A Physiology Based Three Layer Diffusion Reaction Model For In Vitro To In Vivo Translation Of Intravesical Mitomycin C Exposure In The Human Urinary Bladder Wall

Technical Summary: A Physiology-Based Three-Layer Diffusion-Reaction Model for In Vitro to In Vivo Translation of Intravesical Mitomycin C

Problem Statement
Non-muscle-invasive bladder cancer (NMIBC) is frequently treated with intravesical chemotherapy, specifically Mitomycin C (MMC), following tumor resection. However, clinical response varies significantly between patients, and current dosing regimens (dose, volume, dwell time) are chosen empirically rather than based on the actual drug concentration reaching the target tissue layers. While the therapeutic efficacy of intravesical therapy depends on drug concentrations within the bladder wall—particularly the lamina propria where residual disease persists—routine care does not measure tissue concentrations. Existing pharmacokinetic models often treat the bladder wall as one or two well-mixed compartments, failing to resolve the steep concentration gradients that occur across the anatomically distinct layers (urothelium, lamina propria, and muscularis). Furthermore, while layer-resolved concentration data exist for both human tissue in vitro and in vivo, no prior analysis has successfully estimated layer-specific transport parameters from in vitro human data and transferred them unchanged to predict independent in vivo clinical tissue concentrations.

Methodology
The authors developed a one-dimensional, physiology-based partial differential equation (PDE) model to describe MMC transport across the human bladder wall. The model structure and workflow included:

1. Model Structure: The bladder wall was represented as three distinct anatomical layers:

   • Urothelium: 0–200 µm.
   • Lamina Propria: 200–1500 µm.
   • Muscularis: 1500–4000 µm.
     Each layer was governed by Fick's second law of diffusion coupled with a first-order loss term (representing local degradation, binding, and uptake). The governing equation for layer $i$ is:
     $$\frac{dC(x,t)}{dt} = D_i \cdot \frac{d^2C}{dx^2} - k_i \cdot C$$
     where $D_i$ is the layer-specific diffusion coefficient and $k_i$ is the first-order loss constant. Flux continuity at interfaces was maintained using a harmonic-mean effective diffusion coefficient.

2. Data Sources:

   • Parameterization (In Vitro): Data from Di Stasi et al. (1999) involving full-thickness human bladder specimens exposed to MMC in a passive diffusion setup. Concentrations were measured in the three layers at 5, 15, 30, 45, and 60 minutes.
   • Validation (In Vivo): Data from Wientjes et al. (1993) involving 7 patients who received intravesical MMC (20 mg/40 mL). Tissue biopsies were taken at 120 minutes at depths corresponding to the urothelium (140 µm), lamina propria (700 µm), and muscularis (~2000 µm).

3. Parameter Estimation Strategy:

   • Stage 1: All six parameters ($D_{uro}, D_{lam}, D_{mus}, k_{uro}, k_{lam}, k_{mus}$) were estimated simultaneously from the in vitro data. This revealed poor identifiability for four parameters due to correlation and low signal in the muscularis layer.
   • Stage 2: To resolve identifiability, three parameters ($D_{mus}, k_{lam}, k_{mus}$) were fixed based on the Stage 1 estimates and sensitivity analysis. Only $D_{uro}, D_{lam},$ and $k_{uro}$ were re-estimated.

4. In Vivo Translation: The Stage 2 parameter set was transferred to a clinical simulation without re-estimation. The only modification was the luminal boundary condition: the constant in vitro concentration was replaced by a dynamic urine-dilution function to simulate the clinical dwell period (120 min).

Key Results

• Parameter Estimates: The urothelial diffusion coefficient ($D_{uro}$) was estimated at $1.164 \times 10^{-7}$ cm²/s, which is approximately 84-fold lower than the lamina propria diffusion coefficient ($D_{lam} \approx 9.8 \times 10^{-6}$ cm²/s). This quantifies the urothelium as the dominant permeability barrier. The urothelial loss constant ($k_{uro}$) was estimated at $1.673 \times 10^{-3}$ s⁻¹.
• Model Fit (In Vitro): The reduced three-parameter model achieved an $R^2$ of 0.95 against the in vitro concentration-depth-time data.
• Clinical Prediction (In Vivo): When applied to the clinical simulation:
  • Urothelium (140 µm): Predicted concentration was 14.4 µg/g vs. observed mean of 11.65 µg/g (Ratio: 1.24).
  • Lamina Propria (700 µm): Predicted concentration was 7.05 µg/g vs. observed mean of 5.63 µg/g (Ratio: 1.25).
  • Muscularis (2000 µm): Predicted concentration was 3.77 µg/g vs. observed mean of 1.56 µg/g (Ratio: 2.42).
    The predictions for the urothelium and lamina propria fell within the pre-specified 2-fold criterion and specifically within the 1.25-fold range for close agreement. The muscularis prediction overestimated the observed mean by 2.42-fold.
• Sensitivity Analysis: In the urothelium, exposure was most sensitive to $D_{uro}$ and luminal concentration ($C_0$). In the lamina propria, both $D_{uro}$ and $D_{lam}$ were influential. For the muscularis, $D_{lam}$ was the dominant factor, indicating that transport through the lamina propria is the rate-limiting step for deep penetration.

Significance and Claims
The paper claims that a mechanistic, physiology-based model parameterized solely from human in vitro permeation data can predict clinical tissue concentrations in the bladder wall layers that define the NMIBC target (urothelium and lamina propria) without re-fitting tissue parameters.

• Translational Success: The study demonstrates that parameters with physical meaning, placed in an anatomically faithful structure, can translate from a controlled diffusion-cell experiment to a patient scenario. The transition required only a substitution of the boundary condition (from constant to dynamic dilution), validating the core premise of in vitro to in vivo extrapolation (IVIVE) for this specific application.
• Mechanistic Insight: The model successfully distinguishes between the barrier function (low $D_{uro}$) and the retention/clearance function (high $k_{uro}$) of the urothelium, a distinction often lost in compartmental models.
• Limitations and Scope: The authors modestly restrict their translational claim to the urothelium and lamina propria. They acknowledge that the over-prediction in the muscularis is likely due to the absence of vascular perfusion (systemic clearance) in the in vitro in vitro data used for parameterization, which the current first-order loss term cannot fully capture for the in vivo vascularized lamina propria. Consequently, the model is presented as a basis for setting regimens against the NMIBC target layers, but extending this confidence to muscle-invasive disease (muscularis) would require incorporating vascular loss terms informed by in vivo absorption data.