CPV Shines New Light on Old Manufacturing Processes

Biopharmaceutical manufacturers say that implementation of continued process verification has helped to update “archaic” manufacturing processes for legacy products and uncover problems with raw materials. They have reaped some business benefits as well.

Biopharmaceutical manufacturers say that implementation of continued process verification programs for legacy products is shining a bright light on old processes and helping to answer why some processes are in a state of control and others are not.

They also point out that monitoring systems need to be carefully designed as to not set off false alarms; and that not every shift is a problem.

Representatives from Eli Lilly & Co. and Merck & Co. Inc. offered case studies showing how CPV has helped them delve more deeply into their processes and equipment to ferret out sources of variability.

These case studies were presented at a CMC Strategy Forum on “The Evolution of the Biopharmaceutical Control Strategy through Continued Process Verification” in Gaithersburg, Md., on July 21-22. The case studies focused on insulin and a vaccine

They explained how they incorporated elements of a case study on CPV implementation developed by an industry organization, the BioPhorum Operations Group (BPOG), into their monitoring programs.

Some of the challenges of implementing CPV were also discussed as well as the merits of different trending programs.

An official from GlaxoSmithKline PLC described the manufacturer’s use of change control charts to monitor processes for 30 legacy products while another official from Baxalta Inc. warned of the danger of relying too heavily on control charts to resolve process problems.

Heather Eurenius, director of engineering with Merck’s global technical operations, said that one benefit of CPV is that it “ensures that data is routinely available and not collected ad hoc. The fact that we have all this stuff collected and archived and analyzed on a regular basis gives us a framework and process control.”

Warren MacKellar of Eli Lilly said that another benefit is the constant learning that occurs. He said that through CPV, Lilly has learned how much they don’t know about their manufacturing processes. “We think we know a lot more about our process than we actually know.”

Case study to assist industry

Joerg Gampfer of Baxalta, who was speaking on behalf of BPOC, said that it was important to develop the case study to agree on a “generally accepted roadmap” to speed the pace of CPV implementation.

Gampfer said that the goal of CPV is to answer two main questions: what is critical and how can this be controlled?

“The main question in continued process verification is do we have the right answers to these two things: Are we identifying everything that is critical? Or are there other things that are critical? Or are we seeing things that we assume they are critical but are not? Are the control mechanisms that we have the right ones?”

The case study was released in July 2014, and was prompted by FDA’s process validation guidance issued in January 2011. A pharmaceutical expert previously noted that the industry is intimidated by the sheer magnitude of implementing a huge program like CPV (Also see "Overcoming Fear of ‘CPV Monster,’ Drug Makers Launch Continued Process Verification Programs" - Pink Sheet, 29 Apr, 2015.).

CPV is the last stage of process validation, and requires manufacturers to develop more frequent sampling and testing of products that are on the market than is currently the case but leaves the details up to industry to figure out.

CPV is generally considered to be divided into two stages: Stage 3A is the initial post-commercialization period of enhanced sampling while Stage 3B is viewed by many in the industry as an optimized level of monitoring for the remainder of the lifecycle (Also see "Pharmaceutical lndustry Proposes Ideas for Number of Batches, Monitoring Production" - Pink Sheet, 28 Feb, 2013.).

The recommendations from the case study were based on a mock CPV plan for a typical cell culture production process for a fictitious monoclonal antibody product initially described in the A-MAb quality-by-design case study developed in 2009 (Also see "New Ways to Ascertain Critical Attributes of Monoclonal Antibodies Explored" - Pink Sheet, 1 Jul, 2009.). The initial study was based on identifying critical quality attributes, or CQAs.

The case study was developed by Amgen Inc. , AstraZeneca PLC , Baxter International Inc. , BMS Laboratories Ltd. , Biogen Inc. , Genzyme Corp. , GlaxoSmithKline PLC., Novartis AG, Janssen Biotech Inc. , Lonza Biologics, Merck & Co. Inc. , Patheon Inc., Pfizer Inc., Eli Lilly & Co.and Genentech Inc.

The case study recommends that manufacturers accumulate around 30 batches to set action limits based on statistical significance.

It recommends that manufacturers establish a written CPV plan as well as a CPV monitoring plan.

The CPV plan should reflect the risk assessment of the potential CQAs impacting the manufacturing process and how these risks are to be controlled. This plan then feeds into a monitoring plan that contains the trending and monitoring used to control the CQAs.

The case study further recommended that the CPV plan be product specific and generally facility specific. In addition, a cross-site review is recommended where the same product is manufactured at multiple facilities.

The plan should be available to regulators upon inspection and need not be submitted.

Elements of case study used

Eli Lilly has used elements of the BPOC case study in developing a CPV plan for a legacy insulin product, explained Warren MacKellar of Eli Lilly.

MacKellar explained the company’s evolution from 30 years ago “when no one knew what a CPP or a CQA was” to today where full-scale validation is performed for all new products and some legacy products.

He said that over the past six years Lilly has been going through an “insulin technical agenda” in modernizing its processes, facilities, procedures and analytical procedures. Lilly is also changing the way verification is done as a result of the 2011 guidance.

He described how validation has evolved from a one-off activity to a constant state of monitoring.

“In the past we would do some work and we would validate and then we would just walk away from it. With CPV this is a continual learning curve.”

MacKellar explained that the huge size of Lilly’s insulin operation was initially a challenge in devising a CPV program.

The insulin facility in Indianapolis totals 1.38 million square feet, or “18 soccer fields,” with 5,530 tanks, 6,850 pumps with a fermentation and purification distribution center. The plant is open 24 hours a day. “You can have thousands of data points you can collect in a CPV program, which doesn’t make a lot of sense.”

The company now has more insight on its manufacturing processes with CPV.

Yet he said that there is one given fact with producing insulin: there are always changes in raw materials. “I can assure you that with over 30 years of manufacturing that this happens all the time with our raw materials.”

Worth the cost

MacKellar explained that the first step to implementing CPV was establishing a CPV plan with a control strategy and then setting up a monitoring plan.

He said that “one of the things that the newer molecules have that we did not have with some legacy models is the development of a good control strategy up front. To really have a good CPV program you need to have a good control strategy.”

The control strategy has four components: processes, equipment, analytical and people.

Data integration is a key part of the control strategy for new processes. Lilly is using Discoverant software from Biovia, San Diego, Calif., to aggregate data from its SAP enterprise application software and its laboratory information management systems into one central data repository.

He said that “what we have is a wealth of data. In fact we have too much data. Unfortunately not all of that data has been archived. But we have at least 50 years of data.”

After this data is collected, there are “routine data reviews by our process teams. This is happens daily. They look at trends in the data right then before things happen.”

MacKellar said that investing in this new data management system was not cheap but it was worth the cost.

“We have actually had operators save lots of material on the floor that actually paid for the system. If you save a couple of lots of insulin not going down the drain it will help pay for a system like this. The value is immeasurable.”

Another part of the controls strategy is validating every step used to make the insulin.

This includes validation of fermentation, precursor isolation, initial purification, drug substance purification and the drug substance.

The CPV program is incorporated into Lilly’s quality system and integrated into the firm’s preventative maintenance and calibration program, the annual product review, change control, deviation management and the periodic review of facilities, utilities, equipment and computer systems.

The next step after designing the control strategy and CPV plan was coming up with a monitoring plan.

The plan has a list of parameters such as reaction start pH and final yield. It also includes the type of monitoring method used such as batch production record/control chart, control chart or the data historian.

It also lists who is responsible for collecting the data as well as how often the process should be monitored, either quarterly or by each batch. Lastly it also includes the additional data required for that parameter such as the quality target product profile.

Success stories

MacKellar said that CPV has helped the company uncover problems with raw materials as well as situations where certain raw materials are no longer necessary.

In the first case, he said that CPV was used to uncover the root of a problem with a buffer component that had been used for 20 years.

“We’ve never had a problem with that buffer component. Suddenly we had a new adduct showing up in our insulin. We talked to the manufacturer of the raw material. They said they never changed it. They did. They just didn’t know what it was. Continual monitoring will show you where you may have problems.”

As a result of this problem Lilly had to change some of the specifications for their raw materials. Adducts disrupt the biological activity of insulin.

In the second instance, after doing a validation exercise for fermentation, officials found that they were using the same media component for 30 years, but were not sure why it was being used.

“We had media components in our fermentation that we really didn’t know why they’re there. They may be in someone’s notebook someplace. We’re not really sure. But when we looked at the yield in the fermenter it doesn’t really do anything so why not drop them out? They are expensive and you have to have scrubbers on the roof to take out materials that they degrade to. Let’s get rid of them.”

Drinking from a fire hose

Merck is using CPV to improve the manufacturing process for a legacy vaccine and to understand the sources of inter-plant variability for automated cell counters, said Merck’s Heather Eurenius.

Eurenius said that “the product has seen growth since 2000, so in addition to the launch facility and doing a capital expansion we built two additional facilities and added manufacturing staff … it is a very hands-on manual intensive process and each facility has their own engineers.” About 700 people work at the three facilities.

She explained that “when it was developed it was an older generation platform so the process itself is quite archaic to say the least.”

The vaccine has been on the market for 20 years, “long before QbD.”

“Imagine having three facilities with data sets with multiple CQAs and CPPs, then add process parameters to that and then routine process parameters. Trying to manage all of this data is like drinking from a fire hose.”

One CQA Merck focused on was execution variability. Merck’s team wanted to reduce the variability in the automated equipment used to count cells. They wanted to understand why there were major discrepancies between the two automated machines producing cell counts.

They trended the data on live cell counts from two automated cell counters. There was a “phantom shift in the data on viable cell counts in the data set from the second” and there was a wide discrepancy in the results produced by the two cell counters.

Cell counts very important

Eurenius said that cell counts are very important and “determine the quantity of the virus addition.”

It turned out that one of the cell counters was taken out of the facility for maintenance in December but was never recalibrated upon its return.

“It was probably put in someone’s car but was never recalibrated so this is why one [cell counter] was producing faulty cell counts.”

“Through this observation from CPV we were able to update the instrument procedure and to do a calibration control check to make sure that the machine is producing the same results as its partner.

The lessons learned are that “you need to be as vigilant with the automated system as with the manual system that it is designed to replace.”

Benefits of real time monitoring

A key part of CPV monitoring is being able to measure whether a manufacturing process is in a state of control, and being able to react to dramatic shifts quickly, said Brian Nunnally of Biogen.

Yet he said that monitoring programs also need to have effective controls in place so that not every shift is deemed a major problem necessitating alerts.

He said that like Lilly, Biogen uses Biovia’s Discoverant software to integrate data from disparate databases and paper records such as laboratory information management systems and batch records into a centralized database. Biogen is in the process of converting new drugs as well as legacy products onto this platform.

The software aggregates this data and produces control charts trending the data in real time to detect process shifts.

Nunnally said that it is important for firms to set appropriate parameters first so that analysts are not tracking every parameter and sending out red flags for every alert.

“The current system still requires an analyst’s judgment. Until we start talking about artificial intelligence it will be hard to get past some type of analyst intervention.”

“You can automate some of these steps and try to take analysts’ judgment out of loop at some occasion. Not all violations are important.”

He also said that when choosing which type of control chart to use, moving range charts work better at discerning process shifts than individual charts.

“If your control charts do not having moving range control charts within them you are doing it wrong. You need to have that moving range chart because it has a lot of information.”

The moving range charts show the linkages between different manufacturing units and are better for seeing the big picture.

The individual range charts plot trends going on in one particular area, say production or the lab, but do not give the big picture.

In the scenario when only individual charts are used “production calls the lab and starts screaming at the lab and says ‘it’s all your fault we have a problem fix it. The lab has the ability to say ‘nope our control charts are fine. Something is going on in manufacturing.’ This works both ways.”

Use of change point analysis by GSK

GSK has implemented a monitoring plan that uses a software system developed in house to monitor the performance of 30 legacy products, said Bernhard Pasenow-Gruen. The program has been in effect for about a year.

He said that process data generated between 2006 and 2015 have been converted into the system.

The software converts the data into a statistical tool called change point analysis. Pasenow-Gruen said that this tool is capable of handling large amounts of data yet can detect subtle shifts in processes.

The software integrates data from more than 1,000 parameters and is capable of performing a full product update in less than 10 minutes.

Biotech not amenable to mouse clicks

Michael Kraus of Baxalta said that without process understanding, however, control charts are of limited value. Kraus urged manufacturers to take a QbD- and risk-based approach to process monitoring.

He said that process understanding should drive the monitoring and the monitoring should not drive the process.

Kraus said that by nature, biotech processes are highly dependent on external variables such as media components, including yeast and detergents; reaction components such as heparin and chromatography gels; and contact surfaces such as filters and skids.

These variables cannot always be accounted for in Shewart control charts.

He said that the objective of monitoring is to obtain relevant signals and not statistical ones.

“To understand whether a process is stable requires more than just a simple mouse click.”