NIRS is having a renaissance;but not all the pieces are moving at the same speed

NIR has been a mainstay of the food and agriculture scene for over four decades. For over three decades, textiles, polymers, and chemicals have been in love with the technique. Whilst some initial work was done in the Pharma industry (e.g. yours truly, in 1984 for raw materials qualification), the acceptance was slow in coming. There are numerous reasons for this, but they breakdown into three categories: economics, talents/education, and statutes (law).

Addressing the third category first, it is easy to see why “the rules” make progress slow in Pharma:

Only recently have the US Food and Drug Administration (FDA) and European Medicines Agency (EMA) added large numbers of qualified inspectors. In the first several decades of the FDA’s existence, the only requirement for inspectors (and reviewers of documents) was to have a “degree in science.” It seems in retrospect, that a degree in meteorology or geology might not fully prepare an inspector to drive modern practices.

(One good example is liquid chromatography. In 1970 (when I started), if a Pharma company submitted a new drug application (NDA) containing HPLC as the release method, it was out-of-hand rejected. Why? The FDA had neither a scientist familiar with LC nor the instruments on which to run it. It wasn’t until the 1980s, when colleges began turning out chemists with LC experience that the industry was allowed to utilize HPLC. THEN, by the 1990s every NDA release method had to be HPLC for the FDA to review it.)

The largest product in any Pharma company is paper. I once joked that there are standard operating procedures (SOPs) for writing SOPs, for which there are SOPs. The massive number of documents needed for each single batch is amazing. Also, any change in methodology, equipment, ingredient, and especially, analysis method requires large amounts of validation effort. (This goes to the extreme of changing the brand of an HPLC column, even when packed with the same material as the previous column. Work must be performed to prove that “same is the same.”)

Despite encouragement from the Regulatory Agencies, many companies in the industry have reservations about “updating and streamlining” their entire process of manufacturing. The agencies issue “guidances” that suggest ways to integrate “new” technologies such as NIRS (and Raman, TeraHertz, etc.), but seldom issue clear regulations, stating that guidances are, basically, the current thinking of the government. Hence, the highly (some say over-) regulated industry is in fear of ANY change, so they stick with 1960s technology and processes.

Let’s look at the financial side now. Unless you are running a charity or non-profit organization or are a government agency, you attempt to make a profit for your shareholders. In this, the Pharma industry is unique: it makes, on the whole, a boatload of money (especially in the US, where the industry has the ears of Congress):

When compared with most other industries (where a 1–5% profit is the norm), the Pharma industry can make very large profit margins (25%+ is not an anomaly). (The common mantra is “we spend so many billions on research, we need the income stream.” The reality is that the amount of money spent on advertising often exceeds the amount of money needed to develop a particular drug.) Therefore, there is seldom any incentive to modernize when all they need to do is raise prices to make up for rejected batches and cost overruns.

When purchasing replacement equipment, the cost of implementing new technologies has to be taken into account: facilities redesign, power and HVAC requirements, training the operators, validating the new equipment (assuring the product is still within acceptable limits), and writing (and validating) new SOPs. Simply replacing old equipment with similar equipment saves time and effort and, since (in the US, at least) the company can raise prices to make up for lack of speed and accuracy, the easy way is usually chosen.

Then we have the expertise element. To modernize to a fully monitored (mostly spectroscopically) process, you need (aside from the proper equipment) trained operators, an instrument group to install, calibrate, and maintain the hardware, scientists to generate calibrated methods and maintain the software, and, as with all changes, monstrous mounds of documents to please the quality assurance (QA) department (the group that explains and apologizes to the FDA/EMA). So they need to follow the following steps:

After designating/training/hiring a group leader to move the process measurements to the 21st century, begin to build the group.

Before choosing the hardware, an analyst with experience in the technology (obviously, considering the name at the top, I mean NIR) to evaluate the best equipment for the purpose intended, not “the best equipment” available. (A BMW is a nice car and a Ford F-160 is a nice truck; which would you choose to haul bricks and sand which to take your date to dinner?) That means, choose which process is to be monitored, then obtain the best instrument for that process.

Training on NIR equipment is sometimes problematic. When I started in NIR, the companies who were the leaders (Technicon, later Bran+Leubbe, and Pacific Scientific, later NIRSystems) included a week-long course at the suppliers’ facility. These included lectures on theory, math, chemometrics, sampling, sample prep, and sample presentation. Now, it is more likely that a tech rep delivers the (usually lab) instrument, performs qualification (does it work?), and spends (perhaps) a day with the customer, showing him/her which buttons to push to run the equipment.

Public courses on NIR theory and practices are seldom seen, anymore. This coincides with an overall reluctance of companies to send any technical staff to any technical meeting or technical short course. Between newly-minted scientists’ over-reliance on internet posts and financial problems, following the 2008 melt-down, Pharma companies have moved away from third-party training and attending short courses. As a consequence, the in-house expertise needs to come from poaching NIR-savvy people from either other companies or instrument companies. (The latter further depletes technical staff at instrument companies, making training courses even less likely to be offered. Much like eating seed intended for future crops, no?)

Not all of the blame rests with the inertia of the Pharma industry. Since, as was once pointed out, the gross income of the entire NIR instrument industry is not as large as a single day’s output of Lipitor (when it was on patent, of course). As a consequence, when Pharma considered process monitoring, the R&D budget for smallish instrument companies was quite limited. That meant that early attempts at “process” instruments were simply lab instrument, placed into NEMA enclosures to avoid explosions. That often added $30–40,000 to an instrument that could already cost $60–70,000… and was “OK,” at best.

The problems were cost, size, and speed. For example

A typical lab-based, but hardened NIR instrument took approximately one minute per scan, not making it suitable for real-time process control. For biological processes, typically taking hours or days, measuring the mix every 10 min may be considered “real-time.”

Pfizer set up several lab-based units in an enclosed lab, built between two process lines (for Viagra), came closest to a process measuring system. Samples were taken and run throughout the production. This allowed a much better analysis of a batch, but still offered no true control (there was an hour lag between seeing an out of trend sample and the ability to correct the process).

The units placed into the process stream were large and bulky. This limited either mobility or where they could be installed.

The earliest breakthrough came when Pfizer changed the paradigm for design and installation of NIR equipment. Pfizer teamed up with Zeiss instruments to develop and employ for wireless blend uniformity testing. This set a new template for process instrumentation:

Since instrument manufacturers are not flush with money, development of new models used to be aimed at the “least common denominator,” able to be used in any place, much like Swiss Army Knife. For process control, more specialized equipment is needed, often in small numbers.

In addition to instrument qualification (IQ), operational qualification (OQ), and process qualification (PQ), we have a say in design qualification (DQ). DQ simply means working with the instrument company to design units specific to the needs of that particular company or industry. But, the cost of development for a small number of sales hinders development. The new approach involves incremental development steps:

The Pharma rep contacts the instrument company and outlines their needs.

An overall cost for development is generated.

The Pharma company proffers, say, 20% of that cost to develop plans and possibly a breadboard unit.

Another 20–25% could be paid when a working model.

Another piece could be paid when a unit is delivered and works as intended.

When qualified/validated, the final increments would be paid out to the instrument company.

Since the earliest process equipment successes, numerous instrument companies are now specializing in fast and accurate process instruments. Small units, based on MEMS and other rapid analyzers, are showing up in large numbers. Several instruments, based on “push-broom” imaging, are capable of qualitative and quantitative measurements at a speed of 100,000 tablets an hour. Concomitantly, software manufacturers are taking advantage of computers with “blinding” speeds to build algorithms to take the spectra from the new units and use it to actually control the production of products… in real time!

So, much as DaVinci designed a helicopter, but needed an internal combustion engine to fly it and Michaelson developed an interferometer in the 19th century, but needed computers to run M. Fourier’s equations to generate infrared spectra, NIR could not be applied to pharmaceutical process control until faster computers and instruments were developed.

In short, the pieces were there; the science was there; all that was needed was (money) and modern equipment to make our dreams come true. It did take Messiers Beer, Bougier, and Lambert well over 100 years to develop the simple formula A = abc. Science moves in pieces and, sometimes, it takes many years for “a plan to come together.”

Conclusions

The strengths of NIRS were recognized and employed by agricultural and consumer products (textiles, prepared foods, baked goods), largely because they were both less regulated and worked under tighter financial pressures. As regulations became modernized and hardware and software for rapid (validated) analyses came on the scene, the Pharma industry has begun to join the rest of the commercial world.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.