Knowledge Management and Process Monitoring of Pharmaceutical Processes in the Quality by Design Paradigm - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 234

• Squared prediction error (SPE) charts detect deviations from normal behaviour
that are not defined by the model. These are based on the error between raw data
and the PCA model fitted to those data;
• Contribution plots provide an indication about the process variables that are
responsible for a particular deviation;
• Loading plots represent the relative contribution of each process variable
towards a particular principal component;
• Variable importance for the projection (VIP) plots provide a method to quantify
the relative importance of various variables used in the PLS model;
• Batch control charts are used to identify the time point at which the process
starts deviating from the normal behaviour on the basis of the ±3SD limits.

These charts are used in combination to detect variation and for predictive
modelling of the process. A combination of Hottelings T 2 charts and contribution
plots provides a powerful method for fault detection and diagnosis, where the T 2
chart provides an indication of the shift and the contribution plots are then used to
determine the cause of the deviation [ 14 ]. These charts also prove useful for real-
time monitoring of bioprocesses. This is illustrated by two case studies, where this
approach could save future batches by identification of equipment issues related to
lower viability in a particular bioreactor and probe failure in a chromatography
column as the cause of the out-of-control trend [ 23 ]. Score plots have been used as
a preliminary aid in identifying abnormal batches based on a model developed
from historical batches that can take into account multiple input parameters [ 23 ].
These plots also act as a fingerprint for the process and help to identify batch
evolution and progression trends [ 10 ].
Over the years, the term ''process monitoring'' has evolved from being not just
a basic review of process trends and has come to include various types of advanced
statistics. The applications of ''process monitoring'' are not limited to fault
detection and diagnosis but have proliferated into real-time monitoring, feedback
and control as well. However, the success of these advanced tools requires instant
access to process information, which mandates efficient process knowledge

management as a prerequisite.

3 Knowledge Management and Process Monitoring
in the Quality by Design Paradigm

3.1 Quality by Design

Quality by design (QbD) is defined in the ICH Q8 guideline as ''a systematic
approach to development that begins with predefined objectives and emphasizes
product and process understanding and process control, based on sound science and
quality risk management'' [ 18 ].The publication of the Food and Drug Administra-
tion's (FDA) PAT—A Framework for Innovative Pharmaceutical Manufacturing