Over the last 25 years of consulting, I have lost count of how many potential clients have called and said, “I’ve just purchased a near infrared (or Raman, etc.) instrument; now, tell me what I can do with it.” [Since I always want to stay in business, I never tell them what I think they can do with the instrument.] As an analogy, both a Ford 150 truck and a Mustang are good vehicles. Before buying one, you need to ask why are you buying a vehicle? If I’m hauling bricks, the truck would be excellent; if I’m taking my wife to dinner, I might prefer the Mustang. Before you drop tens of thousands of dollars on a process monitor, first ask why you want it? There is a process to help decide.

Do you want to do off-line analyses (lab-type)? At-line? On/in-line? Do you want to perform qualitative analyses (ID, etc.) of incoming raw materials or in-process or finished product? Do you want to follow blending, drying, or coating processes? All these applications could require a different piece of apparatus. You need to define what and where you are measuring before you think of finding (or having a vendor build) an instrument. You might want a NIR or a Raman, or perhaps and on-line chromatograph or NMR, etc., etc., etc. to do your analysis. First choose a measurement and level of accuracy. (That is, do you want to see trending or single point analysis?) If you don’t know what information you need, you are never going to be successful.

First step: know the difference between data and information. Any number generated by a measurement is data.  However, for example, knowing that the wall of the lab is painted light blue is data; it has nothing to do with your analysis of the sample, however, and is therefore not “information.” Another criterion is the time from measurement to the evaluation of the information.

Clearly, if you are still running your production facility under a traditional GMP paradigm, speed is not necessarily critical. That is, you make your product, encase it in drums, drive it to the warehouse, and send the requisite number of samples to the QC lab and await the results. These results could trigger 1) release of the product, 2) reanalysis of the samples, or 3) rejection of the batch. So, a few days after the end of the production process, the batch is sold or destroyed. So, it appears that the “C” in quality control is merely a euphemism, since this pathway exerts no “control” over the quality of the product. So, in truth, only real-time measurements allow the production staff a chance to exert influence over the product as it is being made. Yes, PAT/QbD is the program to which I am referring. Assuming you make the leap into the 21st century, there could be a rocky learning curve in store for your analytical staff.

Let’s start with a less than obvious source of problems: the placebo, also known as the matrix. When a tablet of capsule is analyzed in a lab setting, using traditional “wet” methods (titration, HPLC, GLC, etc.) the analyte is measured by first removing it from the matrix.  For clinical trials, it is customary for formulators to replace the missing API with one of the excipients. This does not bother any “wet” analysis, but spectral methods “see” all the ingredients of the matrix and the ratio of excipients makes a difference with, say, NIR methods. In this case, when making a series of calibration tablets/capsules, the matrix is kept constant with the amount of API increasing.

Then, decide whether the analysis you are designing is for qualitative or quantitative purposes. For qualitative methods of finished products (e.g., clinical trials), it is sufficient to use a simple Principal Components Analysis (PCA) of the dosage form versus the placebo. This is good because it is 1) much, much faster than HPLC and 2) you can check 100% of the cards for each patient. We published the blueprint for this several years ago.¹ This paper shows how to generate and validate a working NIR equation for clinical supplies in a single day. For a last-minute check before packaging, the equation should include any potential products that appear similar to the one being packaged.

For quantitative analyses, the process is a bit more complicated. It requires generating samples of the dosage form involved with API levels outside the normal range expected in production. [This is necessary for several reasons, one of which is NIR is not good for extrapolating outside the range of standards used, just interpolating values.] The spectra need be obtained, either in reflection or transmission, an equation generated, then validated.^2,3 The USP, ICH, and FDA guidelines are easily found on-line, but they cannot cover all details of the procedure of generating a PAT/QbD (process) analytical equation.

One difficulty that is often overlooked is that the act of compressing a tablet imparts a great deal of energy on the granulation. As a consequence, a freshly compressed tablet is both warmer and denser than a sample allowed to “relax” for several hours (the normal interval between compression and the sample being analyzed in a laboratory). Thus, an equation generated in a lab setting may need to have a bias adjustment when transferred to production. [Of course, generating the spectra on-line and using these spectra to build a calibration v. lab data would solve that problem.]

Another mistake I have seen is using PAT tools improperly. One powerful tool used to determine which portions of the process are deemed “critical” and need be monitored and controlled is a Design of Experiment program. The software is used to generate a matrix of conditions (temperature, compression force) and ingredients (binder, disintegrant, lubricant) that could affect a critical performance attribute (dissolution profile, bioavailability, ejection force).

The appropriate number of batches are generated and analyzed and the results fed into the DoE program. The numerical and graphic outputs will show (statistically) which are critical points to measure so instrumentation can be inserted at the appropriate points.

I have seen this DoE approach used for quantitative and qualitative analyses of production products. That included ranges of the API from 80 to 120% of label claim as well as randomly varying the excipients. The “specificity” of the analysis was then attempted by using Principal Components Analysis (PCA). In short, PCA can be used to find variance between groups and, when plotted on a 2 or 3-D graphic will cluster samples that are similar and show separate clusters of samples that are different. So, for a properly built sample set, there will be clusters of 80, 85, 90, 95, 100, etc. % of label claim.

However, that ability is severely compromised when all the other excipients are also randomly varied. The resultant plots are more like a Jackson Pollack painting than groupings of like materials. The proper way to generate the samples for either a qualitative or quantitative set is to hold the excipient ratio constant and vary the API content. For a qualitative test, other than a yes/no of active vs. placebo, a smaller range of API or a comparison of dosage levels if, as in Oxycontin, there are several levels of active in similar looking (same colored and shaped) tablets, then a PCA can quickly ascertain which dosage level you have. Also, for quantitative tests, do not vary the excipient ratios and expect to use a PCA test.

Now, as we were discussing at the start of this column, a process analytical method isn’t the same as the lab type. When you decide to “dip your toes in the ocean” that is PAT/QbD, you will want to have the input of production, marketing, purchasing, QA and QC, and your statistics groups helping you decide which test to perform first. That is, which will be both an easy first step, yet give a satisfactory return on investment (ROI) to justify the expense and time.

Most companies opt for the simplest (and first to be run in a Pharma company) test: raw materials qualification.  This doesn’t require synthetic samples or extra work for the QC analysts. You merely scan incoming materials and allow the QC department to do its magic. Just as they normally check out RMs: particle size, moisture, ID, etc. You just gather “good” and “failing” batches, scan them, and label them accordingly. From these, you can build the “library” for qualification.

In all cases, I would advise you to attend some meetings where PAT is explored and caht with people who have had experiences, both good and bad. So, when you decide to join the “PAT nation,” you will have some ideas of equipment, validation, and which tests with wish to begin your program. Trust me, you won’t regret joining the 21st century.

References
1. “A General Test Method for the Development, Validation, and Routine Use of Disposable Near-Infrared Libraries,” J. Near Infrared Spectroscopy 9, 165-184 (2001) Ciurczak, E.W., C. Tso, G. Ritchie, L. Gehrlein.
2. “Validation of a NIR Transmission Spectroscopic Procedure, Part A: Validation Protocols,” J. Pharm Biomed Anal 28 (2), 251-260 (2002). Ciurczak, E.W., G. Ritchie, R. Roller, H. Mark, Cindy Tso, S. MacDonald.
3. “Validation of a NIR Transmission Spectroscopic Procedure, Part B: Application to Alternate Content Uniformity and Release Assay Methods for Pharmaceutical Dosage Forms,” J. Pharm. Biomed. Anal., 29 (1-2), 159-171(2002). Ciurczak, E.W., G. Ritchie, R. Roller, H. Mark, Cindy Tso, and S. MacDonald.

---

Emil W. Ciurczak, also known as the NIR Professor, has roughly 50 years of cGMP pharmaceutical experience and more than 35 years of Near-Infrared Spectroscopy (NIRS) experience with industries, universities, and instrument manufacturers. Emil teaches courses in NIRS, NIR/Raman, Design of Experiment, and PAT/QbD; has designed and patented hardware and software (including hardware and software related to anti-counterfeiting; written numerous technical texts and chapters; published extensively in journals; and presented hundreds of technical papers at many conferences, worldwide. He 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. For more info: emil@ciurczak.com.