Reference : Bayesian Design Space applied to Pharmaceutical Development
Dissertations and theses : Doctoral thesis
Physical, chemical, mathematical & earth Sciences : Multidisciplinary, general & others
http://hdl.handle.net/2268/126503
Bayesian Design Space applied to Pharmaceutical Development
English
[en] Espace de Conception Bayesien pour le développement pharmaceutique
Lebrun, Pierre mailto [Université de Liège - ULg > Département de pharmacie > Chimie analytique >]
21-Jun-2012
Université de Liège, ​​Belgique
Doctorat en Sciences, orientation statistiques
Boulanger, Bruno
Haesbroeck, Gentiane mailto
Albert, Adelin mailto
Eilers, Paul
Govaerts, Bernadette
Hubert, Philippe mailto
Lambert, Philippe mailto
Peterson, John
[en] Given the guidelines such as the Q8 document published by the International Conference on Harmonization (ICH), that describe the “Quality by Design” paradigm for the Pharmaceutical Development, the aim of this work is to provide a complete methodology addressing this problematic. As a result, various Design Spaces were obtained for different analytical methods and a manufacturing process.
In Q8, Design Space has been defined as the “the multidimensional combination and interaction of input variables (e.g., material attributes) and process parameters that have been demonstrated to provide assurance of quality” for the analytical outputs or processes involved in Pharmaceutical Development. Q8 is thus clearly devoted to optimization strategies and robustness studies.
In the beginning of this work, it was noted that existing statistical methodolo- gies in optimization context were limited as the predictive framework is based on mean response predictions. In such situations, the data and model uncertainties are generally completely ignored. This often leads to increase the risks of taking wrong decision or obtaining unreliable manufactured product. The reasons why it happens are also unidentified. The “assurance of quality” is clearly not addressed in this case.
To improve the predictive nature of statistical models, the Bayesian statistical framework was used to facilitate the identification of the predictive distribution of new outputs, using numerical simulations or mathematical derivations when possi- ble.
By use of the improved models in a risk-based environment, separation analytical methods such as the high performance liquid chromatography were studied. First, optimal solutions of separation of several compounds in mixtures were identified. Second, the robustness of the methods was simultaneously assessed thanks to the risk-based Design Space identification. The usefulness of the methodology was also demonstrated in the optimization of the separation of subsets of relevant compounds, without additional experiments.
The high guarantee of quality of the optimized methods allowed easing their use for their very purpose, i.e., the tracing of compounds and their quantification. Transfer of robust methods to high-end equipments was also simplified.
In parallel, one sub-objective was the total automation of analytical method de- velopment and validation. Some data treatments including the Independent Com- ponent Analysis and clustering methodologies were found more than promising to provide accurate automated results.
Next, the Design Space methodology was applied to a small-scale spray-dryer manufacturing process. It also allowed the expression of guarantees about the quality of the obtained powder.
Finally, other predictive models including mixed-effects models were used for the validation of analytical and bio-analytical quantitative methods.
http://hdl.handle.net/2268/126503

File(s) associated to this reference

Fulltext file(s):

FileCommentaryVersionSizeAccess
Open access
thesis.pdfAuthor postprint44.45 MBView/Open

Bookmark and Share SFX Query

All documents in ORBi are protected by a user license.