) of the individuals using oral solution, gastric emptying time (GET) played a vital role in refining the IVIVC. Factoring out this inter-occasion and inter-subject GET variability during individual deconvolution evidently helped to improve the correlation.
Establishing virtual bioequivalence
Predicting in vivo equivalence of drug products virtually is a subject of great interest for pharmaceutical scientists and regulatory agencies. A PBPK modeling approach can predict the population PK variability of a formulated API in a ‘virtual population’ and enable assessing the likelihood of ‘product bioequivalence’ via virtual trials. Accounting for “variability” in these virtual trials can impact several areas of drug product development including formulation safe space design, clinically relevant dissolution specification setting, aiding justification of biowaivers, formulation changes in late stage development and beyond. Recently, a validated PBPK model of tramadol was used to run virtual bioequivalence (BE) trials; this approach can inform setting dissolution specifications and, consequently, building a safe design space based upon Weibull function parameters.
Additionally, PBPK based virtual trials coupled with pharmacodynamic (PD) models have been used to assess the clinical relevance of bioequivalence criteria. Colleagues at the Brazilian Health Surveillance Agency (ANVISA) constructed a PBPK model for the non-steroidal anti-inflammatory drug (NSAID) ibuprofen and coupled it with two published PD models: antipyresis and dental pain relief. With the help of a validated PBPK-PD modeling approach, the authors demonstrated that the current PK based BE approach may be too restrictive for ibuprofen products.
Bioequivalence studies are typically conducted in healthy volunteers, but the indicated patient population may have different physiology than healthy population. PBPK models hold an advantage over other modeling approaches as they account for both the drug formulation characteristics and the underlying physiology of the species studied and its co-variates within a population. Hence, PBPK models can “extrapolate” to other populations such as pediatric or bariatric surgery patients, where conducting clinical studies is quite challenging! In this context, scientists demonstrated how PBPK based virtual trials can assess product performance of two weakly basic drug compounds— ketoconazole and posaconazole— in a variety of patient populations and clinical situations.
A novel biopharmaceutical-IVIV_E paradigm
In vitro- in vivo extrapolation (IVIV_E) techniques translate parameters derived from in vitro experiments to their corresponding in vivo counterparts to predict the in vivo behavior of drug candidates. The Simcyp In Vitro Analysis (SIVA) toolkit is a user-friendly software package, designed to help pharmaceutical scientists analyze complex data generated from dissolution techniques such as USP II, USP IV, transfer model, two-phase dissolution model, etc. This approach may also help formulation scientists to estimate unknown/uncertain parameters of the drug product i.e. particle size, drug precipitation parameters etc., that are generally unavailable in early product development. Moreover, this approach streamlines and optimizes designing in vitro experiments to potentially reduce the cost and time of formulation development. To learn more about using SIVA to get the most from your in vitro experiments, please watch this webinar by the Simcyp team.
Various examples of biopharmaceutical IVIV_E viz. danazol (Modeling USP II Dissolution), dipyridamole (conventional USP II Vs. two-phase dissolution modeling), ketoconazole (transfer experiment modeling), and posaconazole (changing dissolution media modeling) demonstrate that PBPK modeling informed by mechanistic modeling of in vitro experiments increases confidence in the quality of the input parameters and mechanistic models used for in vivo simulations.
Over the years, PBPK modeling has transformed from merely an early stage modeling tool to a mature field with proven potential to reduce and refine clinical trials to study drug-drug interactions and drug effects in special populations. However, its applications in biopharmaceutics and formulation studies have not been explored extensively. Recent advances and applications in the use of Simcyp represent an opportunity for formulation/experimental scientists to explore modeling in designing and/or screening formulations. Using validated predictive modeling techniques will lead to more rational drug development.
All information presented derive from public source materials.
To learn more about mechanistic approaches to IVIVC, please watch this webinar.