Comparison of conventional In Vitro - In Vivo Correlation methodology with non-linear mixed effects modelling
Gaynor, Clare (1), Dunne, Adrian (1) and Davis, John (2)
(1) UCD School of Mathematical Sciences, University College Dublin, Ireland;
(2) Clinical Pharmacology, Pfizer Global Research and Development, Sandwich, England
Objectives: Many methods for establishing In Vitro - In Vivo Correlation models for modified release formulations have been proposed, which generally fall into two categories: those based on convolution and those based on deconvolution. Some fundamental flaws in the deconvolution approach have been highlighted [1] and it is clear that, the use of this method is not advisable. A convolution technique using non-linear mixed effects modelling which does not suffer from the same inherent problems has been developed [2]. This new approach should, in theory, produce superior results and the objective of this study is to quantify the extent of this difference in performance in terms of bias, efficiency and the FDA validation criteria.
Methods: A simulation study to compare the conventional (deconvolution based) methods of establishing an IVIVC to an alternative non-linear mixed effects modelling (convolution based) approach was undertaken. This involved repeatedly simulating an IVIVC study for which the true model and parameter values were known. In the true model the In Vivo and In Vitro dissolution profiles were coincident and intra- and inter- subject variation were incorporated in a similar manner to that suggested by O’Hara et al [2]. These simulated data were analysed using the deconvolution method proposed by Hovorka et al [3] and a convolution method based on that developed by O'Hara et al [2]. The IVIVC models established were used to predict In Vivo plasma concentrations. The predictions made using each method were evaluated by examining values of bias, efficiency and whether or not they would meet the FDA criteria for establishing an IVIVC [4].
Results and Conclusion: Our findings confirm the expectation that the convolution based method would outperform its counterpart. The concerns about the deconvolution method, in particular that it would produce biased and inefficient (highly variable) estimates are supported by the results. While these may seem like purely statistical considerations they have a very striking practical effect. The FDA guidance recommends assessment of the prediction error for both the area under a plasma concentration curve (AUC) and the maximum plasma concentration (Cmax), when developing IVIVCs. The results of this study demonstrate that, where an IVIVC relationship exists, the deconvolution method fails the FDA validation much more frequently (32%) than the convolution method (0.1%). This corroborates the statistical theory and quantifies the superiority of the convolution based approach.
References:
[1] Dunne A., Gaynor C. and Davis J. (2005) Deconvolution Based Approach for Level A In Vivo-InVitro Correlation Modelling: Statistical Considerations. Clinical Research and Regulatory Affairs, 22, 1-14.
[2] O’Hara T., Hayes S., Davis J., Devane J., Smart T. and Dunne A. (2001) In Vivo-In Vitro Correlation (IVIVC) Modeling Incorporating a Convolution Step. Journal of Pharmacokinetics and Pharmacodynamics, 28, 277-298.
[3] Hovorka, R., Chappell, M.J., Godfrey, K.R., Madden, F.N., Rouse, M.K. and Soons, P.A. (1998) CODE: A Deconvolution Program Implementing a Regularization Method of Deconvolution Constrained to Non-Negative Values. Description and Pilot Evaluation. Biopharmaceutics and Drug Disposition 19: 39-53.
[4] Food and Drug Administration (1997) Guidance for Industry: Extended Release Oral Dosage Forms: Development, Evaluation, and Application of In vitro/In vivo Correlations.