1. What, in your opinion, is currently the single largest trend with respect to PAT/NIR?
JC:I think one of the trends is to re-think what the appropriate role is for PAT/NIR in the fields of pharmaceutical research, development and manufacturing scheme and how the technology can be best utilized under the current regulatory environment.
RM: Continuous processing with real-time process measurements has become a huge trend as processes change from small batch process to larger scale manufacturing processing. These processes require robust instrumentation with immunity to vibration in process environments with continual internal reference and fully integrated software that is easy to use from the developer level down to the operator level.
KB: The trends are really to start early in the product development cycle to use PAT as it can bring tremendous value in process understanding. Also it is important to engage people at all levels of the organization in PAT, so that there is ownership, and it is not just another thing being passed over from R&D for example.
SD: A shift to focus more in development, whereas the initial PAT thrust seemed to be more towards manufacturing control. This ties in very well to the Agency’s shift to Quality by Design (QbD), where higher data density in development gives a more thorough basis for process and product understanding. It has always been our strategy to frontload the PAT work, because a chemistry or engineering “fix” for the process is generally preferable to monitoring. Real-time analytics will also have significant impact in continuous processing, and I think folks are again seeing the benefits to be as large, and possibly larger, in development as compared to manufacturing.
RC: The progress and interest in PAT has been extremely slow since the concept was introduced to the pharma industry with the release of the PAT guidance document in September 2004. This slow rate of change makes it difficult to identify any trends or patterns. The closest thing to a trend that I’ve observed recently is it does seem there has been an increase in interest in PAT. For example, the number of positions being posted by pharma/biopharma companies that include “PAT” in the job responsibilities and desired skills seems to be on an upward trend. There also seems to be more interest in topics such as using DOE and risk assessment tools (e.g. FMEA, HACCP) during process development in discussions on intranet sites such as LinkedIn.
RV: The FDA’s PAT Guidance published in 2004, has triggered the pharmaceutical industry to consider a more risk and science based approach for their production processes.
The first PAT applications were mostly related to moving some of the off line quality testing closer to the production line by means of real time quality measurements. However, this first generation of PAT applications was usually not interested in controlling the process based on these results. The application was mainly interested in monitoring the process in real time and stopping it when the process was drifting out of specification limits. Therefore, no true financial gains were achieved yet, apart from stopping a process during the execution of a batch up front when a quality issue is being developed, rather than only discovering the quality problem at the end of the batch execution phase.
The second generation of PAT applications started to embrace more the Quality by Design principle. This means having a more integrated approach of the online quality results and the control system, and therefore being capable to anticipate quality problems, rather than dealing with them after the facts or not continuing the run. Furthermore, by making use of these online quality measurements as an input for the control system, parametric release of the product can be achieved.
Today, we see the third generation of PAT applications with some very innovative drug manufacturers; using PAT as the enabler to achieve continuous manufacturing. Thanks to the existence of more and more powerful data collector and modeling systems, entire batch driven processes can be converted to continuous manufacturing processes. Important to successfully convert processes in the pharmaceutical industry to continuous processes is the understanding of your product’s Critical Quality Attributes (CQA) and related to that the process’ Critical Process Parameters which influence these CQAs. Therefore, it is important to really well define upfront what it is that makes your product a quality product. Only then, one can define which process parameters to control with which control strategy.
2. Are there any disadvantages or limitations in using PAT/NIR applications in the pharmaceutical industry?
JC: There are many advantages of implementing NIR in pharmaceutical industry. However, the biggest hurdle for NIR technology is that it intrinsically lacks sensitivity and selectivity compared to other tools such as IR and Raman spectroscopy. NIR absorption bands are generally broad and overlapping. To extract useful information from NIR spectral data often requires the aid of multivariate data analysis tools such as PCA and PLS. The introduction of multivariate data analysis tool to interpret NIR data presents the perception that it is done in a “Black Box”, which in turn hinders the acceptance of NIR applications. In addition, without proper training and understanding, it is possible to build multivariate models based on purely correlation instead of causality. A model built on correlation alone is very likely to fail during method transfer. Furthermore, the usage of the multivariate data analysis makes the method validation very challenging. The definition for the method figure of merits such as selectivity and sensitivity for multivariate models are not well defined and accepted.
RM: NIR is empirically based. Therefore, qualitative or quantitative models must be developed for each unique process. Because NIR is not a separatory technique and the samples are measured directly in their natural matrix changes in formulation, for instance, can have a great effect on predicted results. Once the models are developed they work well on routine measurements as in QC or process control but are not useful for identifying or quantifying samples dissimilar from those in the models.
KB: In choosing any PAT tool, it is important that the tool be appropriate for the measurement. There have been instances of people trying to develop applications for which NIR is not best suited (metals analysis in soils?). As long as the system under study has relevant absorbances in the NIR spectral region, and the instrument can be interfaced to the process to give representative sampling NIR can be used. It does have the disadvantage of not being capable of trace level detection (ppms), but has many applications for which it is a good choice.
SD: There seems to still be some uncertainty in industry about harmonization and consistent interpretation by regulatory bodies. The disadvantage may be that the industry has historically kept things close to the vest; there is not an environment of shared learning and full disclosure. This is obviously changing, for example in the appearance of numerous white papers and faux submissions, but for regulated industries such as ours this serves to slow down a process that is by its nature somewhat sluggish. The other issue with being in a regulated industry is that for installed manufacturing processes, there are substantial barriers to implementation of PAT equipment with regard to modification of registered processes and possibly opening your facility to additional regulatory scrutiny. This seems to be the case even though the economic benefit may be much easier to determine than for new compounds just being launched.
RC: Most of the concern about disadvantages or limitations, with respect to using PAT, have to do with the possibility of increased regulatory uncertainty. Use of PAT is a new concept in the pharma industry. Use of this new technology and approach to product quality may be viewed in different ways depending on what regulatory agency, division, and/or reviewer is reviewing a regulatory submission. This concern is further magnified when the company must deal with numerous regulatory agencies in order to market the product in multiple countries. Worldwide agreement on the use of consensus standards (e.g. ASTM, ICH) would go a long way towards minimizing these concerns.
RV: Regardless of the many benefits PAT has demonstrated in many studies published, the pharmaceutical industry has been reluctant to embrace the topics of PAT and Quality by Design. The perceived difficulties of PAT applications are:
- The need for a change of mind set with employees and management
- The need for upfront investments in analyzer-, IT-, process infrastructure
- The need for revalidation of the existing process with no in-house references of a Design Space approach
- The need for multidisciplinary teams not in place today
3. Why is the use of PATs lagging behind in the pharmaceutical in comparison to other industries?
JC: The biggest difference between the pharmaceutical industry in comparison to other industries is that it is a highly regulated industry. Any deviation from established practice brings potential risks. The implementation of a new technology or practice has to be justified with significant improvement with subsequent benefits. The change takes much longer time to be accepted by both regulators and the industry itself.
RM: There continues to be a great amount of uncertainty about the FDA’s support of alternate or parametric analysis methods. At the CDER top level the initiative has support and optimism for improving risk management by gaining process understanding and control. At the regulator level the “devil is in the details” particularly with new FDA personnel coming from the pharmaceutical industry or recent graduates that may not understand the ultimate goals of PAT. They look at compendia methods as “tried and true” and are skeptical of in-process methods that may indeed yield a better analysis in the long run.
KB: Clearly the regulatory hurdles of the pharmaceutical industry have inhibited the expansion of PAT in this area. The fact that processes historically could not be changed without a large paperwork burden once validated and approved has held the pharmaceutical industry back from being innovative in their processes. Since the PAT and QbD initiatives have been launched by the FDA this has changed. But indeed it takes time not just to develop the resources, but to install a newer way of thinking in any company and industry. This is now happening, and some very good uses of PAT coupled with a thorough consideration for process understanding and control are being seen from the pharmaceutical industry.
SD: In part, timing. As described previously, there is a reticence to install PAT in existing processes; so many folks are imbedding PAT in late-stage development programs at the same time that the number of new molecule launches has gone down. In addition, it takes investment (people, equipment, and development of new work processes) just as the industry is being buffeted by a number of factors putting pressures on product pricing and margins. At the end of the day, these same pressures will help drive the deployment of PAT applications, because the commodity industries have shown us that well-engineered control systems will pay for themselves many times over.
RC: There are multiple contributors to why the implementation of process analytics in the pharma industry lags other industries such as petrochem who has by far the longest history with this technology. It’s important to keep in mind that evidence of process analytics in the petrochemical industry dates back to the World War II timeframe or about 10 times longer than the FDA’s PAT initiative has been in existence. A major factor in the use of process analytics is the type of manufacturing process utilized. The greatest implementation of process analytics is in industries utilizing continuous processing rather than batch processing as pharma primarily uses. Obviously, there is a large risk of producing huge amounts of off-spec product if a continuous process relied on off-line, lab-based testing for process monitoring and control. On the other hand, batch processing, such as used historically in the pharma industry, is much more tolerant of the process delays created by off-line testing. Process efficiency is another major contributing factor. The petrochemical industry is extremely dependent on operating a highly efficient operation; both from a profit and market demand perspective. In contrast, the pharma industry typically has little concern about manufacturing capacity not meeting market demand. The exception to this would be in the case of a widespread contamination issue shutting down an entire manufacturing site as in the case of a sterile product such as a flu vaccine. Process efficiency has an additional impact from the perspective of cost of manufacturing. The pricing of pharmaceuticals is typically based on patient value rather than manufacturing costs so process efficiency from a cost perspective is typically not a major concern for the pharma industry. This may change in the future as the industry faces increased pressure from a product pricing perspective. Probably the major reason for a lack of PAT implementation in the pharma industry has to do with culture. The pharma industry is very conservative from a regulatory risk avoidance perspective. As long as the PAT approach remains a voluntary choice rather than a requirement, it will likely continue to be implemented very slowly as few companies will be willing to implement something that is viewed as new and potentially risky from a regulatory approval standpoint even though there are measureable benefits to implementation. The current business climate impacts this heavily. With the prevalence of layoffs, M&A’s, outsourcing, etc. it is less likely that anyone will want to push for implementing a new approach to manufacturing and quality control when there is concern their job could be in jeopardy. Obviously in this kind of business environment, who wants to be singled out as the person responsible for implementing a new approach to quality control that put manufacturing at risk. Lastly, I believe there is a perception that the use of on-line monitoring and control is a less secure approach to ensuring product quality than off-line, lab-based testing of the finished product. This argument seems weak; particularly when one considers the amount of product that is tested off-line with respect to lot size. It becomes an even more difficult position to defend when the typical pharma process variability is factored in.
RV: Two main reasons why PAT implementations are slow:
- The pharmaceutical manufacturing industry is by definition a risk averse and traditional industry which, due to strict regulations, was not able to change anything in the process control or release strategy once the three reference batches were produced and the process got locked in.
- The pharmaceutical industry has long been privileged not to have to look into optimizing their manufacturing processes as product margins were high. Pharma CEOs were not lying awake at night wondering how to get more efficiency inside their production plants, but they were missing out on sleep wondering how to get the next blockbuster out in the market as soon as possible. Today, the pharma CEOs must be really restless, as the blockbuster model is not being met anymore and therefore profit needs to come from elsewhere too, the manufacturing facilities are being looked upon as cost cutting targets. Slowly, PAT and QbD solutions are finding their way into process optimization programs.
4. How, in your mind, has the landscape of PAT/NIR research shifted from when the technique was first introduced to how it is being implemented today?
JC: The research field for NIR in pharmaceutical industry has expanded exponentially from early applications such as determining moisture content and raw material ID. Nowadays, NIR is widely used in both drug substance and drug product stages, mostly for process understanding and control of unit operations such as chemical reaction, isolation, distillation, drying, crystallization, granulation, blending homogeneity and content uniformity.
RM: More attention is being directed toward continuous processing and in-process measurements rather than just replacing wet chemistry methods with faster, low sample prep NIR methods. Granulating, drying and coating as well as at-line content assay methods are being studied and published. On the biologics side, investigators are using in-situ NIR to better parameterize cell cultures and fermentations.
KB: NIR continues to be heavily used in the agricultural industry, with continuous innovations in instrumentation and in chemometric methodologies (including networked systems), which contribute to the value of NIR. Now there are numerous NIR instrument vendors, and a larger body of knowledge in how it can be used. We have learned from past experience, and are conscious of some of the mistakes made in the past. Instrumentation is robust, commercial software with numerous tools is available and more effort in being put into understanding the measurements, and not just treating it as a black box. There are many opportunities for closed-loop control of processes with NIR in the pharmaceutical industry, as is the case in other industries already.
SD: I think the two major elements are the shift upstream to R&D concurrent to the QbD initiative discussed previously, and a more practical approach in the description of PAT. Regarding the latter, I’ve heard much more about manufacturing efficiency in the last year than in the preceding five years, so the focus of the conversation is shifting. Many of the initial presentations about PAT were either didactic or involved technologies that have no place in a manufacturing environment. It may not be as “sexy”, but I am seeing more talks about the nuts and bolts of practical installations, and concepts like ROI and NPV for their installations.
RC: My observations are that most of the research is in the area of monitoring and controlling continuous processes, developing “platform solutions”, technology miniaturization, and using PAT to control cellular processes. As stated earlier, on-line monitoring and control of continuous processes is extremely important and fairly self explanatory with respect to benefits. The platform concept is one where the PAT tool is designed to be an integral part of a standard unit operation. The unit operation is then replicated across numerous production processes. One particularly active area is the development of sensors that are compatible with single use (i.e. disposable) equipment used in bioprocessing. Examples of the platform approach are on-line HPLC to control the collection of product from large scale purification columns or NIR to determine the endpoint of a blending operation. Miniaturization, such as MEMS technology, is another area I see being explored. One use of MEMS is as a means to increase robustness (e.g. new MEMS based designs for NIR spectrometers). Another use of MEMS is “lab on a chip technology” to enable utilization of wet chemical techniques in on-line analyzers. Miniaturization is also being used to move large on-line analyzer systems towards devices more closely approximating on-line sensors such as the integration of miniaturized liquid chromatographs onto NeSSI sampling technology to enable the entire analyzer and sampling system to be coupled directly onto the process pipe or reactor. Lastly, we are seeing research groups utilizing PAT to help control cellular reactions such as using on-line HPLC to monitor and control critical media components that influence glycosylation.
RV: The main drivers for PAT research and PAT implementation have been coming from 2 main fields:
Standard and community initiatives:
- FDA’s Guidance for Industry as part of the cGMP Initiative
- ASTM E55 standard proposals
- OPC UA data communication standards
- Communities like IFPAC, PAT COP Groups of ISPE
Technology improvements:
- Measurement techniques and sensors more aligned with pharma processes
- Disposable sensors for bioprocess monitoring
- Data management systems which can handle large amounts of multivariate data every second
5. Where do you see the biggest application for PAT/NIR in the future?
RM: The biggest future PAT/NIR application is in-process instruments specifically designed for dryers, blenders, granulators and in-situ bioprocess applications. These instruments need to be ruggedly designed for the specific process/manufacturing environment versus laboratory bench top instruments. They will need to be able to withstand the vibration and harsh environments of the manufacturing area. Also, the software needs to be designed for in-process NIR dedicated interfaces.
KB: The biggest application for NIR will be that that has the greatest business value. At this time, we see that is making a large impact in counterfeit detection. I do think that PAT tools such as NIR will become more commonly used, and be part of process control systems, allowing for continuous processing with low risk of producing out of specification materials and products. As we develop a greater process understanding using PAT, we will be in a position to decrease the amount of final product testing that is currently done.
SD: I think there are a lot of avenues for potential growth and two that I see having significant impact are supporting the shift to continuous processing and providing real-time demonstration of product quality in a globalizing industry. Continuous processing is a natural fit for real-time analytics, and this is an area that has a lot of interest across pharma, for both bulk and product. With regard to product quality, when I worked in the commodities world, we could call up on-line instruments from some of our competitors that shared headers into community pipelines, and those were frequently located in remote and harsh environments. As we deal with extensive outsourcing of pharmaceutical manufacturing, it’s not unreasonable to see a time in the not too distant future where innovative companies will want to see and track product quality metrics from their contract manufacturing partners, and those who offer such services may have a competitive advantage.
RC: The opportunities for application of PAT are still wide open. One area that I think we’ll see expand greatly in the future is in the area of cellular control in bioprocesses. Historically, the process control applied to bioreactors has been focused on making the environment for growing cells favorable, i.e. increasing cell density to maximize yield. There has not been a widespread focus on attempting to control what occurs within the cell. I believe we will see further integration of on-line HPLC with mass spec, and chemometrics software to move metabolomics from the lab to on-line to enable directly influencing cellular metabolic pathways that control product expression. From a low hanging fruit, low risk perspective, there are plenty of opportunities for companies to move from lab-based measurements to field measurements such as using handheld NIR and/or Raman instruments for raw material ID testing and on-line HPLC to monitor the endpoint of CIP processes, i.e. continuous validation of the cleaning process.
This article was printed in the November/December 2010 issue of American Pharmaceutical Review - Volume 13, Issue 7. Copyright rests with the publisher. For more information about American Pharmaceutical Review and to read similar articles, visit www.americanpharmaceuticalreview.com and subscribe for free.