Right the First Time

July 1, 2004
Predictive Quality Control Is Aim of New Program at Pfizer

By Agnes Shanley, Editor in Chief

Last year, New York-based Pfizer launched a radical new program, "Right First Time" (RFT), designed to take it from a reactive approach to a predictive approach to manufacturing and product quality. Leveraging process analytical technologies (PAT) and information technology (IT), the program aims to deepen knowledge of processes, to focus on those that are "critical to quality," and to develop metrics that will allow Pfizer to improve quality throughout its manufacturing operations, documentation procedures, and laboratories. Deep process knowledge, Pfizer believes, will improve product quality, reduce cycle time and improve supply chain management. Pharmaceutical Manufacturing caught up with Robert Miller, Ph.D., Pfizer's director of quality operations for North America and Puerto Rico, to learn more about the program.

Pharmaceutical Manufacturing - "What were the drivers for RFT?

Robert Miller - We wanted to get to a point, as far as overall quality, predictability and reproducibility of processes, where no pharmaceutical company has ever gone. We wanted to take a revolutionary step, to move beyond where the pharmaceutical industry is today.

PM  - What is the foundation of RFT?

RM - A very thorough investigative process--a process that allows us to dig much deeper than traditional methods, to help us understand the root causes of variability and to define processes for preventing that variability.

PM - Is this a business process reengineering project?

RM - It hasn't been a top-to-bottom organizational change. Sites have dedicated groups working full-time on RFT. They've dedicated a lot of resources--in terms of both funding and people'stime--and work on very well-defined projects.

PM - How many facilities are implementing RFT currently? How extensively are RFT principles being applied now?

RM - Every one of the company's sites is involved in some aspect of RFT. We've started small: every site has been asked to identify from one to three RFT projects. Some sites have completed the first set of projects and moved on to others, others are still working on the first set, so the involvement varies from site to site. As the program matures, RFT tools, training and opportunities will spread out through each organization. Within the next five years, we hope to get to a point where these won't be specialized projects any longer, but the way we do business.

PM - What tools are you using for RFT?

RM - Most, if not all, of our sites have some kind of Six Sigma black belt training. We're also developing IT-based tools, such as an online system that will allow us to investigate problems or deviations in a process or a lab in a systematic fashion. We're also using a program that takes data and develops more meaningful specifications--internal criteria based on process capabilities.

In addition, on the lab side we are developing an e-compliance platform based on Velquest's software at one of our facilities which will be made available throughout the U.S. on an as-needed basis. On the manufacturing side, we're looking at programs with embedded logic to allow this same automation to be achieved in batch recordkeeping and management. We're also in the process of installing manufacturing execution systems at our manufacturing facilities.

PM - How do you go about developing "Critical to Quality" metrics?

RM - True root cause, determined through a very robust investigative process, makes it easy to determine CTQ attributes. Then, PAT and other capabilities allow you to monitor them regularly. What we're doing is developing a learning organization, where we can take a step back and ask what the dataare really telling us--to be diligent and committed to really dig down and get that true assignable cause, then put measures into to place to prevent it from recurring in the future.

PM- What successes have you seen so far, and what goals do you have for the program over the next five years?

RM - Even though it's early in the process, we've already had some successes in reducing reject batch rates and improving our understanding of many of our most challenging processes.

PM -How will RFT affect maintenance and calibration?

RM - We're working on ways to improve preventive maintenance and calibration, as well as the documentation process. We need to trend, continuously, and understand why valuesvary, so that we don't merely ask "does it pass or does it fail," but "what is the trend?" For example, are we seeing drift? With drift, equipment may still be operating within its acceptable range, but being able to read into the data and say "Gee, this may be operating within spec, but these data are trending out." I can now anticipate that, within six months for example, based on the trend, it may go out of spec.

Once you see the historical data and trends, you can then take the appropriate corrective action--perhaps work with a new vendor, source a specific type of equipment, or change the way that your facility handles preventive maintenance. The goal is to learn from what you see, rather than take an instrument reading and treat it as a distinct data point.

PM- What changes will the program make to batch documentation and related processes?

RM - We'll be eliminating non-value-added documentation requirements. Today, for example, someone told me of a case where we're putting the same piece of data, handwriting it, into four different documents. In another example, in batch records, the person said, they have to make 20,000 entries of the same type. We're all human. When you have those types of processes, the opportunity for failure is quite high.

We'll also ensure that all the appropriate data are in the record, asking "do we need to provide extra information for clarity, or can we delete some because it doesn't add value?"

PM -How much of the batch record process has Pfizer automated so far?

RM - We're just at the beginning. We haven't migrated to automated batch record-keeping as a company, although some of our sites are already doing electronic batch records. At this point they are the exception rather than the rule.

Right now, sites are focusing on simplifying paper batch records, to optimize the forms before they go electronic. And when we do go to electronic systems, I'm not interested just in electronic, but electronic with logic.

PM -What capabilities would the built-in logic add?

RM - It would check and double-check, query and alert appropriate people as needed. Take, for example, a specific batch record that says we have to operate a given blender at 25 rpm, the program would be set up so that, if you didn't put in 25, it would question you. Then, if a value other than 25 rpm were filed, the software would query and alert others that the entry is outside of filed specifications, so if that were the case, operators as well as their supervisors would be alerted for approval, if necessary.

PM- Does such software already exist?

RM - On the lab site, Velquest's package has that logic. There's nothing yet on the manufacturing side, but we believe that we can adapt Velquest;s product. We're also evaluating other products that only recently became available.

PM - Have the company's PAT projects changed the way that Pfizer handles QC?

RM - Our ultimate goal is not getting rid of traditional methods. We envision three stages to PAT implementation. First, get a better process understanding, then, once you have that understanding,identify variables critical to quality and monitor them, then, finally, replace traditional quality assurance and control.

We think that, in many cases, we won't need to go to stage three. We don't want to make QC lab extinct. We just want a greater process understanding so that when we make a batch, we're absolutely comfortable that that batch will be well within specification. And we won't have deviations. That's the real driver for PAT.

PM - What impact has FDA's risk-based validation and compliance directive had on RFT?

RM -There's been a wonderful handshake there, since, if we understand our processes, we'll know where the risks are. If we want to make a change, we'll know much, much earlier whether that change will have any impact on product quality or not. As a result, we will be able to take a change viewed today as having a high potential to impact product quality, and, based on our knowledge and process understanding, take that down to little or no risk. This ability will allow us to change the whole paradigm of validation.

After all, validation is a very discrete event--with PAT, every lot has a potential of being a validation because we'll be able to monitor many attributes within every lot. We'll really understand what's going on inside the blender or the tablet press, and monitor the parameters that will cause product variation.

Companies generally fall into one of three groups: those who truly want to know more about their processes and are pursuing PAT, those who don't care to know, and others who are waiting to see how the PAT leaders are received by FDA. Six months ago at a conference in Washington, during a discussion, someone from a big company said something to the effect that the industry can't trust FDA. We've got to grow up. We've got FDA telling us they're supportive and we've seen that support. If we don't change where we're going as an industry, we never will. It's a wonderful time, a window of opportunity that's open for all of us. Either we take advantage of it or let it close. It's up to us.

PM - What has been your biggest challenge so far, both on the technical side and as far as people are concerned?

RM - The biggest challenges are yet to come. For one thing, what will we do with all the data? We'll be able to monitor thousands of samples. Because we'll be looking so much deeper, we're re going to find deviations we've never even seen before. How will we respond to all that information and how will our specifications need to be modified? If we are not careful, we may ultimately delay QC decision making because of data overload.

Then there will be tremendous changes in who will actually be doing the quality control work and generating the data. In the future, much of this work will be done not in the lab but by operators on the plant floor. We'll have people, not only operating blenders, but looking at all the uniformity data being generated in real time. In addition, we will need people with more engineering background in quality operations, and the entire role of quality will change.

PM - Any advice for manufacturing professionals who are developing a new quality strategy?

RM - The key is a consistent message through all levels of the organization. I've been part of a few situations where we had a choice between whether to slip into "don't ask, don't tell" mode, or to work aggressively to understand our processes. We've taken the latter approach every time. People realize that management is very serious, and that RFT is the way it's going to be at Pfizer.