IVIVC is a predictive mathematical model that describes the relationship between an in vitro property of a dosage formulation and an in vivo pharmacokinetic response. The IVIVC Object uses a two-stage Level A IVIVC to build and validate a model that relates Fraction Absorbed (found by deconvolving in vivo data) to Fraction Dissolved (from in vitro data).
The process starts with dissolution data (in vitro) from different formulations that have different release rates. The dissolution data for each formulation is fitted with the user-selected model to generate interpolated values at many more time points than the observation times.
Users identify each formulation as having internal or external use.
The internal formulations will be used in the process of determining the correlation model and also for internal validation.
The external formulations are not used in determining the model but are used for external validation. (The Target formulation used later for Prediction should not be used for Internal or External validation.)
The process uses the in vivo dataset collected for these formulations. One formulation is specified to be the in vivo reference data, and it should be IV or immediate release oral form, to describe the elimination of the drug. A UIR model is fit to this formulation for each subject for the selected formulation.
The in vivo data for each formulation other than the reference formulation are deconvolved with the UIR models to get Fraction Absorbed for each formulation and subject. (Other options are available for using mean profiles or using an already-deconvolved dataset.) The Target formulation must be in the in vivo data if Prediction will be done.
The correlation model is determined. If linear kinetics are assumed, the Fraction Absorbed curves should superimpose the Fraction Dissolved curves, except for possibly a scaling factor or a shift in time which are the parameters in the correlation model. The best correlation model that correlates these curves, across all formulations, is determined. The Levy plot and the Fabs vs Fdiss plot can be used to determine which model to apply. For example:
If Fabs vs Fdiss does not intercept (0,0), then use a model with AbsBase;
If Levy does not intercept (0,0), then use a model with Tshift.
If each formulation requires a significantly different scaling factor, then Level A IVIVC cannot be used.
The correlation model is validated by testing how well the model fits. Only a single correlation model has been determined even though there are multiple formulations.
The correlation model is now applied to the dissolution data for each formulation to get fraction absorbed.
Fraction absorbed is convolved with the UIR to get the predicted in vivo data for each formulation.
Cmax and AUC are computed for each formulation, to see how well they match Cmax and AUC for the actual in vivo data.
The goodness-of-fit is computed for each formulation (both the internal formulations that were used to determine the model and for the external formulations which show how good the model is for new data) and also averaged across internal formulations.
Acceptable %PE for internal and external depend on the relevant regulatory guidances.
Next is prediction, the overall goal of IVIVC, which is to show bioequivalence of new formulations based on the IVIVC model that was developed and validated, without having to collect in vivo data for the new formulations. The Target is the proven drug (e.g., innovator drug) that the user wants to show bioequivalence to. The Test data is the dissolution data from a new formulation, but no in vivo data is needed. There is no model fitting at this stage, other than fitting the new dissolution data, because the IVIVC correlation model has been determined and validated by proving it to be robust for a variety of dissolution rates.
The correlation model is applied to the new dissolution data, which is then convolved with the UIR to predict in vivo data.
The predicted in vivo data is compared with the observed in vivo data from the Target, by comparing the Cmax and AUC of these two in vivo curves and calculating prediction error.
Acceptable %PE again depend on regulatory guidances, but small %PE indicate bioequivalence to the Target.
Legal Notice | Contact Certara
© Certara USA, Inc. All rights reserved.