Several recently published articles have promoted the process analytical technology (PAT) initiative as the path to the real-time release (RTR) of finished product. Many strong arguments favor gaining a better understanding and control of our manufacturing processes, but it is not clear that RTR is one of them. The US Food and Drug Administration and other regulatory agencies worldwide are working diligently to base regulation in science, but it is reasonable to ask whether RTR through PAT is compatible with the ideal of science-based regulation. Because my education, background, and training are in the field of microbiology, I will examine the question from this perspective.
What about microbiology in PAT?
One noticeable correlation in the current literature on the promise of PAT to deliver RTR is that the authors are virtually all chemists or regulatory specialists. In the rare publication discussing microbiology and RTR through PAT, this goal is mentioned in only the most off-hand manner (1). This avoidance of a clear discussion of microbiology within the PAT initiative as a contributor to RTR is striking. A review of aseptic manufacturing indicates that the microbiological portions of a manufacturing process delay product release to the greatest extent (2). In fact, the time required for microbiological in-process and release tests virtually defines the critical time line of manufacture and release of product for sale.The second puzzling aspect of this omission in the current PAT discussion is that the first approvals under FDA's PAT initiative were for microbiological methods (3). The role of microbiological testing in RTR is too important to ignore.
PAT and the microbiology of nonsterile products
Nonsterile products are subject to process controls from a microbiological perspective. Though these controls are not as stringent as those for aseptic manufacture, the US Code of Federal Regulations (21 CFR) clearly requires the manufacturer to produce products free of "objectionable organisms." It must be stressed that FDA's concern about "objectionable organisms" is not the same as the compendial tests for "specified" organisms (4–6), and so the finished product should also be evaluated for the presence of "objectionables" while in quarantine (costing money for warehousing and not being sold).
The implementation of PAT to allow the RTR of nonsterile product would require the manufacturer to establish controls on the process sufficient to ensure that no potentially objectionable organism would gain entry (via raw materials, the environment, personnel, or any other means) into the finished product. In addition, it would require that we have sensitive and accurate measures in real time to confirm this state of control and allow changes to be made in real time to correct any potential issues as they develop. Although it is intellectually possible to envision a system of HACCP-like evaluations in which all contaminants are instantaneously counted, identified, analyzed, and reported, current technology does not come close to delivering such a system (7, 8).
PAT and the microbiology of sterile products
Sterile products can be divided into two main types: terminally sterilized products and aseptically filled products. For clarity, we shall evaluate them separately. RTR through PAT is feasible for sterile products manufactured using terminal sterilization. This process is called "parametric release," which is cited in the PAT guidance and has been well described by FDA (9), the European Medicines Evaluation Agency (10), the Pharmaceutical Inspection Cooperation Scheme (11), the US Pharmacopeia (12), and in the literature (13, 14).
So, what about the aseptic manufacture of sterile product? RTR through PAT requires extensive process understanding and control. Current in-process monitoring capabilities in microbiology do not allow for the determination of finished-product attributes because of excessive variability (15, 16). The plate-count method has a significant gap between the limit of detection (LOD) and the limit of quantification (LOQ). Though the LOD for plate counts is 1, the LOQ has been established to be in the range of 25 (17) to 30 (18).
Here is a major problem of current "regulatory science" being at odds with "science-based regulation." Regulatory expectations are that plate counts can accurately distinguish between a count of 2 colony-forming units (CFU, acceptable quality) and 4 CFU (unacceptable) or between 4 CFU (acceptable) and 6 CFU (unacceptable) for surface monitoring under various conditions. In truth, the plate-count method cannot reliably quantify any number less than 25 CFU (at best).