Whereas process parameter risk quantification is conducted using mathematical calculations (Monte Carlo simulations) based on a process model, risk analysis is a semi-subjective methodology, of value for both early, as well as for late process development work. Herein we will focus on early process development, when little experimental information is available.
In spite of its seemingly “trivial” nature, such risk analysis can take a significant amount of time in some cases, without necessarily leading to valuable results.
One reason for this undesired outcome is the size of the team executing the analysis; in some cases ten scientists participate (draining company resources), in others only one scientist does the work (with the inevitable bias of an expert). These are undesirable extremes. In our experience, a diverse group of three-four scientists typically suffice, and this considered best practice.
Another challenge with such risk estimation is the methodology used; for early development, the Kepner-Tregoe™ works well, whereas for late development the Failure Mode and Effects Analysis (FMEA) is most suitable. For a variety of reasons, many prefer to use FMEA also for early development work, and a certain level of confusion exists regarding the analysis of the modes and of the effects. For the calculation of the Risk Priority Number (RPN) we need to rank the impact of the variability of a certain process parameter on one or more product Critical Quality Attributes (CQA’s). Specifically, we calculate RPN = Severity x Occurrence x Detectability. Severity is ranked based on the effect of the parameter variability, whereas Occurrence and Detectability are ranked using the mode of the parameter variability. For example: if temperature is the factor analyzed, and if increased temperature (within a certain range) can lead to an unacceptable level of an impurity (CQA), the “Severity” is ranked very high because of the effect of the temperature increase. On the other hand, when analyzing Occurrence, we must ask what is the likelihood that the temperature will increase (or decrease) during the process. The same goes true for Detectability, about which we must ask about the possibility that a temperature change would go undetected.
In part 2 of this note we will show in more detail how the RPN can be calculated.
In a future post we will also address the risk analysis of Critical Quality Attributes.
Risk Analysis for Quality by Design (QbD) (part 1)
Recent articles from the lab
Living in a “Material World”
Written by Jonathan Moseley, April 17th 2024 No relation to the Madonna song, but rather the title of a book by Ed Conway, an economist and journalist, regarding the Material World in which we live. And if like me, you have a general interest in chemistry, as well as some interest in economics and history…
Put and Take Shakes It Up
In this post I’m highlighting a couple of papers that are essentially two sides of the same coin. As a process chemist, I’ll be the first to admit that the chemistry I’m about to describe is not going to end up in your manufacturing route any time soon, but as a former card-carrying medicinal chemist…
Some Elements are More Equal Than Others
In George Orwell’s classic 1945 novella ‘Animal Farm’ the revolutionary livestock develop the slogan “Four legs good, two legs bad” to summarize their views on the farmer and other humans. As the novella progresses, we see that things cannot be simplified so readily, and so I was encouraged by the recent excellent article by Brianna…