Bayesian Approach to Design Space for Drug Product Dilution and Fill

Before proceeding to the filling of syringes, the bulk material must usually be diluted to achieve a desired concentration. Also release of the final product requires that the average content or result is greater or equal to a lower specification limit (LSL).
Given the inherent uncertainty for the estimate of the content of the bulk material, there is a risk that the dilution factor derived from a point estimate can lead to under dosing the final product. To avoid this possibility, a typical practice is to apply a margin of safety by slightly under-diluting the sample to ensure specifications will be met. But this carries the risk of over-dose of the final product. In addition, the final release of the product is evaluated using the same assay, ie the release also is made with an uncertainty. Finally, the precision of the assay is itself determined with some uncertainty.
In this poster we’ll present a Bayesian model and simulation technique that take into account the three uncertainties to optimally dilute the bulk: the estimate of content of the bulk to derive the dilution factor, the estimate of the content of the final product and the estimate of the precision of the assay (repeatability and intermediate precision).
The results will be presented to minimize the risk of rejecting a lot while minimizing the risk of over dosing the final product. In addition, various sample size [number of runs and number of replicates per run] will be examined to propose a practical sample size given time constraints within the laboratories.