📄 medicine

Development of a Dissolution-Informed PBPK-Assisted In Vitro–In Silico Correlation Framework for Predicting Food Effects of Lurasidone Hydrochloride

This study developed and validated a dissolution-informed PBPK-assisted in vitro–in silico correlation framework that successfully predicts the food effects and systemic exposure of the poorly water-soluble drug lurasidone hydrochloride by integrating biorelevant dissolution data with physiologically based pharmacokinetic modeling.

Original authors: Minal Narkhede, Sonal Mhaske, Jitendra Nehete

Published 2026-08-12

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Minal Narkhede, Sonal Mhaske, Jitendra Nehete

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to get a message across a crowded room. If the message is written on a slippery, water-soluble piece of paper, it might dissolve before it reaches the other side. But if the paper is thick and waxy, it might not dissolve at all, leaving the message unread. This is a bit like how many medicines work inside our bodies. Some drugs are like that waxy paper: they don't mix well with the watery fluids in our stomach and intestines. Because they can't dissolve easily, they can't get absorbed into the bloodstream to do their job. Scientists call this a "dissolution-limited" problem.

Now, imagine that the crowded room changes its rules depending on whether you just ate a big meal or are still hungry. If you just ate, your stomach produces more special "detergents" (called bile salts) that can help break down that waxy paper. This is the "food effect." For some medicines, eating a meal makes them work much better; for others, it might make them work worse. Predicting exactly how a specific drug will behave in these different conditions is tricky. Scientists use two main tools to solve this puzzle: they test how fast a drug dissolves in a cup of liquid that mimics the stomach (called in vitro), and they use powerful computer models to simulate how the drug moves through a human body (called in silico or PBPK modeling). The big question is: can we combine these two tools to predict exactly how a drug will act in a real person, especially when food is involved?

This paper dives into that question using a specific drug called Lurasidone hydrochloride, which is used to treat conditions like schizophrenia. Lurasidone is a classic example of a drug that struggles to dissolve in water and has a huge "food effect"—meaning its performance changes dramatically depending on whether you take it on an empty stomach or with a meal. The researchers wanted to see if they could build a "digital twin" of the human body that could accurately predict how this drug behaves, using real-world data from their lab experiments as the starting point.

The Lab Experiment: Testing the "Stomach" Waters

First, the team went into the lab to see how Lurasidone behaves in different types of "fake stomachs." They didn't just use plain water; they created four different liquid environments to mimic the human body under different conditions:

Fasted Stomach: Empty stomach, acidic.
Fed Stomach: Full stomach, less acidic, with food-like properties.
Fasted Intestine: Empty intestine, neutral pH.
Fed Intestine: Full intestine, rich in bile salts (the body's natural detergents).

They dropped a 40 mg tablet of Lurasidone into each of these four liquids and watched how much of the drug dissolved over two hours. The results were clear and dramatic. The drug dissolved the most in the Fed Stomach fluid (85.1%), followed closely by the Fed Intestine fluid (78.8%). In contrast, it dissolved much less in the empty stomach (61.8%) and the least in the empty intestine (57.7%).

This confirmed what doctors already knew: food helps this drug dissolve. But the researchers wanted to know exactly how much better it would be in a real human body, not just in a beaker.

The Computer Simulation: Building a Digital Body

Next, the team took those real-world numbers from the beakers and fed them into a sophisticated computer program called PK-Sim. Think of this program as a video game where you build a virtual human. The scientists programmed this virtual human with all the right details: how big the stomach is, how fast it empties, how much blood flows to the liver, and how the liver breaks down drugs.

They didn't just guess how the drug would move; they used the actual dissolution data from their lab experiments to tell the computer exactly how fast the drug was releasing in each of the four "stomach" scenarios. The computer then ran the simulation to predict what the drug levels would look like in the blood over time.

The Big Reveal: The Computer Got It Right

The results of the simulation were impressive. The computer predicted that taking the drug with food (the "Fed" state) would lead to a much higher amount of drug in the blood compared to taking it on an empty stomach.

In the Fed Stomach simulation, the peak drug level (Cmax) was predicted to be 64.26 µmol/L.
In the Fasted Stomach simulation, the peak was only 46.68 µmol/L.

This means the computer predicted a 37.7% increase in the peak drug level just by adding food to the mix. The total amount of drug absorbed over time (AUC) also jumped by nearly 60% in the fed state.

But how do we know the computer wasn't just making things up? The researchers compared their computer predictions against real data from actual human clinical studies that had been published in other papers. The match was uncanny.

The computer's prediction for the peak level was off by less than 4% compared to real human data.
The "Fold Error" (a way to measure how close the prediction is to reality) was 1.04, which is almost perfect (a score of 1.0 would be a perfect match).
The correlation between how much the drug dissolved in the beaker and how much the computer predicted would be in the blood was extremely strong, with a score of 0.9545.

This proved that the "digital twin" was accurate. The computer successfully translated the simple lab experiment (drug dissolving in a cup) into a complex prediction of what happens inside a human body.

What Drives the System?

The researchers also played a game of "what if" using their computer model. They changed one variable at a time to see which part of the body mattered most for how much drug got into the blood. They found two main drivers:

Solubility: How well the drug dissolves. This was the biggest positive factor. If the drug dissolves better, more gets into the blood.
Liver Cleanup: The liver uses an enzyme called CYP3A4 to break down the drug. This was the biggest negative factor. If the liver works harder, less drug stays in the blood.

Interestingly, things like how fast the stomach empties or how long the drug sits in the intestine mattered, but they weren't as critical as the drug's ability to dissolve or the liver's ability to clean it up.

Special Populations: What About Sick Livers?

Finally, the team asked a crucial question: What happens if the "digital human" has a liver that isn't working perfectly? They simulated a scenario for people with hepatic impairment (liver damage).

In these virtual patients, the drug levels went up. The peak level (Cmax) rose from 64.2 to 77.0 µmol/L.
The total exposure (AUC) increased from 5138.05 to 5395.20 µmol·min/L.

This suggests that people with liver issues might absorb even more of the drug, which is important information for doctors to know when prescribing the medication.

The Takeaway

This study didn't just guess; it built a bridge between a simple lab test and a complex human body. By combining real dissolution data with a computer model, the researchers created a reliable tool that can predict how Lurasidone behaves in different people and under different conditions.

The paper suggests that this approach—using biorelevant dissolution tests to feed into PBPK models—works very well for drugs that struggle to dissolve. It shows that we can predict food effects and even how the drug behaves in people with liver problems without needing to run as many expensive and time-consuming clinical trials on real humans. While the study focused on just one drug and one type of tablet, the method they used is a powerful new way to understand how medicines work, potentially making drug development faster and safer for everyone.

Technical Summary: Dissolution-Informed PBPK-Assisted IVISc Framework for Lurasidone Hydrochloride

Problem Statement
Lurasidone hydrochloride, a second-generation atypical antipsychotic classified as Biopharmaceutics Classification System (BCS) Class II, exhibits low aqueous solubility and high intestinal permeability. Its oral absorption is dissolution-limited and significantly influenced by gastrointestinal physiology, resulting in a pronounced positive food effect (approximately 2–3-fold increase in bioavailability under fed conditions). Conventional dissolution testing in simple aqueous media often fails to predict the in vivo performance of such poorly water-soluble compounds. While Physiologically Based Pharmacokinetic (PBPK) modeling and biorelevant dissolution testing are established tools, there remains a gap in integrating experimentally generated dissolution profiles from multiple biorelevant media directly into mechanistic PBPK simulations to establish a robust in vitro–in silico correlation (IVISc) for predicting food effects and systemic exposure.

Methodology
The study employed a systematic workflow integrating experimental biorelevant dissolution testing with PBPK modeling using PK-Sim® (version 12.2):

Biorelevant Dissolution Studies: Dissolution profiles of a 40 mg lurasidone hydrochloride tablet were generated in four simulated gastrointestinal media: Fasted State Simulated Gastric Fluid (FaSSGF), Fed State Simulated Gastric Fluid (FeSSGF), Fasted State Simulated Intestinal Fluid (FaSSIF), and Fed State Simulated Intestinal Fluid (FeSSIF). Experiments were conducted at 37°C using USP Apparatus II.
Kinetic and Statistical Analysis: Dissolution data were fitted to Zero-order, First-order, Higuchi, and Korsmeyer–Peppas kinetic models. Similarity factors ( $f_2$ ) were calculated to compare profiles between fasted and fed states.
PBPK Model Development: A whole-body PBPK model was constructed using a bottom-up approach. Drug-specific physicochemical parameters (e.g., pKa, logP, intrinsic solubility, fraction unbound) and physiological parameters were sourced from literature and the PK-Sim database. The model incorporated region-specific gastrointestinal transit, pH-dependent solubility, and CYP3A4-mediated hepatic metabolism.
Integration and Simulation: Experimentally derived dissolution profiles from the four media were integrated as time-dependent inputs into the PBPK model. Simulations were performed for virtual adult populations under fasted and fed conditions to predict plasma concentration–time profiles and pharmacokinetic (PK) parameters ( $C_{max}$ , $T_{max}$ , $AUC_{tend}$ ).
IVISc and Validation: A PBPK-assisted IVISc framework was established by correlating the extent of dissolution at 120 minutes with predicted systemic exposure ( $AUC_{tend}$ ). Model performance was validated against external clinical PK data using fold error (FE), average absolute fold error (AAFE), goodness-of-fit analysis, and visual predictive checks (VPC).
Sensitivity and Special Population Analysis: Sensitivity analysis identified key drivers of systemic exposure. The validated model was further applied to simulate pharmacokinetics in a hepatic impairment population (Child–Pugh B).

Key Results

Dissolution Behavior: Dissolution was significantly influenced by gastrointestinal conditions ( $p < 0.001$ ). The highest drug release was observed in FeSSGF (85.1%), followed by FeSSIF (78.8%), FaSSGF (61.8%), and FaSSIF (57.7%). The Korsmeyer–Peppas model provided the best fit for most media, indicating non-Fickian (anomalous) release behavior.
PBPK Predictions: The model successfully translated dissolution inputs into PK outcomes. FeSSGF simulations yielded the highest predicted $C_{max}$ (64.26 µmol/L) and $AUC_{tend}$ (5138.05 µmol·min/L). Compared to fasted states, fed-state simulations predicted a 37.7% increase in $C_{max}$ and a 59.9% increase in $AUC_{tend}$ for gastric conditions, and 36.3% and 40.2% for intestinal conditions, respectively.
IVISc Correlation: A strong linear relationship was established between the percentage of drug dissolved at 120 minutes and predicted $AUC_{tend}$ ( $R^2 = 0.9545$ ), confirming that dissolution extent directly drives systemic exposure in the model.
Model Validation: The model demonstrated excellent predictive accuracy. Mean fold error (MFE) and average absolute fold error (AAFE) were both 1.04. Prediction errors for $C_{max}$ and $AUC$ ranged from 3.67% to 3.98%, well within the accepted $\pm 20\%$ criteria. Goodness-of-fit analysis showed strong agreement ( $R^2 = 0.97$ for $C_{max}$ , $0.98$ for $AUC$).
Sensitivity Analysis: Solubility and CYP3A4-mediated clearance were identified as the principal determinants of systemic exposure, followed by oral dose and effective permeability.
Hepatic Impairment: Simulations in subjects with hepatic impairment predicted an increase in $C_{max}$ (from 64.2 to 77.0 µmol/L) and $AUC_{tend}$ (from 5138.05 to 5395.20 µmol·min/L), reflecting reduced metabolic clearance.

Significance and Claims
The paper claims that the developed dissolution-informed PBPK framework successfully translates biorelevant dissolution behavior into pharmacokinetic outcomes with high predictive accuracy. The integrated PBPK-assisted IVISc demonstrates robust prediction of food effects and systemic exposure for lurasidone hydrochloride.

The authors position this work as a practical application of Model-Informed Drug Development (MIDD). They assert that the framework supports:

Formulation Optimization: By linking specific dissolution profiles to predicted exposure.
Food Effect Prediction: Mechanistically explaining the positive food effect through solubilization and dissolution kinetics.
Special Population Assessment: Demonstrating the model's ability to predict PK changes in hepatic impairment.
Regulatory Utility: Aligning with FDA and ICH guidance (including ICH M15) regarding the use of PBPK for biopharmaceutics applications, virtual bioequivalence, and reducing reliance on extensive clinical studies for poorly water-soluble drugs.

The study concludes that this approach provides a mechanistic alternative to empirical in vitro–in vivo correlations (IVIVC), offering a validated tool for evaluating the impact of gastrointestinal physiology on the absorption of BCS Class II drugs.