September 1, 2012 (Vol. 32, No. 15)

Bioprocess monitoring and control have been around since the beginning of therapeutic biotechnology, borrowing methodologies from industrial biotech and chemical/pharmaceutical processing. During the 1980s and 1990s, new analyzers arose from medical instrumentation for measuring pH, dissolved oxygen, carbon dioxide, etc. This instrumentation, in one form or another, serves every bioreactor in use today, down to milliliter-sized scaledown vessels.

This “traditional” bioprocess monitoring is of course still a very big thing, often involving some very substantial organizations.

Sartorius Stedim Biotech has incorporated multiparameter monitoring and control in its BIOSTAT® Qplus parallel bioreactor system, which uses 0.5 and 1 L vessels. Intended for design of experiment (DOE) for process optimization, the system incorporates O2 and CO2 off-gas analyzers for monitoring, O2 supplementation capability, and BioPAT® software for supervisory control.

Since DOE mimics conditions in much larger bioreactors, it demands the same analytics and control. Yet one of the knocks against continuous, real-time bioprocess monitoring is what to do with all the data. Since BIOSTAT runs up to 12 culture vessels in parallel, the data-handling system must be robust. Sartorius has incorporated a full MFCS SCADA supervisory control package with Qplus to go along with the monitoring functions.

Sartorius Stedim Biotech has addressed data bottlenecks in real-time bioprocess monitoring by including a supervisory control package within its BIOSTAT Qplus parallel bioreactor system.

From Diagnostics to PAT

An internal collaboration between Roche Diagnostics and Roche Pharma arose from the need to improve process monitoring and control during drug substance manufacturing. “Roche Pharma was struggling with the gold standard technology at the time,” says Andreas Schneider, vp for life science alliances at Roche Diagnostics. Instrumentation consisted of three analyzers from a U.S.-based international vendor that had begun as a manufacturer of clinical analyzers.

Schneider and his colleagues realized that Roche Diagnostics already possessed the technology to duplicate the performance of the three process analyzers they were using at the time. Like their vendor’s system, this technology originally served medical diagnostics and monitoring markets. “We asked ourselves, is there a possibility of modifying an analyzer designed for clinical use into a process monitor?”

After considerable development work, they came up with the Cedex Bio Bioprocess Analyzer, a continuous, random-access, multichannel bioprocess analyzer. Cedex measures up to 14 parameters simultaneously through a combination of electrolyte testing (similar to standard process monitors) and “photometric” analysis through a built-in, 12-wavelength spectrophotometer. Roche is targeting cell culture processing from cell-line development to large-scale manufacturing.

Reducing the number of process analyzers, Schneider says, will improve consistency, minimize human intervention and maintenance, and reduce technical support.

Cedex Bio hits all the standard analytes: glucose, lactate, LDH, ammonia, potassium, sodium, glutamine, and glutamate. A version with IgG measurement, and one for microbial cultures, will be available in future upgrades. Roche claims time-saving and greater accuracy compared with existing bioprocess monitors. “And this plays to QbD,” Schneider says, “where you need accurate data to design studies.”

Roche is collaborating with Bayer Technology Services, the engineering arm of Bayer Healthcare, on an autosampler that feeds Cedex Bio with process fluid, allowing real-time analytics and feedback control. The two companies are investigating how best to integrate the two systems to make measurements more accurate and fully automated, providing data around the clock.

PAT and QbD

The FDA’s Process Analytic Technology (PAT) initiative, promulgated in 2004, had the immediate effect of opening up the panoply of chemical analysis tools to bioprocessors. Vendors operating in traditional spectroscopy markets—near-infrared, Raman, and UV/visible spectroscopy—were breaking into biopharmaceutical markets, as medical device companies had done a decade earlier.

A scan of the literature shows how radically new sensing modalities depart from traditional quantitation of pH and glutamate. PAT also inspired development of novel sensor and monitoring technologies for “old” parameters, particularly for instrumentation and data handling.

From the perspective of operations, PAT’s major industry-wide success has been bringing process monitoring closer to the process. FDA defines analytics as being offline (far away), at-line (in the same room), and inline, which roughly correspond to cycle times of days/hours, hours/minutes, or more-or-less real time.

Bioprocessors now have a better appreciation for shorter analysis times and greater proximity to where the action is. Bioprocess monitors that measure routine parameters (pH, DO, etc.) are most often deployed in-line, but other technologies are moving toward the bioreactor at a snail’s pace. One often-cited reason is that production suites were designed for manufacturing, not analytics, and that space can be tight.

Regardless, eight years after PAT, analyzers have become more sophisticated and robust, while biomanufacturers have a much more refined sense of monitoring, quality, and risk minimization. This is due in no small part to FDA’s well-reasoned connection of PAT, process understanding, and quality by design (QbD). These three factors have come to form the predominant business case for advanced process monitoring and control.

Their connectedness look great on paper, and sound great in theory. Building a strong business case for higher quality, products that are safer and more effective, and fewer batch failures is easy. Nevertheless, the uptake of advanced sensing and monitoring schemes has been painfully slow.

PAT’s promise to deliver “process understanding” and enable QbD is based on the premise that analytics will provide an actionable (or predictive) understanding of critical process parameters affecting quality, which will lead to higher-quality product and/or less waste.

Slow Implementation

Yet, in a survey conducted by Bioprocess Technology Consultants, just 29% of respondents indicated they practiced QbD. Twenty percent of those surveyed believed that QbD is overly time-consuming, 15% felt it would delay regulatory approval, and just 26% felt it reduced time to market.

Furthermore, the 9th Annual Report and Survey of Biopharmaceutical Manufacturing, a 2012 study from BioPlan Associates, reports that PAT ranks at the bottom of currently implemented quality initiatives. The numbers: risk analysis (61%), QbD (43%), lean six sigma (43%), PAT (21%).

The study cites implementation times (71%), insufficient personnel (61%), costs (54%), and regulatory uncertainty (47%) as the principal hurdles to implementing PAT. Respondents citing time and costs were down since BioPlan’s previous survey, which Eric Langer, managing partner at BioPlan and a booster of PAT, attributes to the industry “recognizing the benefits that PAT provides.”

But Langer admits that despite its promise, “implementation of PAT remains slow and uneven, leading some to ask when this initiative will achieve its promise” but is confident that innovation in validated measurement techniques and control systems will eventually win the day. “The success of PAT and QbD will depend on better analytics, which will help biomanufacturers make a strong business case for using these tools to maximize yields, obtain purer product, and minimize quality defects.”

Given the time since FDA’s PAT guidance, and the fact that bioprocess monitoring has been ongoing since the birth of manufacturing biotech, one wonders if biomanufacturers have calculated return-on-investment analyses and simply not liked the numbers.

For example, the batch failure rate for bioprocessing is around 7%. The most common causes for facilities operating above the 1,000 liter scale are contamination (30%), operator error (20%), equipment failure (18%), and failure to meet specifications (13%).

For facilities operating below 1,000 liters, the top three factors are material failure (29%), equipment failure (26%), and product cross-contamination (17%). PAT might avert some failures related to meeting specifications or equipment failure if the specific analytics related to some quality attribute were in place, and if timely remedial action could turn the batch around—big “ifs”. PAT is unlikely to salvage batches that fail due to microbiological contamination or cross-contamination, operator error, or material failure.

Similarly, quality attributes such as protein folding, aggregation, and post-translational modification are difficult to evaluate in real time, particularly in-line or at-line. The possibility of repairing the batch in mid-production assumes processors “understand” the exact cause of the excursion from specifications and can act in time—neither of which is guaranteed.

So questions remain about PAT and QbD—their workability, cost-effectiveness, relationship in the analytic-quality space, and their ability to deliver as promised.

According to Christian Hakemeyer, Ph.D., fermentation manager at Roche Diagnostics, classical analytics still prevail at his company. These include offline biochemical characterization methods for proteins, and online pH and pO2 for cell culture processes. These analytics do not differ from those employed in the manufacture of therapeutics, he says.

That is unlikely to change any time soon, as Roche Diagnostics has no official PAT initiative, “only sporadic technology projects evaluating new analytical methods.”

“So far, the relationship between PAT and QbD remains within the realm of theory,” Dr. Hakenmeyer says. “With the exception of spectroscopic methods for characterizing raw materials, almost no successes are known in bioprocessing.”

Vaccine Process Analytics

Cell-based influenza vaccine manufacturing has been gaining on traditional egg-based production for reasons that are well-known. The switch is occurring at modest pace despite advocacy from the U.S. government, approval of such vaccines in Europe, and the completion of Novartis’ cell-based manufacturing plant in North Carolina. With cell culture comes the potential for applying advanced process monitoring and PAT to vaccine production. However, a recent survey by Bioprocess Technology Consultants suggests that PAT is far from a sure shot with vaccine makers.

Process monitoring did not make the list of the top five technologies that will have the greatest impact on vaccine manufacturing over the next five years. Flexible manufacturing, 36%, and novel delivery systems, 24%, were the top two. And while 73% of respondents believed that PAT and QbD could accelerate vaccine production, only 12% thought the benefits would accrue during upstream processing. The remainder were evenly split among formulation and purification.

Even when both are conducted in cells, vaccine and antibody production differ in one significant respect: Vaccine makers sell immunological activity, not grams of product. Where protein titer is a good indicator of the state of the cell culture, virus or VLP titers are less informative. Furthermore, immunological assays can take many hours, and vaccine production cycles are much shorter than for therapeutic proteins.

Still, the de-emphasis on upstream processing is puzzling given vaccine makers’ interest in infection rate, cell viability, and indicators of optimal harvest time, all of which are potentially “PAT-able.”

Cell-based vaccines provide somewhat more opportunity, as a group from the Netherlands Vaccine Institute is demonstrating with their whole-cell whooping cough vaccine.

If At First You Don’t Succeed

Grifols, manufacturer of plasma-based products, initiated a large-scale PAT initiative several years ago. Principal engineer Paul Spencer explains that bacterial contamination is the primary source of batch failure at his company. “The value of each batch can exceed $2 million, and a contamination problem puts several batches at risk.”

Bacterial endotoxins are the main reason for rejecting batches. Contamination can arise from issues related to equipment, cleaning, or the process itself. At certain levels of contamination steps can be taken to rescue the batch, but more often than not it is destroyed.

Grifols had been using an off-line test for polysaccharide endotoxins to monitor bacterial buildup in its processes, but the assays take as long as 12 hours. During that time a process may be over and another begun in its place. As many as four different products can run simultaneously, and contamination in one implies potential contamination in others.

At the time Joydeep Ganguly, an electrical engineer working at the time for Grifols’ predecessor company Talecris (and who is now with Biogen Idec), began investigating real-time, at-line endotoxin sensors. The probes were capable of determining endotoxin levels in 30 minutes, in time to prevent product contamination and batch failure.

“If we’d saved just one batch per year, the product would have been worth it,” Spencer confesses. “But the detection limit of the sensor was too high, and the false negatives were killing us. By the time a high reading kicked in, the batch had already been compromised.”

Previous articleLegalizing Medicinal Marijuana
Next articleTop 20 Pharma-Friendly Congress Members