Know Thy Process

The U.S. Food & Drug Administration has sent the pharmaceutical industry a message with its guidance document supporting process analytical technology (PAT), part of a broader initiative on current Good Manufacturing Practices (cGMP). Quality, the agency asserts, cannot be tested into products; it should be built-in or be there by design. Make it so, the agency concludes, by better understanding and controlling manufacturing processes, and regulatory oversight will come, as it's needed, based on scientific, risk-based assessments.

According to FDA's definition, PAT is a system for designing, analyzing, and controlling manufacturing through timely measurements of critical quality and performance parameters. It defines "analytical" broadly to include chemical, physical, mathematical, and other analyses used in an integrated way. Available PAT tools include multivariate data acquisition and analysis, analyzers and other systems to monitor and control processes and production end points, and continuous improvement and knowledge management capabilities.

The goal, the agency adds, is to design and develop processes that consistently ensure quality products. PAT has been used for many years in other manufacturing industries but has yet to penetrate deeply into pharmaceuticals, where rigid production processes are validated and, once approved, seldom changed. Drug manufacturing also tends to use off-line and end-of-line testing--that is, checking the final product for quality in a lab and then releasing it for shipping.

PAT involves on-line, in-line, or at-line measurements at critical processing points for feedback and control. It is expected to improve efficiency and manage variability while preventing rejects and waste, to reduce errors and increase safety, and to yield shorter production cycles and possible real-time release of the final products. Drug developers are starting to work PAT into their processes, but it has yet to become a factor in regulatory filings.

"It does require a significant mind shift for the industry as well as for FDA," says Jeff Dudley, manufacturing director for Dowpharma. Since Dow has been using PAT for decades in its non-cGMP operations, Dudley believes a transition to using it in pharmaceutical production will be relatively easy. But overall, he adds, "this kind of change will require the training of people both at FDA and in industry, and the development of operating procedures and quality assurance practices."

DRUG PRODUCERS aren't yet required to use PAT, and there is some uncertainty and concern about what regulatory action will be triggered when they do employ it. The industry is hesitant to take "that first step without understanding completely how FDA will react to it," Dudley says. "The real key will be FDA's willingness to put some clear boundaries as to their intent: Do they mean process optimization through this technology, or do they mean process tweaking and 'don't wander very far from where you are?' "

The agency is also proposing a major shift in regulatory oversight. It has faced an increasing burden on its resources and difficulties keeping up with inspections. According to the agency's Office of Pharmaceutical Science, there's been a growing trend in manufacturing-related problems and an increase in product recalls, the latter of which rose from 176 in 1998 to 354 in 2002. FDA believes, however, that better directed regulation will improve the situation.

FDA anticipates an "inverse relationship between the level of process understanding and the risk of producing a poor-quality product." Thus, if manufacturers can show high levels of understanding, the agency says it will be less restrictive in its approach and more flexible when it comes to change. Such scientific, risk-based regulatory decision-making is also designed to serve as a framework in which drug producers can innovate. But some observers wonder how flexible the agency really will be at a time when it is under tremendous public pressure and its activities are under scrutiny.

The current regulatory environment and the newness of using PAT in drug product applications generate some uncertainty. Despite this, "the consistent message coming out of large pharma is that they are committed and engaged with people, money, and strategies," says Brian Davies, director for PAT at Thermo Electron. What he doesn't see, however, is as much activity among smaller firms that may be less able or willing to fund such efforts. Among the larger companies, Davies says, AstraZeneca has a factory in Germany largely controlled by PAT, although the company hasn't yet sought to license it.

Likewise, Pfizer is putting PAT systems all over its older manufacturing plants, but not relicensing, to improve processes and gain internal benefits, he says. Pfizer also has an often-cited cooperative R&D agreement with FDA to develop "chemical imaging" for process monitoring and control, but the company didn't offer any details on its progress.

Sanofi-Aventis, meanwhile, is among the first companies to work with FDA on PAT-based applications. The company selected a solid oral dosage form of an existing product and implemented PAT tools to study critical manufacturing steps including impurity levels, moisture content, and tablet uniformity. After about two years of work, FDA recently approved the company's comparability protocol detailing the planned PAT-based changes.

Financial returns that will help justify the investment for implementing PAT are there if people know where to look, Davies says. "There are benefits if they can understand where the economics and the actual processing overlap and then use that to pinpoint an area where, if they implement a PAT system, they can not only get better quality but can radically change the economics of a manufacturing process."

Matching critical process attributes and the appropriate analytical technologies is key. So is making sure that the critical process parameter actually relates to final product quality and can be controlled. Various organizations, including the Pharmaceutical Research & Manufacturers of America, American Association of Pharmaceutical Scientists, Center for Pharmaceutical Processing Research, Center for Process Analytical Chemistry, and International Foundation Process Analytical Chemistry, have efforts to address these and other issues arising around PAT. Conferences, including FDA-sponsored meetings and workshops, have become common.

Instrumentation companies--such as Thermo, Bruker Optics, Mettler Toledo, and others that supply much of the equipment being employed--have become active, offering technical and consulting support. Thermo's Davies, for example, brings experience as a chemist at GlaxoSmithKline. Process chemists involved in the manufacturing of active pharmaceutical ingredients (APIs) have long been comfortable with process analytics in monitoring pressures, temperatures, pH, and other reaction conditions, he says, "because that's how you run a chemistry plant--they just haven't called it PAT."

Much of the push for PAT is in secondary manufacturing, such as formulation and crystallization, where it's hoped that analytics will give insight into poorly understood and poorly controlled processes. And here, near-infrared (NIR) spectroscopy is an established technology. It handles powdered and solid materials extremely easily, with no sample preparation or chemistry needed, and provides pertinent chemical and physical information, Davies explains.

Numerous examples of process chemists conducting PAT-based studies are appearing in the scientific literature. Organic Process Research & Development has just published a special feature on PAT in its May/June issue. Schering-Plough scientists, in an earlier paper (Org. Process Res. Dev. 2004, 8, 488), demonstrated the use of IR measurements to monitor and control supersaturation to get the desired crystal size of an unidentified API in development.

Experiments were conducted using an in situ Mettler Toledo ReactIR probe linked to an automated reactor to measure concentration. Crystal size was monitored by laser backscattering with a Lasentec Focused Beam Reflectance Measurement (FBRM) probe. Both real-time monitoring systems fed information back into the automated reactor's computer to control cooling rates. The same techniques, the researchers concluded, can be modified and applied to evaporation as well as other crystallization techniques.

Mettler Toledo has reviewed PAT for understanding and optimizing production-scale batch crystallization processes (Org. Process Res. Dev. 2005, 9, 348). And Pfizer researchers have constructed a PAT-based strategy for API crystallization (Org. Process Res. Dev. 2005, 9, 360). Their emphasis is on analytics along with high-throughput experimentation in process development for "measurement and understanding, rather than as a release test or control tool" in production.

Pfizer Global Manufacturing's process development group has used on-line redox monitoring of excess bromine in producing a dibromopenicillanic intermediate, as well as Lasentec FBRM particle size monitoring in the crystallization and granulation of an unidentified API (Org. Process Res. Dev. 2005, 9, 332). They call both studies an "investment in process knowledge," where real-time results led to greatly improved process robustness and "right-first-time" processing. "These examples, along with other successful PAT implementations, have helped in driving the discussion on integrating PAT into manufacturing away from strict cost-savings arguments toward understanding the process," they write.

"The most important thing is simply understanding your process," agrees Koen De Smet, a scientist in chemical development at Johnson & Johnson Pharmaceutical R&D. "Once you understand your process much better, you can also control it much better." The benefits, he explains, include reduced variability, greater reproducibility, and thus fewer downstream problems. J&J process chemists start considering PAT during development to determine if it's relevant--for example, in handling problems that often occur in scaling up reactions.

De Smet and his J&J colleagues have used Fourier transform infrared spectroscopy (FTIR) to monitor and control the enolate formation step in the one-carbon homologation process of 4-methylcoumarin with chloromethyl pivalate (Org. Process Res. Dev. 2005, 9, 356). They also have employed NIR to control selectivity in a hydrogenation step critical in the synthesis of the antipsychotic Paliperidone (Org. Process Res. Dev. 2005, 9, 344). 

THE CHALLENGES, De Smet says, include introducing analytical equipment into the process environment and getting representative samples of what's really occurring in a reactor. In the hydrogenation process, NIR results were found to correlate well with high-performance liquid chromatography (HPLC) results to predict concentrations of reactant, end product, and an undesired dehalogenated product. The results helped define a clear real-time end point of the reaction and minimize variability and were then confirmed on a production scale.

It is often easier, process chemists say, to think about introducing PAT during process and product development. Many researchers are combining in-line analytics with other development tools, such as microreactor technology (see page 43) and parallel experimentation (see page 54). In the former case, PAT is especially applicable to continuous, rather than batch, processes, and FDA encourages innovations in this area.

"Looking into calorimetry in the early stages has allowed us to catch a lot of problems," says Christopher J. Welch, a scientist in analysis and preparative separations at Merck. "And in crystallization, we are bringing a lot of analytical power to bear, including Raman, IR, and X-ray fluorescence spectroscopies--tools that five years ago were infrequently used in early development and now are used routinely in process research."

Welch says there is also more information flow between process R&D and manufacturing. "Generally at the process stage we know where the 'vulnerable points' are that need monitoring," he explains. "PAT is a useful analytical tool in providing a window into what's going on. And our PAT group, headed by Zhihong Ge, is highly skilled in this area." Nevertheless, practical considerations for production operations may involve "swapping out an analysis tool with something that is more robust and user-friendly," he adds.

For example, Merck researchers have used in situ NIR and IR to track events during the formation of a vinamidinium salt, a key intermediate in the synthesis of Merck's COX-2 inhibitor Arcoxia (Org. Process Res. Dev. 2005, 9, 141). They decided to move to real-time monitoring of the corrosive reaction mixture since identifying a usable HPLC assay proved difficult. Although nuclear magnetic resonance spectrometric assays were also possible, they were inconvenient for operations in pilot-plant or manufacturing facilities. The researchers found they could effectively track not only starting-material disappearance and product formation, but transient intermediates as well.

Similarly, Joel M. Hawkins, senior research fellow at Pfizer, has been exploring in situ ultraviolet/visible spectroscopy to profile transition-metal chemistry. In a heterogeneous palladium-catalyzed Heck coupling, Hawkins believes he can follow a complex catalytic process that involves the formation of palladium species, including Pd(0) nanoparticles. "The ultimate goal is to use the UV probe to understand catalyst activation, selectivity, and palladium removal at the end of the reaction," he explains, by targeting the correct form of palladium.

Safety is another major consideration. Bristol-Myers Squibb researchers developed what they call a safe and scalable oxidation process for preparing 6-hydroxybuspirone, a metabolite of the company's antidepressant BuSpar, which, like the drug, demonstrates a strong affinity for the human 5-HT_1A receptor (Org. Process Res. Dev. 2004, 8, 616). The most attractive strategy, they reported, was a one-step hydroxylation of buspirone. After optimizing process parameters and gaining a better mechanistic understanding to avoid various impurities, they added in-line FTIR process control to monitor the conversion and ensure product quality.

NOVARTIS AND SOLVIAS process chemists have recently reported improved safety of a fast and highly exothermic Grignard reaction using on-line NIR monitoring (Org. Process Res. Dev. 2005, 9, 365). Their quantitative method--developed on a lab scale using chemometric models and HPLC calibration to follow formation of the Grignard reagent--was tested successfully in a 630-L pilot reactor and installed on a 2,500-L production reactor. The method is applicable to a subsequent cross-coupling reaction with the Grignard reagent to determine reaction yield in real time as well.

Previously, Novartis and Solvias scientists had looked at using online IR and Raman analysis to avoid handling toxic test samples in the synthesis of tributyltin azide and to detect an unstable hydroxylamine intermediate, which could lead to uncontrolled decomposition reactions, in the hydrogenation of 1-chloro-2-nitrobenzene (Org. Process Res. Dev. 2003, 7, 1059).

While drug researchers around the world are exploring where PAT can be best applied, international regulatory harmonization is slowly moving forward. European regulators have been sharing the stage with FDA officials to offer views on PAT implementation and have indicated that the European Union's PAT team agrees with FDA's framework. PAT is also part of discussions within ICH (International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use), specifically in new ICH Q8 Pharmaceutical Development guidelines under review.

Many involved say PAT implementation promises to be one of the most radical changes in pharmaceutical manufacturing in decades. Modifications are already occurring all the way from the drug development stage to the manufacture of active ingredients and the final products. Although the changes are anticipated to be large, many expect gradual, rather than dramatic, shifts. "I believe it's going to be a 'walk before you can run' process that FDA and the industry work through," Dowpharma's Dudley says.