Those of us working in the commissioning and qualification field have the opportunity to see the work and efforts of many others before us, the scientists who discover and create the target molecules, engineers who developed the process and specify the equipment, architects who design the buildings, designers who work out the details, manufacturers who build the equipment, and skilled craftsmen who put it all together. A few years and 100’s of millions of dollars later there is an operating cGMP pharmaceutical plant. There are numerous design reviews, FATs, SATs, commissioning, IQs, OQs and so on along the way. A year or two further down the road the cycle of re-qualification begins. And occasionally we see things that could be done better next time around.
Aside from reviewing the performance of a system or a piece of equipment, a good re-qualification process would also include how that system interacts with other systems and operators, its maintenance history, its performance and ultimately how it affects the quality of the product. What are the re-qualification engineers looking for when they walk-down a system and compare it to the P&ID? They are ensuring that all the components are there in their proper relative position and orientation¹ with the proper attributes like slope where specified. The P&ID has been the one governing document since the beginning of engineering and could easily be at or beyond its fifth revision by re-qualification time.
The concept of “Lessons Learned” comes from project management, “Continuous Improvement” comes from the quality field, and “Corrective And Preventative Action” comes from operations. Together these are the backbone of the GMP and GEP environment. When an operator/maintenance error has been detected, it is noted and the information passed to the proper group to be corrected. This might trigger an additional review to discover the root cause and a develop a mitigation path to reduce the probability of the error recurring, or maybe eliminated entirely with a design change. Possible solutions might involve operator retraining, additional signage and/or engineering methods. For the Owner, this might be sufficient to meet their regulatory requirements and this system might successfully continue operating under GMP compliance until the end of the equipment’s life-cycle. But should the lessons learned stop here or go back further to benefit the next generation of systems to be built?
A typical diafiltration system consists of a batch tank, membrane modules, pumps and their associated instruments, pipe, valves and a control system. A lab or pilot plant system might have all these components assembled on a single cart or maybe the components are bought à la carte and connected with flex hose ad hoc. While this latter approach may be sufficient for development work, for a production-level manufacturing system, the equipment is larger with vastly increased volume and an engineered system approach is used. Ideally you would want to keep this system as a single unit but at some point, this may not be possible and the design of two or more skids may be more practical. Or there may be other reasons to break up a system into multiple skids e.g., existing building layout, doorway size limitations, or transportation issues. Hopefully, the misconception that competitively bidding a group of individual subsystems with delivery the lowest cost is not adapted. While the multiple supplier approach may yield the lower initial cost, other issues later on will become more prevalent driving up the life-cycle cost. The initial cost of the equipment needs to be compare to the value of the product and the cost of manufacturing over the life span of the product. A little loss of product due to poor efficiency or the occasional loss of a batch could be significant. Sometimes at the early project phase when these decisions are made it is difficult to evaluate life cycle costs. The rational of considering an overall system evaluation method early in the design phase will be discussed in the following example.
Here is the case from a recent re-qualification which brought this topic to mind. The system consisted of a filter module skid near a pumping skid connected by flex hoses. Since it is typical engineering practice to bring connections to the edge of a skid, the interconnecting hoses were the same length and the Feed and Retentate lines having similar flow rates were the same diameter and in this case they were also next to each other. So, the hose sections looked identical and without any special obvious marking could easily be switched, which they were. In addition, the hoses had integral manual isolation valves which more than doubled the spool piece weight; this also presents an ergonomic issue. Looking at the existing equipment it is hard to understand why the system was designed this way. Without reviewing the design record, we can only speculate that the isolation valves were incorporated late in the design phase. Maybe the skids were already released for fabrication or on site and the installation of the options for locating the valves without causing major rework or delays was limited to the interconnecting sections. If one had the opportunity to redesign this system without any restraints the valves should be within the skid boundary so they could be fully supported. Ideally this system would be built as a single skid, but if the design needs to go to two skids at least they should be located adjacent to each other. If it were necessary for the skids to be apart, one of the connections could be recessed; hence the flex hoses would be different lengths which would prevent inadvertently reconnecting them incorrectly.
Flex hoses do have their applications as they can provide vibration isolation and can make-up for small dimensional differences. Over time flex hoses can soften, elongate and sag potentially creating pocketing and drainage issues. Having two skids abutted to each other has several advantages, the use of flexible hoses would be eliminated, all sloped lines would be fully supported, and a system of minimal volume could be designed.
- Minimize the use of flexible hose
- Support all valves in fixed locations, minimize use in removable sections/spool pieces
- Avoid designing identical removable spool pieces unless they are straight thru pipe or hose
Another underlying issue behind this discussion is the “Minimal Volume Concept”. The process objective of diafiltration is to concentrate the target molecule while eliminating unwanted material. In this process the batch is recirculated through the filter modules while smaller molecules are rejected. Buffer solutions are added to aid this process.
Minimal Volume Concept followers assert that by reducing the final batch size to the smallest practical volume the overall target molecule yield will be increased as less of the batch is lost when transferring to the next process step, usually chromatography. To build a minimal volume system, the tank, pump, and membranes need to be as close to each other as possible with the shortest piping design. This is hard to do if different companies are building separate subsystems. Interconnecting skids even if they are butted up against each other still adds extra pipe length hence extra volume.
The other camp of subject matter experts essentially believe that the complete batch gets transferred forward regardless of the final batch size and the length of the system and transfer pipe, therefore the yield is the same and that the expense of a minimal volume system is not justified. The initial cost of a minimal volume system could easily be twice as much than the separate system approach on paper.
Engineering firms recognize the special process expertise associated with bioreactor and will single source a qualified vendor. The same recognition of the process expertise that membrane manufacturers may be lacking.
Returning to this case, both the Feed and the Retentate interconnecting assemblies were identical and without checking the valve tag numbers one would never guess that they had been switched. The Feed and the Retentate lines flow in opposite directions, therefore when the hose assemblies were inadvertently switched both streams were flowing in the reverse direction in relationship to their valves. In this case the valves were of bi-directional design so the errant reassembly did not pose a serious issue. In other GMP applications unidirectional valves might be specified which could cause improper draining, substandard cleaning and a possible product contamination with subsequent batches.
¹ P&IDs do not show absolute location.
Written By : David Campanella, Engineer IIIJoin our network, follow us and share this on LinkedIn, Google +, Twitter and Facebook so you can see other new and exciting news and discussions being posted by ICQ Consultants.