Don’t blame government regulations
While “PAT, APC, and beyond” (May/June 2012 InTech) makes some excellent points, the first part of the article paints a rather black and white picture of regulatory constraints limiting the use of inline data collection and real-time product quality analysis in life science industries.
The article indicates that before 2004 (when the FDA launched its PAT campaign) a regulatory stranglehold on process innovation had kept the industry in the technological dark ages. In actuality, the pharmaceutical industry has utilized innovative state-of-the-art instrumentation and automation technology for decades and has even been an innovation leader in several areas. Some pharmaceutical companies were interfacing analytical systems (e.g., mass spectrometers) to their production fermentors as early as the late 1970s. In the 1980s, there was significant use of on-line HPLCs interfaced to pharmaceutical production chromatography columns, which automated separation operations in the purification portion of processes. Also, in the 1980s, pharmaceutical companies were among the first customers of DCSs. By the early 1990s, several pharmaceutical plants were known to be using real-time expert systems, interfaced to their control systems and historians, for bioprocess monitoring, on-line diagnostics, and intelligent remote alarming, with some of these applications published in two 1999 InTech articles.
So, while the 2004 FDA initiatives summarized by the authors are now encouraging greater use of PAT technologies, 2004 did not mark the beginning of PAT applications. Rather, it marked the “formalization” of PAT, organized and supported with the help of the FDA, that builds upon the known successes of past PAT applications. Some of these were published in two ISA books:
- New Directions in Bioprocess Modeling and Control: Maximizing Process Analytical Technology Benefits, 2007
- Automation Applications in Bio-Pharmaceuticals, 2008
Regarding the authors’ claim that innovation has been stifled by regulatory constraints, there are no regulations barring a company from pursuing instrumentation, data analysis, or process control innovation when a new drug is in development (which is often a several-years window of time). What is a regulatory bottleneck is the ability to make certain improvements to the manufacturing process once an NDA (new drug application) has been submitted to the FDA and the process for making the drug has been transferred to production. Changes to an existing manufacturing process are possible, but formal change control must be followed, which can be bureaucratic and time consuming. On-line/at-line monitoring and control innovation occurs, but usually in development pilot plants, with such innovations often transferring to manufacturing when the process is transferred.
The other major impact of regulation is the requirement for more equipment qualification and process validation than is typical of non-regulated processes. While this usually adds to the time and cost of starting up a new process in manufacturing, it is believed by many to be time and dollars well spent, as it typically results in far fewer post-implementation plant problems.
The authors also note in their introductory comments that in-line data collection and real-time product quality analysis is much more common in less-regulated continuous process industries. While several pharmaceutical plants contain many thousands of I/O points and collect large amounts of on-line/at-line data, any reduced sophistication of automation in such plants is due more to the batch/discrete nature of most pharmaceutical processes than it is to government regulations. Continuous processes tend to be well-understood and, other than start-up and shut-down, are single operational phase processes, operating at steady state. Most life science processes are comprised of many batch steps/phases with most operations time varying and non-steady state. The biology of such processes is often not as well understood as desired, and there is frequently a significant amount of manual involvement in plant operations. Also, many life science operations occur in sterile environments, which are hostile to some traditional sensors. These and other nuances of batch processes have the net result of minimizing resources available to work on APC techniques, such as Kalman filters, model predictive control, optimal control, etc., which are techniques more common to continuous processes. So, automation areas in which pharmaceutical processes lag behind continuous processes have little to do with the FDA regulatory environment, and a lot to do with the nature of batch/discrete bioprocesses.
In summary, most of the article is excellent—but it seems to blame too much of whatever gap might exist between life science and continuous manufacturing processes on government regulations. It also does not recognize some of the outstanding examples of PAT-related innovation that preceded the formalization of PAT as a new FDA/industry initiative in 2004.
Joseph Alford, PhD, P.E., CAP, ISA and AIChE Fellow