Choose Language Hide Translation Bar
TBrisset
Level I

Biological Products and Stability : Linear and Nonlinear Modeling (2020-EU-30MP-471)

Level: Intermediate

 

Thomas Brisset, Stability Platform Manager, Stallergenes Greer

 

Stability studies are a key part of pharmaceutical product development. They help justify shelf and storage conditions. By using a stability data modeling approach, the laboratory can characterize its product and perform shelf life extrapolation. This approach can help also in the definition of acceptance criteria of quantitative parameters. In the context of biological product development, we studied physico-chemical and immunological parameters using different JMP platforms - Graph Builder, Linear Model, Stability - which integrates regulatory constraints. The objective of the presentation is to explain the approach to study stability data, to highlight the different issues and to exhibit how statistical modeling can represent a decision support process.

 

Pharmaceutical Guidelines mentioned in this presentation: US FDA Q1E Evaluation of Stability Data.

 

Comments
Phil_Kay

Really nice presentation. Thanks, @TBrisset . I would highly recommend this for any person interested in how to approach Stability Analysis in JMP.

Great presentation Thomas and thank you for sharing!!

Thank you for sharing! Great presentation!

HenrietteK

Thank you for sharing this very interesting presentation.

Ben_Ingham

Great presentation @TBrisset, very insightful!

Ressel

@TBrisset, wow, thanks! That was extremely useful and I will certainly recycle much of this information in my own work. One question, please.

In this article (bullet 3.), the author recommends taking the log of the data, not time. I can see in your case, that taking the log of the data does not improve the situation for parameter #3 with regards to linearity, but Ln of time+1 does give a nice linear model. I am wondering: can one just go by trial and error and apply logarithm to either the data directly or time+1 and see whether the model approaches a simple linear regression or are there any rules to follow that define in which case one applies logarithm to the data and in which case to time+1?