Specification Limit Normalization for Nonparametric Process Capability Analysis and Its Application in Cleaning...

Area of Science:

• Pharmaceutical Manufacturing
• Chemical Engineering
• Regulatory Science

Background:

• Pharmaceutical equipment cleaning validation (CV) data is often left-censored and limited, hindering traditional statistical analysis.
• Existing methods struggle with heterogeneous upper specification limits (USLs) and small sample sizes, impacting AI model training.
• Regulatory guidelines mandate rigorous cleaning verification, necessitating advanced data analysis techniques.

Purpose of the Study:

• To develop a robust method for analyzing cleaning validation data with multiple USLs.
• To enable the aggregation and standardization of heterogeneous CV datasets for improved statistical analysis.
• To support the development of AI-driven systems for continued cleaning process verification and monitoring.

Main Methods:

• Proposed a USL-normalization method to standardize residue measurements to a unified USL_Pct = 100.
• Introduced the KDEDPonUSLND algorithm for calculating the cleaning process critical quality attribute (CQA) level upper capability index (Ppu) on normalized data.
• Developed CQAWWC_BAKEDPonUSLND and CQAWP_BAKEDPonUSLND models for AI-driven CV Stage 3 monitoring.

Main Results:

• The USL-normalization method effectively aggregates USL-heterogeneous datasets.
• The KDEDPonUSLND algorithm demonstrated improved robustness for nonparametric analysis compared to traditional methods.
• The developed AI models provide a foundation for advanced, regulatory-aligned cleaning process verification.

Conclusions:

• The USL-normalization and KDEDPonUSLND approach offers a robust solution for analyzing challenging CV data.
• This methodology facilitates the effective use of historical CV data for AI applications.
• The proposed AI-driven system enhances pharmaceutical cleaning process verification and monitoring, aligning with regulatory expectations.