A recent interview in this magazine with Rick Friedman and Karthik Iyer of the FDA included a discussion of the subject of “cGMPs and Statistics.” Friedman and Iyer point out that statistics is a necessary tool and yet many pharma organizations do not make proper use of the tool.
Many of the reasons for the slow adoption of statistical thinking and methods in the past are no longer relevant today (Hoerl and Snee 2012). The Internet, combined with effective and easy-to-use software and new learning and application methods, have greatly broadened and increased the use and effectiveness of statistical thinking and methods. These trends and their benefits are the subject of this column.
Lack of use of any effective tool can often be linked to lack of understanding of the tool, the benefits of the use of the tool and how to implement the tool. We see this when we hear questions like: “What statistical tools should my organization be using? What statistical methods do the regulators want us to use? There is a more effective way to address the need.
Better questions to ask are: 1. “How do we use data to design our products and operate our processes?” 2. “What statistical tools can help us collect and analyze our data more effectively?” Some answers to the first question include:
- Design products, processes and measurement systems
- Control and improve processes and measurement systems
- Maintain stable and capable processes
- Create robust products, processes and measurement systems
- Conduct product stability studies
Once we know what the objective is, what the need is, we can figure out which statistical tools (and of course non-statistical tools) can help us meet our objective. For example, if we want to design or improve a product or process, the concepts, methods and tools of statistical design of experiments are very useful--critically essential if your goal is to use Quality by Design approaches. If your goal is to maintain and sustain stable and capable processes over time, the concepts, methods and tools of statistical process control and process capability have been shown to be effective.
People typically avoid applying something they don’t know how to use. Fortunately, we have experience to draw on. The Lean Six Sigma community, of which I have been a practicing member for more than 15 years, has some effective solutions to this problem (Snee and Hoerl 2003). I believe that we can take the learning and application approaches (not necessarily the Lean Six Sigma methodology) used by the Lean Six Sigma community to broaden and deepen the use of statistical thinking and methods in the design control and improvement of pharma and biotech products and processes.
First, it is critical to focus on the purpose of using the tool rather than the tool per se. Why are we using the tool, e.g., improve a product, control a process, etc. The chosen purpose and an understanding of the data involved lead one to the right tools to use. In the interview referred to at the beginning of this column, Iyer pointed out that, “In the end, any tool must be chosen with fitness of purpose in mind.”
Next, we recognize that in most applications there is more than one tool involved. In such cases roadmaps are available (or can be developed) for how to fit the tools together, regarding sequence and linkage; how the output of one tool becomes the inputs for one or more other tools. Such roadmaps enable the users to learn the approach more quickly, remember the approach over time, thereby speeding up project completion.
Linking and sequencing tools and integrating the tools with an improvement framework is at the heart
of Lean Six Sigma. A common fear in applying statistical tools is the use of the wrong formulas and not
getting the calculations right. Fortunately, this is much less a problem today with readily available statistical software such as JMP and Minitab. The software uses the correct formulas and gets the calculations correct.
The analyst’s job is now to collect the right data to answer the posed question (e.g., select the right experiment design), select the right analysis procedure (e.g., analyze factorial design) in the software and then properly interpret the resulting output.
Information on the tools and their use is much more available in this Internet era. The software typically has “Help” functions that provide useful information. Webinars on statistics and its use in solving the needs mentioned above are available. Universities are offering online courses. Books are becoming more and more available online. These trends will continue to grow.
It is my experience, and that of many others, that the combination of project-based learning, using roadmaps to link and sequence the statistical and other tools together, combined with the readily available, easy-to-use software, is a major contributor to the success of Lean Six Sigma projects. Scientists, engineers and other professionals learn the methodology quickly and complete high impact projects at the same time, returning
significant bottom line savings.
This experience suggests that an approach using project-based learning and tool usage roadmaps, and enabled with easy-to-use software, will broaden and deepen the use of statistical thinking and methods in the pharmaceutical and biotech industries.
Hoerl, R. W. and R. D. Snee (2012) Statistical Thinking – Improving Business
Performance, 2nd edition, John Wiley and Sons, Hoboken, NJ
Snee, R. D. and R. W. Hoerl (2003) Leading Six Sigma – A Step by Step Guide Based on
the Experience With General Electric and Other Six Sigma Companies, FT Prentice Hall,
New York, NY