This article was printed in CONTROL's June 2009 edition.
By Rich Merritt
The pharmaceutical industry has taken a beating during the past two years.
Generic drugs, particularly from India and Southeast Asia, are putting price pressures on domestic manufacturers. As patents run out on today’s drugs, more generics will flood the market.
Consumers are getting more vocal and less willing to pay prices that U.S.-based drug companies charge. If President Obama gets his national health care program through Congress, most likely it will include limits on drug prices. This will further cut into drug company profits.
In addition, major quality issues have plagued the drug industry over the past two years, severely damaging its reputation. On April 20, 2009, the Associated Press (AP) reported that some drugmakers are releasing chemicals such as codeine, antibiotics and sex hormones into drinking water.
Finally, hanging over the heads of pharma is the Process Analytical Technology (PAT) initiative, a Sword of Damocles that is one stroke of a government pen away from changing from an advisory to a requirement.
So, the pressure is definitely on pharma to adopt PAT. Meanwhile, while pharma has been dragging its heels on PAT, other industries have been using PAT for years—whether they realize it or not.
PAT’s Not a Rule—Yet
The PAT initiative requires that information from process analyzers be used to control processes in real time. It was developed for the pharmaceutical industry by the U.S. Food and Drug Administration (FDA), which, everyone fears, will actually enforce it some day. So far, it’s just advisory, and not a regulation. The AP story might be the final straw that makes the FDA require drug companies to track emissions with analyzers and integrate those results into the process control system.
Many of the pharmaceutical PAT applications we read about in the trade press aren’t doing process control; they’re just gathering information from analyzers. Therefore, many of those who think they already are doing a PAT application do not; they’re just gathering data too.
Conversely, many users in the chemical, petrochemical and other industries deny they’re using PAT, but they’re acquiring data from process analyzers and using it to control processes. So, they’re performing PAT-like functions in the true spirit of PAT without even realizing it, and without saying so.
The real kicker is that the chemical and petrochemical industries are far ahead of pharmaceutical companies in their use of real-time process analyzers for control purposes. Pharma could learn a lot from other analyzer users, but pharma has its own agenda, economic reasons and political considerations that keep it from embracing PAT.
PAT or Not?
Even when a drug company is doing real-time control with a process analyzer, it’s reluctant to call it PAT. Or maybe sometimes it’s difficult to differentiate between PAT and just good engineering practice. For example, Raylo Chemical division of Gilead (www.gilead.com) in Edmonton, Alberta, Canada, uses on-line analyzers to ensure quality of its active pharmaceutical ingredients and advanced intermediates (Figure 1).
“If our process is off even slightly, the cost implications are significant. A small 1%-2% increase in the target commercial yield is translated as 100% pure profit. Likewise, a consistent loss of 1-2% of the commercial target yield translates as lost profit. You don’t stay in business long with this type of performance.”
Measuring pH is critical to the process, so Pastushak selected a DolpHin pH sensor from Foxboro Measurements & Instruments (www.ips.invensys.com). To ensure that the product comes out of solution with the proper pH, Raylo typically dilutes the organic mixture with water. This mixture must then be measured for pH and adjusted until the right balance is achieved. To adjust the pH, Raylo pump-circulates the solution through the bottom of each vessel to the top, where the sensor measures pH in a slurry loop. Based on signals from the pH sensor, pre-made caustic or acid solution is added until the right level is achieved. If the pH remains stable for two minutes, the pH test for any secondary crystallization level ensues. The pH analyzer provides reactive, real-time pH measurements, which are key to reducing cycle time.
“The results have been consistent from batch to batch,” says Pastushak. “As soon as we add a solution to adjust pH, the probe responds immediately and provides the new pH reading. We’ve found it to be accurate to ±0.03 pH, which is well within our target limits.”
Raylo, however, is only one of a few pharma companies that will admit to using real-time analyzers to control a process.
Where’s PAT?
As Jim Montague reported last year in Control, “There still don’t seem to be more than a half-dozen drugmakers with genuine PAT projects underway,” (“Is PAT a Silver Bullet?”, June ’08, p. 54). A recent article in another publication uncovered only one more new application, in biotech.
Marcus Trygstad, technical consultant, Advanced Measurement Solutions, Invensys Process Systems (www.ips.invensys.com), says, “The biggest problem with PAT so far has not been in concept, but in implementation. Though some companies have embraced the PAT concept fully and integrated a number of systems that enabled them to pull statistics off their process, they haven’t really designed their systems to respond to those statistics. So these companies are applying PAT methodologies to gather data, but the data has nowhere to go; it doesn’t feed into any process design or improvement loop.”
Gawayne Mahboubian-Jones, product development manager at U.K.-based PAT system integrator.Optimal Industrial Automation (www.optimal-ltd.co.uk) agrees. “Most of the companies making claims of doing PAT are not doing PAT at all. They’re either just using it as camouflage for a process that has changed very little, or they’re in the very early stages and only have the measurement and analysis phases in place. That is, they haven’t taken the all-important step of applying control on the basis of the measurements.
“The most common reasons for this are unwillingness to face the perceived regulatory risk or a belief that the cost savings aren’t worthwhile. The first is rooted in fear of the regulators, which is still endemic in the pharmaceutical industry, and the second is often a result of the huge margins which are common in much of the industry. Real PAT is still a rarity in the pharmaceutical industry and will remain so until the financial pressures on the industry force the decision makers to overcome their fear of the regulators.”
Marcus Tennant, product manager at Yokogawa (www.yokogawa.com/us/), adds, “The risks of introducing new technology into a validated, high-margin manufacturing process are perceived as too high.” He says this usually applies to any drug still on patent, where profits are huge, so pharma companies have no incentive to be efficient or improve quality. “As drugs go off patent and margins shrink, manufacturers may consider PAT techniques in initial process design, and incorporate PAT into the validation process when manufacturing is shifted to a lower-cost facility or when a generic manufacturer considers producing the product.”
In other words, pharma makes so much money off patented drugs, it doesn’t need to be efficient. It will consider using PAT only when forced into being more efficient by competitive pressures—which are mounting.
“When compared to other high-technology industries using common manufacturing benchmark indicators—such as defect rate—pharmaceutical manufacturers rate poorly,” says Frank Offenbacher of Honeywell Process Solutions (http://hpsweb.honeywell.com/Cultures/en-US/). “But while PAT has generated widespread discussion and built momentum in the industry, the application of PAT tools in commercial manufacturing is still quite limited.”
Mahboubian-Jones adds, “While the electronic industry is preparing for 8-sigma manufacturing processes, pharma still struggles to achieve 3-sigma.”
Mahboubian-Jones works with the FDA as a trainer of its PAT inspectors, and with Ali Afnan of FDA, who was the primary author of the FDA’s PAT Guidance document. “One battle I’ve been fighting, along with Ali Afnan,is that part of the pharma industry is trying to use the word ‘control’ as a description of the current system of ‘validating’ processes. That is, run a process open-loop three times and get acceptable results, and then say all subsequent runs under the same conditions will be in ‘control’ even if they’re open-loop. It’s complete nonsense, which the new FDA Process Validation Guidance exposes as nonsense. But the underlying practice is so entrenched in pharma that they’re simply desperate to find some way of continuing to use it, even if that means grossly distorting terms that are used with clear definitions in every other industry.”
How long will the FDA sit back passively, and let big pharma get away with essentially ignoring PAT?
PAT is Easy
One wonders what the PAT fuss is all about. Many end users have been running real-time control applications off analyzer data since the 1980s. Although they’re doing virtually everything PAT mandates, they don’t follow the PAT rules and recommendations. Or maybe they do and, again, just don’t realize it.
“I remember that analyzers, in particular gas chromatographs, started to be the solution for problems with distillation towers, reactors and air monitoring in the process industry in the 1980s,” says Helmut Bezecny, a retired control engineer with Dow Chemical (www.dow.com/facilities/europe/germany/) in Germany. “They enabled us to control the product quality to the desired degree and maximize the facility yield.”
In the early days, Bezecny says analyzer applications posed many problems for Dow. “There had to be a real big problem to solve before we’d choose a GC or similar analyzer,” he says. But they did. “This was mainly because there was no other reliable measuring method available. Since then, infrared, pH, Redox, conductivity, oxygen, nitrogen and other new analyzers have been developed with fewer application problems, are designed to work on-line and have an much lower system cost.”
FIGURE 2
Roplex Engineering in the U.K. uses Rosemount X-Stream analyzers to measure CO and butane and control a vapor recovery unit in real time. The analyzer doesn’t need an analyzer house or any other special protection, because it can be put anywhere.
“A prime consideration for these applications is accuracy and reliability,” says John Robson, Roplex’s managing director. “The analyzer must have the capability to be mounted in an open plant environment, eliminating the need for expensive environmental shelter or cabinets. The analyzer must be mounted conveniently adjacent to the sample tap to optimize speed of analysis.”
In another application, Rosemount Analytical Oxymitter 4000 analyzers are monitoring flue gas oxygen levels at an RWE nPower (www.rwe.com/web/cms/en/97798/rwe-npower/) combined heat and power plant in Barry, South Wales. Its gas analysis measurements are used to fine-tune the burner air/fuel ratios for maxiumum efficiency and reduced NOx emissions. The Oxymitter analyzer appealed to nPower Cogen because its sensor mounts directly into the flue gas duct. This eliminates the need for a sampling system and avoids resulting operation difficulties and maintenance.
Paul Gardner, site engineer responsible for nPower Cogen’s instrumentation, is pleased with the improved accessibility. “It’s far more convenient and safer not to have to send engineers right up to the head unit, where access is difficult and scaffolding is required to carry out the work. The Oxymitter local operator interface gives us a display of flue gas oxygen level and temperature, and with the sealed pushbuttons, we can carry out any calibration routines needed on the sensor from ground level.”
Does PAT Pose Problems?
In spite of other industries galloping ahead with on-line analyzers, pharma remains recalcitrant, probably because its members think they’re different from any other industry and have unique problems.
“PAT presents some unique challenges,” says Kent Lohrey, validation project manager at Optimation (www.optimation.us), a system integrator in Rush, N.Y. “The data collection activity in a PAT system is typically much greater, as data is now collected on every single item in a batch. For example, a tablet batch could be 10,000 tablets for a clinical trial and more than 200,000 for a production run. This requires data collection, handling and storage that are very different than just recording temperature, pressure, humidity and other process parameters over time. This data handling—especially the good/bad part decisions—are of critical importance and must be fully tested as that data is the basis for releasing product.”
Ed Stern, vice president at Canary Labs (www.canarylabs.com), adds, “That’s not a problem. We have several customers that collect huge amounts of data with our historian software. It’s not PAT, but there’s little difference between what they do and what pharma wants to collect and analyze.”
For example, Valero Energy (www.valero.com) uses Canary Labs’ software to collect more than 70,000 process variables from 17 different process units, and archives them on two centralized historian computers. Three months of one-second data, one year of one-minute data and five years of one-hour historical data are available to users online. Off-line, high-resolution data is available back to the early 1990s.
Through the improved visibility into plant operations, process engineers have been able to increase productivity from 185,000 bpd to 210,000 bpd, a significant output gain and incremental revenues for the refinery. “Speed is a big plus,” says Kevin Moran, system engineer at Valero’s Delaware City Refinery, Del., refinery. “It allows us to quickly scan and find historical data we want to see. Users can access overview charts and then quickly drill down to see detailed data. The user doesn't have to be a genius to figure things out."
Gallatin Steel (www.gallatinsteel.com) Gallatin County, Ky., logs data from its caster, melt shop, bag house and pump house operations. The process data encompasses material usage, temperatures, flows, electrical usage, environmental, pumps, valves, equipment states and other parameters. The caster trend historian alone logs more than 900 tags per second. “We can’t believe the amount of historical data that we store online and then access quickly and easily with the Canary Labs Trend Link viewer,” says Dan Pridemore, electrical engineer at Gallatin.
It’s hard to imagine a pharma plant requiring more data than either of these two applications. In fact, such applications are commonplace everywhere but pharma.
A control engineer in the oil industry, who asked to remain anonymous, says that refining uses on-line analyzers extensively. “Analytical applications cover sales specifications, intermediate product specifications and regulatory compliance,” he says. “But more applications are coming. All industrial facilities are seeing regulatory requirements tighten, and part of that is installing more emissions analyzers that analyze more species. When sulfur in gasoline dropped from 350 ppm to 30 ppm and diesel dropped from 500 ppm to 15 ppm, it suddenly got very difficult to sell product. It forced the refining industry to make large investments in hydrotreating technology and to do a much better job running them. Future specifications will force refining to pay attention to specific species in gasoline. This will further drive the application of process analyzers.”
“If the early adapters of PAT in the pharmaceutical industry want to look forward to what’s ahead in PAT technologies, they need to examine what is more quietly being done in the refining/petrochemical and chemical industries,” says Marcus Tennant of Yokogawa.
This laser-based analyzer from Analytical Specialties saves Dow Chemical $300,000 to $500,000 per year for each incinerator where it is being used to analyze and control combustion processes.
Yokogawa’s recent acquisition, Analytical Specialties (www.analyzer.com/), is working with Dow Chemical on multiple installations involving analyzers and real-time control of processes. As noted earlier, Dow has been a pioneer in using process analyzers since the 1980s. Dow uses combustion oxygen and CO analysis on process heaters, furnaces and incineration process (Figure 3). Typical energy savings are $300,000 to $500,000 per year for each incinerator. Dow is currently testing TDL(tuneable diode laser) analyzers for ethylene furnaces and other energy intensive processes. Ultimately Dow expects to install hundreds of TDL analyzers with resulting savings of more than $10 million dollars per year in energy costs.Dow also analyzes acetylene (C2H2) in crack gas. This results in product and energy savings in excess of $200,000 per year for each analyzer installed. Other installations include measuring NH3, CO, CO2, H2S and moisture in chlorine, all of which can lead to contaminated product and wasted energy, since the product must be destroyed or recycled through the process.
Regs and Help Coming
New and tougher regulations are coming for all industries. The U.S. Clean Water Act is now focusing on the build up of fats, oils and greases in sewers. Industries will need to ensure they’re not exceeding their permit limits. In Europe, regulators are going after oil-in-water. A 15% hydrocarbon emission reduction target and hydrocarbon trading have all heightened the importance of monitoring oil concentration levels in water. As a result of these new regulations, Wilks Enterprises (www.wilksir.com) reported a rapid growth in demand for its InfraCal portable TOG/TPH analyzers.
Frost & Sullivan (www.frost.com) says, “The imposition of multiple regulations related to safety, environmental issues and quality control in the chemical and petrochemical industries creates manifold opportunities for the use of different types of process analyzers. Regulations enforced by the government and associations such as the Environmental Protection Agency (EPA) call for end users to demonstrate the compliance of their products, thereby fuelling demand for these instruments.”
Some of the biggest vendors can help you out. Thermo Fisher Scientific (www.thermofisher.com) and Siemens (www.usa.siemens.com) recently combined Thermo’s analyzers and Siemens’ SIPAT PAT software into a comprehensive solution for pharmaceutical process control. ABB (www.abb.us) recently released Industrial eXtended PAT software, which combines process data collection and analysis with traditional process control and data management functions to produce a complete PAT solution in a single system. Optimal and Emerson Process Management offer synTQ PAT software to do the same job as the Siemens and ABB offerings.
Yokogawa developed a solution, Advanced Analytical Instrumentation Management System ( AIMMS) for maintenance and management of online analyzers in a process system.
“On-line blend control and optimization is most profitable when products can be certified for customers and regulatory authorities using online analyzers,” says Eric Gildea of IPS. “This requires a rigorous and auditable statistical quality control program. Invensys developed the Analyzer Management & Data Acquisition System (AMADAS) specifically for this purpose, incorporating applicable EPA and EU guidelines for waiver applications. The package can be used throughout the refinery for all online analytical systems.”
Marcus Tennant says OPC is on the PAT case, too. “As part of the OPC foundation’s UA standard, there is currently a draft specification publicly available for comment on Analytical Device Interfaces (ADI), which is a communication standard and information model for integrating on line process analyzers into control systems.”
Looks like on-line analyzers are coming to your industry. If you need a good way to acquire, analyze and manage the data, following the PAT guidelines and using PAT software might work for you.
Rich Merritt is a Control contributing editor.
