I have already dedicated several columns to the technical and financial aspects of modernizing client-contract organization relationships. I am always struck by the “Law of Unintended Consequences” (LofUC) and how often it applies to our industry. It can be described as a marriage between a “good news-bad news joke” and Murphy’s Law. For example, in 1983, the EMA (European Medicines Agency) suggested that incoming materials, both active pharmaceutical ingredients (APIs) and excipients, should be qualified for every container, not merely as a combined sample or even √N + 1 (N = number of containers of a particular lot), but all the containers. Clearly, this wasn’t as bad for European companies supplied by rail with large containers. However, in the U.S., we received our materials in many more smaller truck containers. For example, lactose was received in more than 200 bags per batch. Performing USP or EP testing would be prohibitively long and expensive.

Considering that, in light of the small number of containers in Europe, the Agency didn’t consider that request very burdensome, at least not in Europe. To U.S. companies, it was a huge expense, both in personnel and time spent. Interesting fact: U.S. affiliates of, say, Swiss giants, didn’t make as much income, yet produced more product. The Swiss didn’t have the same anti-trust laws as we did. It was legal, for example for them to “divide up” the continent so as not to have head-to-head competition. Company “A” would sell a sleep aid in Italy and Spain, company “B” would sell a similar product in Germany and France, “C” would have the Benelux countries and Denmark, and so on. In the U.S., direct competition was the norm, so prices had to be competitive, leading to less profit for more product.

When faced with the massive workload and limited ability to expand personnel and lab space, we turned to what other industries were using, namely food and agriculture. Since USDA (U.S. Dept. of Agriculture) was as regulated as U.S. FDA, I assumed there were reasonable precedents for Near-Infrared Spectroscopy (NIRS) in products consumed by humans.
Besides the [O=C-N-H] group was essentially the same for Nylon, proteins, and numerous drug substances. The USDA measured water and starch, so did we. And the rest was easy to imagine.

But, just as Michaelson’s interferometer had to wait for computers to use the centuries-old mathematics of Monsieur Fourier, NIRS—discovered in 1800 and used in agriculture since the 1950s—had to wait for suitable software and hardware. Both occurred in the 1980s and were put to use by the two leading companies of the time: Pacific Scientific (later NIR Systems) and Technicon (later Bran+Leubbe). Qualitative software was developed (H. Mark at Technicon), making the scanning of hundreds of samples in a morning feasible.

So, commercial NIR was just becoming good enough to be used in the Pharma industry, just when it was most needed. As icing on the cake, secure user passwords and archived data were about to become commonplace, well before 21CFR11 was written. Today, NIR units—both hand-held and riding on carts—are becoming ubiquitous in Pharma warehouses. Speed? Yes, but wait, there’s more. Since NIRS is best used and can only be used for raw materials in a diffuse reflection mode, physical parameters affect the spectra.

Why is this important for today’s industry? Remember the LofUC? Quite simply, the information that may be gleaned by a few rapid spectral scans used to take days by traditional means: (ang.) particle size, polymorphic form, crystallinity, moisture, and, of course, identity. Since all these parameters are needed for a successful PAT/QbD program, the industry, both initiator and generic, no longer can cry that knowing these BEFORE production is too hard or takes too long. Of course, when one said “rapid” in the late 1980s, they meant rapid versus the hours and days needed to perform “wet” chemistry. A minute or two per sample was lightning fast by comparison. BUT, not fast enough for real-time production.

One of the barriers to faster, more sensitive instruments was cost. The detectors were either Silicon (Si) for short-range or near-near-IR (400-1100nm) that included the visible range or “normal” NIR of between 1100 and 2500nm, using lead sulfide (PbS) detectors. Interestingly, the NIR range extends to ~3500 or 3600nm, but both the lamps—tungsten-halogen, or wolfram-halogen for the rest of the world—and PbS detectors pretty much stopped measuring at 2500nm. The NIR region was therefore defined, not by theoretical calculation, but by the hardware available at the time.

A better material was available, namely Indium Gallium Arsenide (InGaAs). It had an extended spectra range and sensitivity and a lower noise component. Sounds great, right? The telecom industry thought so, too. They were, in essence, hoarding whatever supplies could be manufactured for their own uses: long distance and fiber optic communications. The speed and sensitivity made the material excellent for both home and office use as well as long-distance, even trans-continental, lines. The speed of re-emitting signals with extremely low noise made InGaAs perfect for communications and spectroscopy.

With the industry with far deeper pockets than instrument companies buying every gram they could, the cost to anyone was high. I asked the owner of a NIR instrument company how much a diode array, not even cooled, with InGaAs would cost. He answered, “About $10,000.” So, I asked what they would cost if you bought them in larger lots. His answer: “Still $10,00 apiece.” Then, in the early 1990s there was a telecom deflation, or “crash,” if you prefer, where many companies merged or went out of business. A number of direct consequences arose as a result:

1. A whole lot of InGaAs for detectors were suddenly available;
2. The detectors were suddenly far less expensive ($10,000 down to $100); and
3. A number of suppliers to the industry were without clients.

Points one and two explain why traditional instruments rapidly became better, but why is point number three important? Well, the choices for the companies who supplied telecom were simple: close down or find a new niche. Some chose the former, packed their tents, and left town. The latter choice was a happy sample of LofUC: these companies took their devices, based on MEMS—”microelectromechanical systems,” or the technology of microscopic devices, particularly those with moving parts, known as micromachines in Japan, or microsystems technology (MST) in Europe—and proceeded in converting them to hand-held spectrometers.

Yes, they were sometimes “two-hand” held systems and, because of high R&D costs, hardly inexpensive. But, they were battery-powered and portable and the first generation  showing that such devices were possible. These made warehouse inspections much simpler and led to hand-held Raman devices, as well. But, they were not designed for process control. That occurred with a happy coincidence that preceded the PAT (Process Analytical Technologies) Guidance (2002 draft; 2004 final): Pfizer (UK) worked with Zeiss Optics (Switzerland) to develop a rugged NIR instrument that could ride on a bin blender. It had no wires and was battery powered, utilizing fiber optics to “see” into the blender, measuring a scan with every rotation.

At the time, this 1-3 second scan was sufficient and the time between scans was long enough to allow real-time calculations. This minor break-through allowed for optimization of blending, the first step in enabling PAT as a practical program. Now, we could characterize raw materials and APIs AND follow blending. It was a quick transition to a stationary NIR unit for drying and granulation monitoring, optimizing both steps. The same equipment with fiber optics was then converted to measure coating, in real time in the coating pans. Real-time tablet measurements weren’t there yet, however.

Another “UC” came out of left field. A company called Spectral Dimensions in Maryland built a wonderfully useful chemical imaging (CI) microscope that showed the components in a sample (e.g., a tablet), where they were, and could quantify each material—excellent for reverse engineering a proprietary formulation, no? It was, sadly, made of expensive components and, in addition to taking several minutes for each sample, it began life at over $150,000. While not adaptable to the process stream, it served to aid in OOS inspections. Eventually, the price escalated to more than $500,000, so production stopped.

But, seeing the usefulness of CI, VTT, the Technical Research Center of Finland,  developed a dynamic CI device, where the sample or the device needed to move through the length of the sample. This “push-broom” technology was adapted to scanning blister packages, in real time. The latest iteration, produced by Indatech (France), is capable of scanning and evaluating up to 200,000 tablets per hour.

It’s interesting how seemingly unrelated businesses and technologies have come together to give us affordable PAT/QbD for the masses. In other words, monitoring and controlling solid dosage forms is not just for the “big boys” anymore. Hence the title.

---

Emil W. Ciurczak
DoraMaxx Consulting

Emil W. Ciurczak has worked in the pharmaceutical industry since 1970 for companies that include Ciba-Geigy, Sandoz, Berlex, Merck, and Purdue Pharma, where he specialized in performing method development on most types of analytical equipment. In 1983, he introduced NIR spectroscopy to pharmaceutical applications, and is generally credited as one of the first to use process analytical technologies (PAT) in drug manufacturing and development.