Understanding the Measurement System
Before delving too far into the failure mode data analysis, it is important to establish that the measurement systems used
to steer and measure process or product performance are capable of distinguishing a good process or product from a bad process.
Most analytical methods in our industry go through a rigorous method development and qualification. However, saying a method
is validated is not good enough to ensure the level of resolution necessary to identify the root cause of the problem. Depending
on the test or the company's quality policy, method errors of 5–7% are often deemed acceptable. This could be a problem when
looking for subtle changes in a process or problem that can be driving the observed failure mode. Understanding the variation
that is native to the method, the sampling technique, and the raw materials is crucial to deciphering what is really happening
with the process. Studying variation in a process is the quest to differentiate between common cause variation that is native
to a controlled process and special cause variation that is driving the failure mode.
Look at the error allocation between the equipment, process, method, and operators. This is often confirmed by a simple gauge
repeatability and reproducibility study that will determine whether the measurement method is capable of distinguishing good
from bad product. If the method is deemed unsuitable, then an alternate measurement scheme will be required by the team. When
conducting an analysis of your measurement system, consider the following four questions:
1. Is the measurement system stable?
2. Are you able to measure differences in the process? This is needed to be able to drive process improvement.
3. Are you able to measure differences to distinguish good results from bad? This is needed for quality control of a nonconforming
product.
4. If you can't measure the differences, then what can you do to improve the measurement system?
Understanding Performance
The abovementioned tools will have clearly categorized what key process variables drive performance and which measurement
tools you can trust to steer your process. If the conclusion from the previous steps does not answer these questions, then
studies will have to be performed to gather the required information. Before moving forward, it is important to understand
the regulatory landscape for the project. What commitments have been made and what are the requirements for demonstrating
that the product is behaving as originally intended? In this step, any studies should consider statistical sampling plans
and all process variable investigations must be statistically unbiased, i.e., orthogonal. Understanding the difference between
attribute testing and continuous variable testing will help steer the group to the correct sample size and conclusions in
terms of root cause identification.
Corrective Action
A good rule of thumb at this step is to assume that if you can turn the failure mode on and off then you will have identified
the root cause for the problem. Integrating this sensibility, actually trying to cause a failure is a key component to establishing
a standard for defensible RCAs.
Monitor Process Stability
Implementing the corrective action should move the process back to where it was behaving predictably. Establishing a baseline
for key performance metrics and establishing alert and action limits for these variables will ensure the process failure does
not reappear unexpectedly.
Summarize Results
Figure 2. The root cause analysis roadmap consists of eight steps to identify the metrics of success.
|
The final step in the process is to capture the previous seven steps in a comprehensive document that clearly illustrates
the rationale and criteria used for the decision and the conclusions made during this investigation. The final report should
definitively establish that the root cause has been identified and be able to use the RCA methodology and final data to defend
this conclusion. All eight steps are illustrated in Figure 2.
CONCLUSION
Establishing an objective, data-driven methodology is a critical component to an effective RCA exercise. You may leverage
established quality and risk management tools as part of the process to help focus the investigation on the key elements that
are critical to the observed failure mode. Maintaining the remediated process through focused monitoring of the key parameters,
which drive process stability and performance, will ensure that the conclusion drawn from the RCA investigation and the remediation
implemented will maintain processability and predictability and drive down or eliminate the compliance risk associated with
substandard or incomplete RCA investigations.
Bikash Chatterjee is president and chief technology officer and Wai Wong is director, both at Pharmatech Associates, Hayward, CA, 510.760.2456, bchatterjee@pharmatechassociates.com
|
