(657d) Uncertainty Quantification and Global Sensitivity Analysis of Drug Delivery Primary Containers to Understand Process Capability and Key Risk Factors
AIChE Annual Meeting
2017 Annual Meeting
Pharmaceutical Discovery, Development and Manufacturing Forum
Critical Quality Attribute Monitoring and Control in Pharmaceutical Manufacturing I
Thursday, November 2, 2017 - 9:15am to 9:40am
A simple high-fidelity model was created to predict the plunger position and air gap in a pre-filled syringe primary container to be used as a component in a combination product. The product-container system has six sources of uncertainty and two uncertain variables related to operating conditions. An uncertainty analysis was carried out to compute the joint distribution of plunger position and air gap performance parameters with the goal to understand the operating ranges for these CQA (critical quality attributes). In order to relate these to operational practices, process capability indices (Cpk) were calculated via a naïve approach where the marginal plunger-depth distribution was fit to a normal distribution in order to enable the index (Cpk) calculations. This simplified approach was compared with a full propagation of factor uncertainties via the rigorous nonlinear model, which resulted in a remarkable difference in the CpK values. This indicates that the Gaussian approximation was unsatisfactory and more refined techniques are required. As a final step, a global sensitivity analysis based on the Sobolâ indices technique was carried out in order to identify the key factors affecting process capability and inform continuous process improvement initiatives. Overall, four of the eight factors were found to be unessential (non-key).
This work demonstrates the use of rigorous mathematical process models and advanced uncertainty/sensitivity techniques to shed light on design and operation performance characteristics (QbD and QbC). The model and techniques were implemented in a web-based deployment platform in order to leverage the rigorous models/techniques and share these simple-to-use applications with all users across the organization, regardless of their background.