formulation design
Nanosuspension, amorphous solid dispersion or lipid formulation?
Using PBPK modelling to guide formulation technology selection.
The challenge
Nanosuspensions, amorphous solid dispersions and lipid-based formulations are the most widely used strategies to address drug solubility issues. Selecting the technology best suited to a given formulation problem is clearly a multivariate decision, governed by factors such as manufacturability, stability, tolerability and cost of goods. In early development, when the intended dose and systemic exposure targets are still uncertain, the primary criterion for formulation technology selection is usually enablement, that is, the capacity of a formulation to enhance absorption relative to unformulated drug. Estimating the absorption-enhancing potential of a new formulation for a new drug is far from straightforward. The classical approach to assessing absorption-enhancing potential involves administering a series of formulations based on diverse formulation technologies to laboratory animals and measuring plasma concentration as a function of time. The formulation that provides the highest exposure then usually gets selected to support further preclinical studies. This trial-and-error approach consumes considerable time, effort and compound.
Utilisation of physiologically based pharmacokinetic (PBPK) modelling, however, enables an a priori assessment of which formulation technology is best suited to achieve adequate absorption at a given dose level – even in early development, when experimental data are scarce. Such model-derived predictions render formulation technology selection a more rational and targeted process, and can prevent resources being spent on formulations that deliver insufficient absorption enhancement.
The following case study illustrates how a simple PBPK model, parameterised from only a limited set of in-vitro data, was successfully applied to guide formulation technology selection in an early development project.
Compound profile
The compound’s properties are summarised below. The only data points collected experimentally were solubility in phosphate buffer and fasted state simulated intestinal fluid (FaSSIF) as well as permeability across a Caco-2 cell line. LogP and pKa were calculated in silico.
Nanosuspensions, amorphous solid dispersions and lipid-based formulations are the most widely used strategies to address drug solubility issues. Selecting the technology best suited to a given formulation problem is clearly a multivariate decision, governed by factors such as manufacturability, stability, tolerability and cost of goods. In early development, when the intended dose and systemic exposure targets are still uncertain, the primary criterion for formulation technology selection is usually enablement, that is, the capacity of a formulation to enhance absorption relative to unformulated drug. Estimating the absorption-enhancing potential of a new formulation for a new drug is far from straightforward. The classical approach to assessing absorption-enhancing potential involves administering a series of formulations based on diverse formulation technologies to laboratory animals and measuring plasma concentration as a function of time. The formulation that provides the highest exposure then usually gets selected to support further preclinical studies. This trial-and-error approach consumes considerable time, effort and compound.
Utilisation of physiologically based pharmacokinetic (PBPK) modelling, however, enables an a priori assessment of which formulation technology is best suited to achieve adequate absorption at a given dose level – even in early development, when experimental data are scarce. Such model-derived predictions render formulation technology selection a more rational and targeted process, and can prevent resources being spent on formulations that deliver insufficient absorption enhancement.
The following case study illustrates how a simple PBPK model, parameterised from only a limited set of in-vitro data, was successfully applied to guide formulation technology selection in an early development project.
Compound profile
The compound’s properties are summarised below. The only data points collected experimentally were solubility in phosphate buffer and fasted state simulated intestinal fluid (FaSSIF) as well as permeability across a Caco-2 cell line. LogP and pKa were calculated in silico.
PBPK model
The model was deliberately kept simple: it predicts the fraction of the dose absorbed (Fₐ) only, and not oral bioavailability (F), as Fₐ is the only step in the exposure chain that is influenced by formulation. Bioavailability additionally requires the gut-wall and hepatic first-pass extraction to be known, and no metabolic in vitro data were available at this stage. Predicting F would therefore have layered avoidable uncertainty on top of the absorption prediction without changing the formulation selection decision the model was built to support.
How the model described dissolution and absorption
Drug dissolution was described by the Noyes–Whitney equation, which assumes that dissolution rate is proportional to the total drug surface area and the difference between the drug’s equilibrium solubility and dissolved concentration in the surrounding fluid. Because surface area scales with the inverse of particle radius, particle size is the formulation lever the model responds to: reducing the median size steeply increases how fast the dose dissolves.
The dissolved fraction is then transported across the intestinal wall at a rate set by the effective intestinal permeability. Two distinct limits can therefore constrain absorption: the dissolution rate (a rate limit, which particle size controls) and solubility (a concentration limit, which particle size cannot change). The purpose of the simulations was to establish which of these two factors dominates at each dose level.
Key assumptions
What was simulated
Two inputs were varied in the simulations. Median particle size was scanned from unmilled crystalline API (D50 ≈ 20 µm) down to a fine nanosuspension (D50 ≈ 100 nm). Dose was scanned at 1, 10 and 100 mg. Fₐ was read out directly for every combination.
What the model showed
The model showed that, at 10 mg, absorption was essentially complete regardless of particle size. At 10 mg, it was dissolution-rate–limited: a nanosuspension below ≈ 300 nm lifted the absorbed fraction (Fₐ) to ≈ 97–99 %. At a 100 mg dose, absorption was clearly solubility-limited: even the finest particles were predicted to yield an absorbed fraction of 12%. To further enhance absorption at this highest dose level, more enabling approaches that supersaturate and/or solubilise the compound in the gastrointestinal fluids would be required.
The model was deliberately kept simple: it predicts the fraction of the dose absorbed (Fₐ) only, and not oral bioavailability (F), as Fₐ is the only step in the exposure chain that is influenced by formulation. Bioavailability additionally requires the gut-wall and hepatic first-pass extraction to be known, and no metabolic in vitro data were available at this stage. Predicting F would therefore have layered avoidable uncertainty on top of the absorption prediction without changing the formulation selection decision the model was built to support.
How the model described dissolution and absorption
Drug dissolution was described by the Noyes–Whitney equation, which assumes that dissolution rate is proportional to the total drug surface area and the difference between the drug’s equilibrium solubility and dissolved concentration in the surrounding fluid. Because surface area scales with the inverse of particle radius, particle size is the formulation lever the model responds to: reducing the median size steeply increases how fast the dose dissolves.
The dissolved fraction is then transported across the intestinal wall at a rate set by the effective intestinal permeability. Two distinct limits can therefore constrain absorption: the dissolution rate (a rate limit, which particle size controls) and solubility (a concentration limit, which particle size cannot change). The purpose of the simulations was to establish which of these two factors dominates at each dose level.
Key assumptions
- Solubility. As the compound is non-ionisable, its solubility is unaffected by the pH gradient along the intestinal tract. However, solubility in FaSSIF was markedly higher than in phosphate buffer, indicating substantial bile-salt-mediated solubilisation. This effect was therefore incorporated into the PBPK model.
- Permeability. The effective intestinal permeability was derived from the measured Caco-2 apparent permeability. No efflux was observed in vitro, so no active (efflux) transport was included in the model.
- Physiology. A healthy human adult in the fasted state, with default gastrointestinal transit times, fluid volumes and pH profile. The same model structure can be rebuilt with rat or dog physiology should preclinical exposure need to be projected.
What was simulated
Two inputs were varied in the simulations. Median particle size was scanned from unmilled crystalline API (D50 ≈ 20 µm) down to a fine nanosuspension (D50 ≈ 100 nm). Dose was scanned at 1, 10 and 100 mg. Fₐ was read out directly for every combination.
What the model showed
The model showed that, at 10 mg, absorption was essentially complete regardless of particle size. At 10 mg, it was dissolution-rate–limited: a nanosuspension below ≈ 300 nm lifted the absorbed fraction (Fₐ) to ≈ 97–99 %. At a 100 mg dose, absorption was clearly solubility-limited: even the finest particles were predicted to yield an absorbed fraction of 12%. To further enhance absorption at this highest dose level, more enabling approaches that supersaturate and/or solubilise the compound in the gastrointestinal fluids would be required.
What it meant for formulation technology selection
In short, the model showed that a nanosuspension was the right tool to achieve complete absorption at low doses (up to ca. 10 mg). Higher doses would call for a more enabling technology such as an amorphous solid dispersion or lipid-based formulation. The model flagged that crossover early on, before any formulation work had been conducted. As clinical doses above 10 mg were anticipated, the project team decided to pursue a solubility-enhancing formulation for this compound. Application of the PBPK model thus avoided spending resources on a nanosuspension that would not have provided adequate absorption for doses higher than 10 mg.
In short, the model showed that a nanosuspension was the right tool to achieve complete absorption at low doses (up to ca. 10 mg). Higher doses would call for a more enabling technology such as an amorphous solid dispersion or lipid-based formulation. The model flagged that crossover early on, before any formulation work had been conducted. As clinical doses above 10 mg were anticipated, the project team decided to pursue a solubility-enhancing formulation for this compound. Application of the PBPK model thus avoided spending resources on a nanosuspension that would not have provided adequate absorption for doses higher than 10 mg.
How Harpago can help
Our team integrates biorelevant solubility and dissolution information with PBPK absorption modelling to guide formulation decisions early, turning formulation development efforts into a predictable, data-driven path to maximal absorption.
Our team integrates biorelevant solubility and dissolution information with PBPK absorption modelling to guide formulation decisions early, turning formulation development efforts into a predictable, data-driven path to maximal absorption.
Contact us for a project consultation