Statistical Quality and Process Control in Biopharmaceutical Manufacturing—Practical Issues and Remedies

Abstract

Statistical quality and process controls (SQC and SPC) are used for monitoring, trending, and ultimately improving biopharmaceutical manufacturing processes and operations. The purpose of this paper is to highlight characteristic features of bioprocess data and their impact on typical SQC and SPC applications, specifically control charts for individual observations (I-chart) and estimation of process performance index (Ppk). Simulated data were used in an attempt to mimic bioprocess data by inducing inhomogeneity, nonstationarity, autocorrelation, and outliers. The first specific part highlights the roles of within and overall standard deviation (SD) estimates for 3σ limits and their impacts on frequently applied sensitizing rules for control charts, i.e. Nelson's rules 1–4. The second part deals with the often-asked question of how many observations are required for estimation of robust 3σ limits. In the third part, five popular approaches for treating censored data (results below or equal to limit of quantification, ≤LOQ) were compared and their impact on 3σ limits and Ppk estimates were assessed. The final section summarizes the typical issues faced by the practitioner in the application of SQC and SPC and provides remedies for setting up robust and efficient control charts for biopharmaceutical process monitoring. Overall, this study shows that process monitoring and subsequent assessment without taking into consideration this atypical nature of biopharmaceutical process can lead to increased false alarm rates, thus impacting the batch release or even possibility of rejecting good batches.