Friday, May 29, 2009



Moving from the lab to manufacturing can be a scary process

B etween the research lab and the manufacturing plant there is a twilight zone where the possibilities of discovery research meet the reality of pharmaceutical production. Sometimes this meeting is peaceful. At other times it is messy, contentious, or even dangerous. Translating the process from the lab scale to the pilot scale often takes time and skill. And yet there is tremendous pressure to get through it quickly so that a compound can enter the next phase of research.

The result is often a hybrid between a laboratory process and a real manufacturing process-an unlovely marriage of speed and scale. This is the stage where mistakes are most likely to be made. Engineers change pipes, valves, flanges, and other fixtures on the reactors and don't always remember to undo the changes later. Errors can result in consequences that range from a loss of money, time, and material, to more dramatic events like explosions. An understanding of how drug compounds in gram quantities "grow up" to the kilogram scale can make the transition more successful.


One of the first steps involved in scaling up a process is obtaining the starting material. Research chemists order their starting materials from catalogs, but it may not be possible or economical to get them from these companies in bulk. Instead, arrangements must be made with chemical manufacturers, who are often specialists in certain types of chemicals. Some starting materials will be made from scratch, adding an extra stage to the process.

Certain laboratory procedures can be problematic at the pilot stage. Chromatography and microwave chemistry are prime examples. Column chromatography doesn't scale up well. Vast quantities of solvent would be needed, even if columns of sufficient size could be created. And microwave chemistry tends not to work well at large scale because the microwaves don't reach far within a large vessel. Both of these techniques need to be worked around.

Ordinary laboratory glassware is another stumbling block. The flasks and tubes used on the benchtop are classic tools of organic chemistry and have been used for centuries. The problem is that no one has invented a 2000-liter Erlenmeyer flask, and if someone did, it would be awfully difficult to pick up and give a swirl. Instead, the process must be adapted to fixed equipment, which is less convenient, less efficient, and quite frankly, awkward in many ways. There are many things that can be done in the research lab that are difficult or impossible in a pilot plant.

Robert Huhn, director of product development at AstraZeneca (Wilmington, Del.), remembers a difficult process centering on nonscalable labware. "We were in the final few stages of the process for a product currently on the market, one of the biggest products we have. When we were in the late stages of scaling that up and transferring to the manufacturing site, we hit one problem. The problem was that after a reaction crystallization we had difficulties in the subsequent processing step." The reaction required the isolation of an intermediate. Low levels of impurities at this stage are disastrous for the next step. In the laboratory process, the intermediate was effectively filtered and washed. But at the pilot scale, things didn't go so well. Pockets and channels in the filter cake harbored impurities.

Employee's at Laureate Pharmaceutical pack a column in the company's pilot plant.


The solution was to substitute the nutsch filter with a centrifugation step in the process until the filtration problem could be studied and solved. It is helpful if these problems can be anticipated during the development stage, but Chris Exon, PhD, vice president of the chemical synthesis group at Roche, would rather not see lab chemists inhibited by fears of using columns, microwaves, or labware. "Our philosophy is they shouldn't worry about it. They shouldn't be restricted in the different compounds they can make by scale-up issues. We love a challenge."


Frank McConville is a senior consultant for Impact Technology Consultants (Lincoln, Mass.) and a trainer for Scientific Update (East Sussex, U.K.). He and his colleagues at Impact advise clients on topics that range from commodities chemicals to biotech, with particular expertise in the area of process scale-up. Smaller companies often don't have the expertise or experience to manage scale-up operations, and that's where engaging a consultant can save the day.

Some of the most difficult issues in pilot stage scale-up, McConville says, are mixing, heat, and reaction times, which are also some of the most basic and straightforward elements of a chemical reaction. In a large reactor, even with a 100 horsepower motor, mixing is not as efficient as it is in the laboratory. Unfortunately, poor mixing can change the chemical outcome of a reaction. Reactions also take longer in larger batches. Just as in cooking, when you double or quadruple a recipe, it takes longer to cook. And, as most of us have learned from experience, a recipe that works fine when making eight servings can fail catastrophically if you try to double it.

Finally, heat can be very difficult to control in a large reaction. "When you have a reaction the size of a house, the only way you've got to cool it is a jacket. You slowly, very slowly, remove the heat. It can take hours and hours to cool down a very large reactor." Exothermic reactions can continue self-heating until they decompose and explode if heat issues are not anticipated and controlled.

Our customers always have aggressive timelines. Process development folks like to have more time to develop a really good process.

-Mike Ultee, PhD, senior director of process sciences, Laureate Pharmaceutical

An illustration of this unpredictability is a project that McConville worked on for a manufacturer in England. "We were operating a new process for the first time at the 2,000-gallon scale. We had worked this process at my company in the lab and in a small pilot plant. One of the key steps in the process was distillation. After the chemical reaction, you had to distill it down to half the volume to remove the water and get it ready for crystallization. We always did that distillation under vacuum."

After McConville and his team had done a great deal of work translating the process to the larger scale plant, they arrived in England to find that the large reactors were not set up to distill under vacuum. After consulting with their team back in the states, they embarked on a 2,000-gallon experiment to adapt the process to non-vacuum distillation. "Now we had to cook this product at a much higher temperature than it had ever been cooked before, much longer than it had ever been cooked before. After the distillation was done, the material looked a little off-color, but not bad," McConville says. "We took a sample-that was the procedure-to make sure that there was no water left."

That's where things began to go wrong. Water appeared in the sample. Another chemist on the team insisted that it was impossible, but the results of the test were clear, so the team distilled a second time. After the second distillation, water appeared in the sample again. "My colleague was adamant-it was impossible. It turned out they had forgotten to dry out the sample pipe. Water had been left in there. Our batch was perfectly dry. Now we've not only cooked it higher than ever, longer than ever, but cooked it twice." The product passed all quality tests, except for the color test. Because it had a brownish color, the product failed specs; several hundred kilos of material, costing several hundred thousand dollars to make, had to be discarded.

I think it's important to understand that scale-up is not just about increasing the scale and hoping it's going to go well. It's about taking those conditions you will have in that scale [and] using chemical engineering know-how and theory to scale it down and mimic it in the lab scale.


A kilo lab supervisor at Roche works on validating systems and equipment.


The ultimate lesson had to do with communication. "A simple thing like cleaning out a line, leaving water in the line-a laboratory chemist has no idea what goes on in the cleaning protocols," McConville says. "Perhaps they have never even been in a plant before. He has to trust the operators, trust the procedures in the plant. Sometimes things happen beyond the control of the chemist who invented the process."


Compared to a chemical process scale-up, making biologicals seems almost boring. No matter how large the reactor, an explosion is unlikely, and the media are safe enough to drink. There is no real need to reengineer the pilot plant for each new therapeutic entity, and in some ways, the scale-up process is the opposite of a chemical scale-up. In a chemical synthesis, each reaction is unique, and the pilot scale process has to be designed from the ground up for each new compound. For a biological, the goal is to engineer the molecule so that it is optimized for a standard set of manufacturing conditions.

For this reason, the challenges of biological scale-up tend to be in human relations rather than in process. Mike Ultee, PhD, is senior director of process sciences for Laureate Pharmaceutical (Princeton, N.J.). Laureate is a contract manufacturer that offers development services, especially clinical-scale manufacture. "The products we're working on are made in mammalian cells...most are 100,000 or 150,000 molecular weight," Ultee says. "We do it with a goal towards getting clinical supply and commercial supply."

A simple thing like cleaning out a line, leaving water in the line-a laboratory chemist has no idea what goes on in the cleaning protocols. Perhaps they have never even been in a plant before. He has to trust the operators, trust the procedures in the plant.

-Frank McConville, senior consultant, Impact Technology Consultants

In order to be ready for pilot-scale production, the protein must be expressed in serum-free, suspension-adapted mammalian cells. Some of the proteins arrive in this condition, but a great many need re-cloning and development work before they can enter the bioreactors. This is an obstacle for some customers. "Our customers always have aggressive timelines. Process development folks like to have more time to develop a really good process," Ultee says. The reason for the customers' eagerness is the need for large quantities of material for clinical studies.

Some of the technical challenges presented by biological scale-ups have to do with the intricacies of the molecular biology involved. Some cell lines and cloned genes are poor producers. Recombinant proteins can be difficult to express, often because the researchers have not taken the time in the lab to optimize expression. Fusion proteins are often very difficult to express, because the protein has never been optimized to express in nature, much less in the laboratory. Thus, some of the work of biological scale-up may involve going back to the molecular biology to build a better producer.


Figure 1. A Software modeling tool

Computational Fluid Dynamics (CFD) is a complex software tool that simulates heat or mass transfer, reaction kinetics, particle formation, etc. This image is of a spray dryer nozzle being used to optimize fluid flow to solve a particle formation problem.


One of the most important tools for a successful pilot scale-up is good communication. Lab chemists need not change their focus or skill set, but if they know in advance that some techniques scale-up better than others, they can make life somewhat easier for the process chemists. Likewise, time invested in understanding the equipment in the pilot plant could pay off in disasters averted, such as water in the sample pipe. Another useful strategy, once the compound is in the pilot plant, is to identify potential problems in order to avoid them. If a particular synthesis step is a frequent trouble spot, then it's worth heading the potential problem off at the pass.

New technologies can help with some of the common challenges of pilot scale production. Through process analytical technology, operators can monitor reactions much more closely. For example, by putting an instrument in the reactor that allows monitoring of the reaction's progress, operators can get an early warning if something is going wrong. The Lasentec particle characterization instrument by Mettler Toledo (Columbia, Md.) is an example of a technology that can help with the monitoring of reactions, especially crystallization. Continuous flow and low temperature reactors are also technologies that could significantly streamline pilot plant production. Software modeling, like the Computational Fluid Dynamics system used by Impact Technology consultants, can answer many "what if" questions, heading off problems before they start (see Figure 1, below).

Finally, taking a look at the process in reverse can be helpful. AstraZeneca's Huhn rarely has trouble with the lab-to-pilot transition, because the company has a policy of scaling down before they scale up. That means taking a process planned for the pilot plant and carrying it out at the lab scale to work out any problems in advance. "I think it's important to understand that scale-up is not just about increasing the scale and hoping it's going to go well," Huhn says. "It's about taking those conditions you will have in that scale [and] using chemical engineering know-how and theory to scale it down and mimic it in the lab scale.... If there is a problem, we'll try to solve it at the lab stage. It's much too expensive to go up and spoil a batch."¦

No comments: