Drug Makers Catch FDA’s Drift, Look to Rid Processes of Variability

FDA officials say that drug manufacturers need to have better trending and monitoring programs in place to detect process drift in drug manufacture. They note that many of the controls needed to prevent process drift are already embodied in existing GMP regulations but are underutilized.

Agency officials say that while drift is just one of five statistically defined types of changes that can happen during drug manufacturing operations, it is the one that concerns them the most.

A U.S. Pharmacopeia official said that over time, process drift can so change the character of a drug that it is no longer bioequivalent to a reference listed drug – and so perhaps additional bioequivalence studies should accompany requests for major manufacturing changes.

Sources of drift and solutions discussed

The sources of process drift as well as potential solutions were shared by pharmaceutical industry participants and agency officials at a Product Quality Research Institute (PQRI) and FDA conference on Dec. 2-4 in Bethesda, Md.

There was agreement that inadequate change management, inadequate process development, aging equipment or facilities, fluctuations in the process parameters or changes in raw material suppliers can all cause process drift.

Pharmaceutical officials say potential solutions to help manage process variability include use of quality by design and process analytical technology tools, an integrated and holistic approach to quality management, trending stability data, and having a robust change management system.

Rick Friedman, director of FDA’s Division of Manufacturing and Product Quality, and one of the meeting’s co-chairs, emphasized the importance of monitoring process drift.

“The so-what question of process drift is that it can directly affect safety and efficacy. Whether your process is properly controlled determines safety and efficacy every single day at your manufacturing facilities. An incorrect dose, poor dissolution, and high impurities are some of the defects that lead to ineffective and unsafe products.”

The consequences of not paying attention to process drift are steep, and include Form 483 reports or untitled observations, warning letters, product seizures, plant closures or consent decrees.

For manufacturers of transdermal drugs in particular, inadequate characterization of raw materials that resulted in process drift has resulted in a series of manufacturing problems, including recalls.

How to boil a frog

Process drift is a powerful force that can do things like boil frogs alive, said Inna Ben-Anat of Teva Pharmaceuticals USA.

“The boiling frog story is a widespread anecdote describing a frog slowly being boiled alive. The premise is that if a frog is placed in boiling water it will jump out, but if it is placed in cold water that is slowly heated, it will not perceive the danger and will be cooked to death. So of course you can understand why process drift is like a slowly cooked frog and if you wait too long it will be too late.”

Meeting co-chair Avi Yacobi said that PQRI decided to take on the topic of process drift at a meeting a year before. Yacobi, formerly of Taro Pharmaceuticals, and now an independent consultant and president of DOLE Pharma LLC, said that at that meeting, “the topic was that when you have process drift funny things can happen.”

The workshop delved into four areas:

• definitions and terms for describing process variation and process drift;

• the most frequent causes of variation;

• the current strategies for monitoring and detecting process variability; and

• what evolutions in industry management approaches and the regulatory environment could promote more proactive process improvement throughout the lifecycle.

Conference participants defined drift as “unintended, unexplained or unexpected trend of a measured process parameter or resulting product attribute away from its intended target value in time ordered analysis.”

How the GMPs address process drift

FDA officials say that although the term “process drift” does not appear in the GMP regulations, it is their expectation that manufacturers maintain a state of control in their manufacturing or control procedures for the purpose of detecting and correcting undesired changes.

McNally called drift the “most insidious because it can happen so gradually, like the boiled frog” while the other changes “are more likely to grab your attention.”

Grace McNally of the Office of Compliance in FDA’s center for drugs said two provisions in the GMP regulations direct manufacturers to ensure that their manufacturing processes are in a state of control.

• 21 CFR 211.100(a) calls for “written procedures for production and process control designed to assure that the drug products have the identity, strength, quality and purity they purport or are represented to possess.”

• 21 CFR 211.110(a) calls for “procedures that describe in-process controls to monitor the output and to validate the performance of those manufacturing processes that may be responsible for causing variability in the characteristics of in-process material and drug product.”

FDA’s draft guidance on process validation, released in November 2008, also outlines FDA’s expectations for drug manufacturers to keep their processes in a state of control through a lifecycle approach to process validation. (Also see "Process Validation Draft Praised Though Closer Alignment with ICH Supported" - Pink Sheet, 1 Mar, 2009.).

She said that FDA’s forthcoming final guidance on process validation will direct manufacturers to monitor their processes to detect “undesired process variability.”

Guidance adrift in debate over wording

The guidance is under review to determine if the word “drift” will be included in the final version. It was included in the draft version but left out of the final because of disagreement over definitions.

Some comments to the draft guidance pointed out that the word can have both a general meeting and a narrower meaning in the stricter statistical sense.

The statistical meaning for drift is a gradual change to or away from the target, said McNally. Other types of undesirable process are: a sudden change in the average, outliers, an increase or decrease in the variability around the average, and reoccurring cycles.

McNally called drift the “most insidious because it can happen so gradually, like the boiled frog,” while the other changes “are more likely to grab your attention.”

Why firms should hire statisticians

The GMP regulations call for the use of statistical procedures, representative sampling plans and statistical quality control criteria to detect, characterize and respond to product and process variability.

GMP requirements at 21 CFR 211.180(e) require periodic evaluation to determine the need for changes in manufacturing and control procedures. McNally said that “this is a very important GMP requirement and one that I think we under-utilize and do not pay that much attention to.”

The International Conference on Harmonization’s Q6A and Q6B guidelines also make references to setting specifications and establishing statistical testing products for monitoring process variability.

“It is known that there will be some manufacturing variability that needs to be considered,” said McNally. “A lot of it depends on how thorough the design work was in its anticipation of what the variability might look like, but it is not a given that what you predict will happen will in fact happen.”

The GMP regulations at 21 CFR 211.110(b) also require the use of “suitable” statistical procedures to evaluate drug variability.

She pointed out, however, that the agency cannot mandate the type of statistical tools that should be used. And if manufacturers do not have in-house experts to statistically trend this data, they need to hire them or contract with those that have them.

“You need to develop a strategy for trending and monitoring that is aligned with your goal. … We cannot tell you what is appropriate. You need to get the expertise and if you don’t have this in-house you need to hire it or contract for it. Practitioners of applied statistics that can apply it in the manufacturing setting will help you. This is the sort of expertise you need to have on your team.”

Manufacturers are not sampling data

Friedman concurred with his colleague that industry needs to have stronger data sampling and trending programs in place to detect process drift.

Friedman noted that “the root cause of root causes is often the failure of management to focus on minimizing unwanted variability, differences and discrepancies throughout the product life cycle. There is a need to fully implement 211.110(a), which is to validate the performance of those manufacturing processes that may be responsible for causing variability; that is our code for statistical process control.”

Friedman noted that “the root cause of root causes is often the failure of management to focus on minimizing unwanted variability, differences and discrepancies throughout the product life cycle. There is a need to fully implement 211.110(a), which is to validate the performance of those manufacturing processes that may be responsible for causing variability; that is our code for statistical process control.”

SPC is needed to measure inter- and intra-batch variability and “not waiting until the annual product review.”

Friedman said that such sampling, which is embodied in Section 211.110(a) of the GMPs, states that “the sample must represent the batch physically at the beginning, middle and end of the batch.”

In underscoring the importance of the sampling provision, Friedman said that FDA has hired several staff with advanced statistics education to look into these issues and to review whether manufacturers have adequate sampling programs in place.

“The test of a firm’s quality system is if it will promptly catch a problem in a batch versus discovering it only after it is marketed. Mistakes are, in many cases, not caught by the individual making the error, but instead through final inspection or a QC test. … If the QC sample is not enough you have not demonstrated sufficient upstream processing controls.”

Moving from manual platforms to more predictable platforms using automation and moving from open systems to closed systems are also other ways that manufacturers can control their processes better and prevent process drift.

More stability testing necessary?

The GMPs generally require stability testing after the product has been on the market three, six, nine, 12, 18 and 24 months after distribution. Only one lot per year is generally tested.

But sometimes this is not enough to detect stability problems, especially for unstable products or inadequately controlled processes. Multiple marketed lots may be found to fail specifications during the shelf life, in which case more frequent testing is often necessary, said Friedman. As a general rule, for manufacturing or stability purposes, “as the number of tested samples increases – if you have an unstable process or product – there is a higher probability of failure.”

He gave the example of a major migraine drug that was tested every month but this was still not enough to detect assay stability problems.

A female hormone product was put on stability testing every fifth lot due to major dissolution problems, and ultimately every lot pending reformulation.

New statistical standards examines variability

Friedman said that ASTM published a new standard that gives manufacturers more precise statistical tools in measuring whether their batches will pass specifications. ASTM was formerly known as the American Society for Testing and Materials.

ASTM E E2709 is a statistical procedure that evaluates the variability in the data and calculates the probability the drug will continue to pass the specification within a certain confidence level.

“For example,” said Friedman, “we would be able to say, ‘We are 99 percent confident that the probability of the batch passing the USP standard in the future is 99 percent of better.’ ”

The ASTM standard can be used to measure such things as assay, content uniformity and dissolution.

“If a batch just barely passes the dissolution or content uniformity criteria once, what is the probability of passing one to nine more times? Depending on the inherent batch variability, it can range from one percent to about 100 percent probability of passing,” Friedman said. “So a very small finished product sample is not likely to tell you if a given lot has meaningful or sporadic variability. And it is really too late, as the problem happened, and should have been detected, upstream.”

In conclusion, Friedman noted that “unwanted variability is the root cause of many problems. Know your raw material, process and product stability, and characterize these well upstream.”

How to ensure post-change bioequivalence

USP CEO Roger Williams said that process drift is a signal that a product may be “in trouble” with regard to showing continuing equivalence.

He recommended that before approving major manufacturing changes and to protect against process drift, one option would be that regulators could require manufacturers to conduct additional bioequivalence studies.

Williams noted that the reference-listed drug product “is the pioneer product that speaks to the safety and effectiveness of the product and the labeling. The transition from Phase III clinical trial material to the marketed dose form is critical. You don’t want to depart from the pioneer product.”

Williams said that an open question is that once market access is gained for both first entry and subsequent entry products, when should bioequivalence studies be repeated?

He noted that FDA answered that question in its Scale-Up and Post-Approval Change (SUPAC) guidance, which said that major changes should generally be subject to additional bioequivalence studies.

However, FDA’s recent guidance on postapproval changes does not address this issue of linking major manufacturing changes with additional BE studies. (Also see "FDA Proposes to Ease Reporting Burden for Low-Risk CMC Changes" - Pink Sheet, 1 Jul, 2010.).

A key question regulators need to address is whether the manufacturing process has changed to such an extent that it affects the continuing pharmaceutical equivalence of a product.

Williams noted that QbD generally does not require a supplement for manufacturing changes that occur within the design space. A simple approach might be to require additional bioequivalence studies if changes are made outside the design space.

The best time to prevent process drift

David Jaworksi, an FDA consumer safety officer, observed that controlling process drift is most crucial in the design and development of a product.

“If you do these things right you reduce the risk [of process drift] and if you have an effective management process and you have the right equipment you will make major strides to reduce process drift in the future. It is best to spend the time in the design and development.”

Some of the sources of process drift in the design and development phase include having the wrong knowledge base, a poor risk assessment mitigation plan, having an ineffective quality management system, an overly complicated manufacturing process, or a poor choice of equipment.

Jaworski recounted his experience working for a pharmaceutical company where he saw the consequences of doing inadequate design work.

“The more complicated the manufacturing process, the tighter the controls you have to get. You have to look at the effects of full scale. I remember in our process we were scaling it up. It worked fine in the lab using small-scale equipment. We went to this 2,000-gallon tank and when we started making a lipid mix the viscosity was so high that we actually twisted a shaft of the mixer in the tank. I mean, we never thought about that. It worked fine in the lab and it worked fine with small-scale equipment,” he said.

“We ended up with a $50,000 repair job and we had to figure out how we were going to make this now. You can’t make it the way you made it in the lab. So we had to go back to the drawing board. It got quite expensive. Those are the things you have to look at.”

In the second stage of manufacture, the qualification stage, the sources of process drift include measuring the wrong parameters, lack of effective management tools and using the wrong equipment or manufacturing steps to transfer the product.

For example firms cannot expect to transfer their product from one site to another if qualification studies are not done on new equipment.

Jaworski said, “The more complicated the manufacturing process, the tighter the controls you have to get. You have to look at the effects of full scale. I remember in our process we were scaling it up. It worked fine in the lab using small-scale equipment. We went to this 2,000-gallon tank and when we started making a lipid mix the viscosity was so high that we actually twisted a shaft of the mixer in the tank. I mean, we never thought about that.”

He described a real-life scenario where this was done and the product did not blend right.

“They had been using this one tablet press at this site and they were using a different model at a different site. They thought it would not make a difference. There were paddles on the force feeder that were round and the other one had square ones or angular ones, so the product did not flow right. And they wondered why they were having issues with content uniformity. These are the things you have to look at in the qualification phase.”

Sources of process drift in production

In the production phase, sources of process drift include equipment wear, inadequate calibration of measurement devices, inadequate monitoring of component quality, effects of process changes not adequately studied or monitored and inadequate SOPs.

“In the production phase you always have to look at equipment. Is your equipment maintained correctly? Look at your calibration devices. In some cases today you are talking about critical applications and you are using just one device to measure critical applications. Why not use two critical measuring devices on the product? If you have two products measuring the same product and if the product is drifting at least once these devices will notice that the product is drifting.”

Regarding SOPs, he noted that “in one case I had an operator who interpreted the SOP one way and then I had a different operator who interpreted the SOP another way to get a different results. Your SOPs have to be clear and concise so that there is no way that it could be interpreted differently.”

The root causes of process drift

Friedman said that there are two overall root causes of process drift: chemical degradation of the product during storage; and the variability of components.

“Some products are by nature unstable … and have a shorter shelf life, but are expected to be able to meet their labeled expiry as supported by previous estimates and predictions,” he said.

“Do you know all the material attributes that really matter that could cause a product to degrade and can you predict that before it leaves the site?”

Friedman gave some examples of how excipients can cause such chemical degradation of products and cited recent recalls over the past couple of years tied to inadequate characterization of raw materials, including the excipients malt syrup, sorbitol, magnesium stearate and hydroxypropylmethylcellulose (HPMC).

“In the malt syrup the issue was mold. Why? The malt syrup was held for a long period of time and they probably opened and closed the bags a lot … they did not secure the bags as a place that would prevent moisture permeation.”

There was another problem with use of HPMC materials that were used to bind the product together. “Many of our excipients are natural materials and can’t be changed, it is what it is. There was an extended release product that used this raw material. It was the critical excipient that determined the release rate properties of the formulation.”

The second source of variability is from components like packaging. “The packaging component supplier did not provide you with consistent packaging. The packaging operation didn’t have the right torque and closure to secure the shipment.”

Friedman said the key theme in his presentation “is that you must have stable processes as well as stable products.”

Costly implications of process drift

The consequences of not adequately monitoring the manufacturing process for drift are field alert reports, inspections and consent decrees, recalls complaints and MedWatch reports.

Hamilton said that process drift is often difficult to quantify and must be monitored closely. “I looked at my face in the mirror this morning when I was cleaning my teeth. I didn’t notice anything new from yesterday or the day before. But I did notice a lot of changes from five years ago.”

While not naming the firm, Jaworski said that in mid-2008 FDA recalled tablets because they were over-sized. The firm said the thickness was from a low-speed tablet press. In the third quarter of 2008 FDA conducted follow-up inspections.

After subsequent inspections the firm was still not able to correct these and other problems. In the first quarter of 2009 FDA recalled all the products manufactured by the site.

In the first quarter of 2009 FDA completes an inspection of all the firm’s facilities and issued a Form 483 report. In the first quarter of 2009 the firm signed a consent decree.

This had a devastating effect on the firm’s stock price, “It is pretty significant in that the stock … almost became worthless. A lot of their problems were tied to process drift in the manufacturing process,” Jaworski said.

“When we look at process drift it really is all about establishing and maintaining a high level of process capability and measuring the uniformity and consistency of the process over the lifecycle of the product, not just in one phase or another.”

How Lilly controlled process drift

Fionnuala Walsh of Eli Lilly said that a robust, integrated and holistic quality management system is one of the main ways Lilly is able to control process drift. The system, she said, “incorporates all the elements of ICH Q8, Q9 and Q10.”

An example of how this approach under Q10 is used at a manufacturing site is that the process is reviewed on a daily, weekly, monthly and annual basis by cross-functional process teams consisting of site management and senior manufacturing management.

Their fundamental responsibility is to maintain the process so that it is in control, is capable, and is continuously improving.

In addition to these process reviews the individual quality systems at the plant site are reviewed annually. These reviews include a holistic look at the quality systems including all the laboratory quality system and the engineering and utility systems to ensure a deep scientific review of these processes and systems.

She said that a slew of manufacturing problems in 2001 led to a quality management change at the company.

At the time, the company experienced a high number of process deviations, following a spate of inspections and 483s in 2001, “which culminated in a significant warning letter for the company.”

An FDA official noted at the time that the agency “did not see the spirit of continuous improvement at our plants. It took a lot of time to see what that was about. We were all about control and staying within the bars on the control chart and not really focusing on continuously improving. And if you’re not continually improving you’re going backwards. It was a very hard lesson to learn.”

As a result of these QMS changes, she noted that “there has been a very significant improvement in manufacturing” with “fewer recalls and operational deviations.”

Effective change management is key

Another way firms can deal with process drift is having effective change management systems, explained Nigel Hamilton of Sanofi Aventis.

He said that knowledge management and risk management are key enablers for effective change management, themes which are embodied in ICH Q10. He gave some examples of how firms can have effective change management systems for better managing process drift.

“I think we can definitely say that change will happen during a product’s lifecycle. There are some things that must be changed where we have no option. This emanates from either a supplier going out of business or customer complaints.”

Andres said that last year Novartis implemented a minor change on a chemical process that wound up costing the company millions to fix because of inadequate change control. This mistake, he said, was dubbed the “lemon of the year” award.

Optional changes include such things as modernizing equipment, upgrading software, improving a QC test method, or implementing continual improvement initiatives.

“Whatever we decide to do in terms of change, we have to look at the risk/benefit of each individual change. The biggest risk by far is unintended change. Small changes do not always equate to low risk. During this week we talked about the importance of numerous small changes and their combined effect.”

He gave some examples of product and process changes and how they can be controlled. A product change involving a change in excipient suppliers is a controlled change that can be managed through a formal change control system.

An example of an uncontrolled change is process drift that occur over a period of time or over a number of batches.

Seeing the process in the mirror

He said that process drift is often difficult to quantify and must be monitored closely. “I looked at my face in the mirror this morning when I was cleaning my teeth. I didn’t notice anything new from yesterday or the day before. But I did notice a lot of changes from five years ago.”

For example, raw material variability is one major source of unnoticed process drift and can give rise to process variation, even if the raw material is within the agreed specification limits. The natural variability needs to be understood and controlled.

Another example of uncontrolled change is when changes are made through fraud, accidents or errors.

For example, he recalled an incident when an operator accidentally dropped scissors into the product during processing.

To control their process and to prevent this incident from every occurring again, it was important to design a process where it would not be possible for scissors to fall into the product.

“We lost a whole batch at half a million dollars for a lost batch because we did not really design the process properly. We did not think it through.”

He stressed the importance of knowledge in the change management system for the things manufacturers can control.

Knowledge comes from experience, increased product understanding and increased process understanding.

“Knowledge is the most important thing we need. If we have increased knowledge, we have succeeded. Where do we get this knowledge from? Sometimes we get this knowledge from experience, either good experience or bad experience.”

Effective change management discussed

One way to capture process knowledge and to have better change control is for firms to have development reports. This could be especially helpful when firms are sold and history and process knowledge are lost when new QC staff comes in.

The development report can include information on the technical transfer that involves anything transferred between two production sites.

This should include validation reports, external or internal information such as OOT or OOS reports, and annual product reviews.

As part of an effective change control system, people also need to be trained and educated at all levels, whether they are operators or management. Operators, for example, need to know the background and the reasons for manufacturing changes.

“Sometimes an operator is left completely in the dark about a change that they have been running for five years. They have so much knowledge but they are not brought into the [decision-making process] or informed of the change,”

As part of an effective change control system, firms also need to look at how their processes are running one year later after traditional process validation.

“During traditional process validation you can see examples when everything was fine and then you go back one year later and then you realize that everything is not running as perfectly as it once did during validation. You have to question why that is. Is it natural process drift due to materials changes or has the equipment changed and has it been maintained properly? Did the operators change? Did the people follow the procedures in there? Or is it a lack of management interest?”

The devil is in the change control

There was a consensus at a panel discussion that inadequate change management can wreak havoc on the finished product by not examining how small process changes can affect the overall process.

Juan Andres of Novartis said that “one of the biggest enemies for me with the process is in change control. We change equipment, we change raw materials, we change filters … you have to be extremely careful with change control. I fear change control more than I fear anything.”

Andres said that last year Novartis implemented a minor change on a chemical process that wound up costing the company millions to fix because of inadequate change control. This mistake, he said, was dubbed the “lemon of the year” award.

“We make millions on this drug and the savings from the change control was $1,200 a year. … We had to spend $20 million to fix a problem caused by inadequate change control.”

The right people to review changes

Friedman said that in managing change control, firms need to have the right people in place reviewing and making the changes.

“In just one process there can be hundreds of variables. FDA can follow the USDA [U.S. Department of Agriculture] model and have one person at every single plant but the form can still fail change control. The answer is in quality systems. … You need to have the people with the right technical skills and not someone one or two years out of pharmacy school and no manufacturing background. That is not right. … You need to have the right skills set and have the right people at the table making the change. … If you evaluate without a risk mindset you may fail to focus sufficiently on the most important factors.”

Walsh concurred with Friedman on the need to have the right people involved in change control.

“I believe when you are making changes you need a really good design of experiments and a really good understanding of the process. About 15 years ago we took a hard look at changes. The smaller changes were made by enthusiastic scientists with limited experience. They now have to come before you with significant data and explain what the change is and how it will impact the process.”

Transdermal surprises

Lino Tavares of Purdue Pharma also stressed the need to have the right people in change control.

“I don’t want to seem like a broken record but if we paid attention to Lilly yesterday, do not make changes to your formulation lightly. Before you make changes make sure they are vetted by whatever procedures you have in your company and address this to keep you from getting into trouble. If you do get in trouble it will require a fairly high-level sophistication to address these things. QbD can help us do better.”

He discussed how inadequate change control and inadequate understanding of raw materials and their interactions with other drug materials led to some recalls of transdermal products.

Examples of transdermal patches either in development or on the market include drugs for depression, Parkinson’s, Alzheimer’s ADHD, anxiety, epilepsy, urinary incontinence and obesity.

Tavares also described how QbD and PAT have been used to correct process drift for these products. Purdue is the manufacturer of Butrans transdermal patch for the relief of chronic pain.

Transdermal patches are generally of two types: one contains a drug reservoir and a control release mechanism. The other is the matrix type where the drug is formulated into the adhesive and allows the skin to regulate the permeation.

Transderm Scop was one of the first U.S. marketed transdermal products for the treatment of motion sickness. The problem started when the product was transferred from the brand division to the consumer division. The new division decided to switch to a less-expensive API without evaluating how the new API would interact with the other materials in the product.

The impurity profile of the product subsequently triggered safety problems and the drug had to be withdrawn in 1994. The product was reformulated and then reintroduced in 1997.

In another case involving a transdermal patch marketed in the European Union, a reservoir type estrogen/progestin combination product ran into problems with process drift. Manufacturers had changed certain characteristics of the membrane without checking to see how this would affect the overall formulation of the product. It failed dissolution tests, which caused the product to be withdrawn from the market.

“It took a lot of effort and a lot of comprehensive design of experiments, but the company identified the root cause and corrected the problem.”

Tavares also cited another case of leaky reservoir type fentanyl patches that were recently withdrawn due to safety concerns. There were safety concerns of a Class II opioid overdose from the leaky patches. “The most likely causes for this product defect were process drift or operator error.”

Another example was in 2008 when Lohman Thereapie-Systeme AG/Schwarz Pharma had to recall its Rotigotine transdermal system because the active ingredient was forming crystals on the patch which may have resulted in a lower dose being delivered (“Drug Product Recalls in 2008: Categorized by Problem Areas,” “The Gold Sheet,” March 2009).

FDA requested that the product be reformulated. Tavares noted that process drift of overlooked material incompatibility was the likely cause.

As a result of these problems, the company changed the way it did business. Now, critical process parameters for solutions such as the order of addition, the mixing speeds, and the times and conditions are identified.

For coating and drying the critical process parameters are the machine speed, the temperature and the air flow rate. For slitting, the CPPs are the machine speed, roll width and tension. For fabrication and pouching the CPPs are the machine speed, the roll width and the tension.

For Purdue, Tavares said that “QbD allowed us to have a better understanding of the formulation and the functionality of the excipients and the stability of the formulation. It allowed us a better understanding of the manufacturing process and better control of the critical quality attributes and allowed us to have a better plan on how to deal with process drift.”

PAT and QbD used to right process drift

Denise Rivkees of Pfizer offered a case study illustrating how use of PAT and QbD helped explain why their product’s dissolution rate was changing.

Shortly after switching to a new excipient supplier, the company noticed the dissolution of the drug had changed.

Pfizer officials were not sure if the drift was caused by the new excipient, the new offsite lab, or something about the packaging or the dissolution method.

“The product was stable from 2003 to 2005 and in 2005 the product started to drift upwards. We found that the storage conditions for this excipient had changed. But it was not out of trend at this point and not OOS and we went forward and in 2008 we got a notification of change from the second excipient manufacturer. There are two excipients for this product.”

Through the use of near infrared sensors that rotated with the blender, Pfizer did a design of experiments and changed the blending times for the various APIs and excipients in the blender to see how this affected the drug’s dissolution.

They found through the DOE and using materials from both the old and the new excipient supplier that it was not the excipients themselves that affected the dissolution rate but rather it was the API particle size that affected dissolution. Through use of NIR, they were able to predict dissolution.

“The conclusion was that increasing particle size decreased the percent dissolved at 30 minutes. There were no indicators that the excipient value affected that change.” The firm asked the API vendor to adjust the particle size.

Trend monitoring to avoid process drift

Zena Kaufman of Abbott discussed how trending stability data over various time points has been used as a tool in preventing process drift.

Stability testing, she said, usually gives a good indication of what is happening to the product and whether it is drifting out of specification.

Kaufman said that “last night I got a wonderful voice mail call from one of my lab managers. After three months in his lab, and it has been a very active lab, they have not had a lab-related investigation, and he calculated that it has been about 12,000 to 13,000 tests that have been run. So what we are going to do is order 13,000 jellybeans to correspond to 13,000 tests so that the whole lab can see what it is they have done so well and so right.”

She described Abbott’s “very robust trend and monitoring program to monitor process drift.”

At Abbott, she said, “we looked at implementing my boss’s vision of no stability failures. The way to do that is you start at the top of the mountain. We want zero regulatory complaints. Why? That’s not good for our patients and our business. So to assure that I’m not sending Rick Friedman greeting cards on a regular basis, we monitor our stability data. And I will show you examples of how such monitoring has prevented stability compliance issues.”

The firm conducts trending studies of individual lots to ensure that the drug’s potency remains intact over time.

From these studies, Abbott is able to detect if any lots are in danger of going out of trend. They use three categories in monitoring stability results.

These include, from the lowest to the highest areas of concern, analytical alerts, process control alerts, and patient or compliance alerts.

An analytical alert is triggered when a single result is aberrant but within specification limits. A process control alert occurs when a succession of data points shows an atypical pattern that was possibly caused by changes to the laboratory or manufacturing process.

A compliance alert defines a case in which an OOT result suggests the potential or likelihood for OOS results to occur before the expiration date.

Kaufman noted that stability data can go out of trend for many reasons; some of these reasons can be variability in the lab, either with the instruments or with the lab personnel. A well-running lab can reduce the number of stability investigations, she said.

“Last night I got a wonderful voice mail call from one of my lab managers. After three months in his lab, and it has been a very active lab, they have not had a lab-related investigation, and he calculated that it has been about 12,000 to 13,000 tests that have been run. So what we are going to do is order 13,000 jellybeans to correspond to 13,000 tests so that the whole lab can see what it is they have done so well and so right.”

Data can also go out of trend because of variability in the test method. “At some point you do need to question your test methods if you are getting out-of-trend data. We are using hazard analysis and we are seeing the risk with these methods and how to improve on these methods.

Stability can also go out of trend if there has been a change in the raw materials supplier.

Closing the door on process drift

Joerg Zimmermann of Vetter Pharma Fertigung GmBh said that use of PAT tools such as isolators and restricted access barrier systems, or RABs, are some examples of in-process controls used in aseptic manufacture to monitor process drift. The company makes aseptically pre-filled injection systems and has offices in Germany and the U.S.

He said that in the EU, aseptic processing requirements are guided by the GMP Annex 1. The annex, he states, requires that “you need maximum product contamination control and the maximum environmental controls and the minimum operator interference with the product.”

Conventional clean rooms are still the most common aseptic processing operation, with about 2,200 lines operating worldwide. The advantage of this method is that it is cheaper to set up because it does not involve as much equipment as the other two methods, and is especially useful for terminally sterilized products.

The disadvantages are that manual processes need much more monitoring and supervision than automated processes and there are fewer assurances that the product is sterile because it is not as automated as the other two systems and there are operators involved in its manufacture.

The second most common type of operation involves isolators, which provide “the highest level of separation between the operator and the product and is the best solution for highly potent drugs.” The disadvantages are that it requires a well-running process to avoid opening the isolator. There are approximately 400 isolator lines operating worldwide.

With RABS, the third most common type of aseptic processing system, there are no open door interventions allowed and this type involves the maximum amount of automation in the manufacturing process. There are about 250 RABS lines operating worldwide.

Yet he acknowledged that no matter how hard firms try to control process drift through in-process controls, even with highly sophisticated automated systems, it is still difficult to control everything.

“If you look at all the parameters going into the process there will always be things that come up in the process. … We use glass syringes and we had an issue of glass breakage for one of our products and we found out that in the syringe manufacturing process at the glass vendor they had air flow fluctuations and this influenced the stability of the syringe. One operator left the door open 10 minutes longer than normal and this suddenly caused a spike in the line. This is one of the 1,000 or 2,000 or 5.000 parameters that go into our process.”

By Joanne S. Eglovitch