Solvent Recovery at Pfizer: A Continuous Solution for Small Waste Streams
When equipment and processes challenge solvent recovery for small-volume streams, an atypical approach is needed, says Pfizer’s Frank Urbanski.
By Paul Thomas, Senior Editor
Solvents are often the primary waste company in API and drug processes, and solvent recovery is routine when it is environmentally and economically feasible. Pfizer has been recovering solvents for decades, and is continuously seeking ways to enhance that feasibility. From 2006 to 2009 at its Kalamazoo site, solvent recovery efforts have led to $65 million in savings, reports Frank Urbanski, Director of Engineering Technical Services, Pfizer Global Engineering.
Recently, Pfizer Global Engineering has worked with students and faculty at New Jersey’s Rowan University to research and develop green approaches to drug manufacturing, including solvent recovery. Led by Urbanski, scientists and engineers from Kalamazoo and Peapack, New Jersey, have explored solvent recovery as an alternative to incineration for small-volume waste streams.
Capital investment for such recovery typically includes piping, tank farms and recovery equipment, Urbanski notes—easy to justify for large volumes and expensive solvents, but not so easy with smaller streams that can’t necessarily be pooled with other solvents.
| The Solvent Recovery and Distribution (SRD) area of Pfizer's Kalamazoo API site. Courtesy Pfizer Global Supply
The researchers performed a case study on several API production waste streams in Kalamazoo, starting with the recovery of acetonitrile from the synthesis of selamectin (the active in the veterinary drug Revolution). Rowan designed a small-scale distillation, solvent-recovery system, and also evaluated it for recovery of isopropanol (used to manufacture nelfinavir, the active in Viracept) and toluene (used to make hydrocortisone).
We spoke with Urbanski to get more feedback on this project and other solvent recovery efforts at Pfizer:
PhM: What’s particularly challenging about solvent recovery from small-volume waste streams? It’s not just a matter of scaling down large-volume practices?
F.U.: The bottom line is it depends on whether your existing plant was designed to handle smaller quantities of solvents in an efficient manner. Certainly, if new equipment were to be purchased, the design can be scaled to the desired flow rates. If constrained to existing equipment, there would normally be some degree of turndown capability, allowing operation at a lower steady-state feed rate. Alternatively, a column could be operated for a shorter duration at steady state. Many times some cleaning will be required between different recovery processes, and more frequent changeovers for small recovery runs adds some degree of inefficiency. So, one might be able to operate an existing column if capacity is available, though less efficiently.
While Pfizer API plants are typically quite flexible, they are not designed to handle every scenario. Excess variability adds to plant cost. So, designs are based on assumptions regarding number of products, batch sizes, number of batches, etc. Similar assumptions would have been made to justify the incremental investment in solvent recovery assets, such as tank farm capacity (number and sizes), transfer lines, and recovery equipment (distillation columns, Pervap systems, etc.). Recovery of 100% of waste solvent streams would not be a typical design assumption and thus a facility would have constraints in the variety and volumes of solvents (virgin, recovered, in-process, etc.) that can be transferred, collected, stored and recovered at any given time. Pooling of solvents can therefore help by reducing the number of isolated streams that must be managed at an operation if it is possible for the particular solvent stream, and if the plant design, lab resources, process validation and quality systems are appropriate.
Solvent recovery operation is secondary to production in that it must react in response to changes in the production schedule, and continually seek to optimize asset utilization while trying to maximize benefits to plant finances. Small spent solvent streams tend to have lower value, and so receive lower priority in the scheduling of equipment, if there is capacity at all. When actual product mix strays from the design basis, and trends toward more variability in products, shorter campaigns, smaller quantities, or increased numbers of solvents, then fitting the solvent recovery demands to existing equipment becomes more challenging.
PhM: Do you think that small-volume waste streams are often overlooked by manufacturers as opportunities for solvent recovery, perhaps viewed as not cost effective or too much trouble? Has this project changed your/Pfizer’s thinking in terms of what is possible, and economically viable?
F.U.: In today’s environment, it is doubtful anyone simply overlooks a hazardous waste that is costing them significantly. Pfizer colleagues are engaged in serious continuous improvement, and reduction in waste volume is a target area. But, there are constraints as described above, and not all streams are being recovered—some because it is technically not possible, but others because capacity is not available or capacity does not fit the situation.
The idea for this study came up as I was thinking about how solvent recovery might operate if we had a continuous low-throughput API manufacturing process. I envisioned solvent recovery would be small scale and fully integrated with the manufacturing process, continuously being recycled, with small inventory in the system. There would not be a large collection tank for the waste stream somewhere out in a tank farm. Rather, the recovery column would be located very close to the manufacturing process, and the waste solvent stream would feed continuously, operated by the same technicians as part of the manufacturing team. The recovered solvent would be continuously monitored by PAT devices and fed back to the process, with little or no accumulation of recovered solvent in a large tank. I thought the same concept could be applied to small-volume batch mode API operations for streams that were not being recovered currently.