I love near-infrared. Anyone who knows me will tell you that I believe that you can measure almost anything using NIR. (Do you feel the but coming?) But any method can be stretched beyond reasonable expectations. During a recent conversation with a friend at USP, we were discussing dissolution prediction using NIR. Now, the only way I will ever learn is to listen to other scientists opinions, so I listened. He had reservations about using PAT measurements to release final products.
Well, I filtered some of his comments because, to a disinterested observer, it might appear that the PAT community is attempting to relegate the USP to the dustbin of history. However, many of his concerns were valid. Last year, we had discussions on whether a single NIR measurement was sufficient to replace the plethora of tests in a USP monograph for raw materials. Even I get a little nervous with a single method (one spectrum) to test ID, particle size, polymorphic form, degree of crystallinity, moisture and eye color of the supplier (I lied about the eye color).
We finally agreed that the NIR spectrum was not the only source of data for a particular raw material. In most cases, the vendor/source had been validated to be sure they were operating under cGMP and there was a valid paper-trail for the lots. In that case, the spectrum of the incoming raw material was, in reality, a last check in a series of tests. The preliminary tests were accepted when the final check was performed. Therefore, NIR wasnt the only test for incoming raw materials. In addition, orthogonal tests such as HPLC, Karl Fischer, and, yes, even USP spot tests are run at periodic times as a cross-check.
Similarly, many, if not most of the on-line tests proposed for PAT are excellent for measuring whether or not an individual dose is within API limits, coated within limits, etc. Earlier, the blend is shown to be well-mixed (as a physical chemist, I cannot use homogeneous to describe a powder mixture); the solvent levels are below pre-set limits; and the hardness is within limits. And then the batch is cleared by Quality Control for final release.
But, do we need the final release tests? After all, all the parts of the process signature have been controlled and/or monitored by other methods throughout the process. We may have used NIR (of course), Thermal Effusivity, Acoustics, etc. to show that the blending, lubrication, granulation, drying, and so forth have been stabilized and are constant (and resemble previous, passing, lots). But (there it is again)
Virtually all these tests are secondary and circumstantial, not specific. That means they are akin to the tests in a USP monograph, such as heating lactose with CuCl to give a red color this proves it is a reducing sugar. A series of circumstantial tests are complemented with a MIR spectrum. Then, lactose is proven.
When working with circumstantial tests, there are always potential interferences. This, for instance, is why I dont like PLS (Partial Least Squares) for blend uniformity end point. If we have variations in particle size, the spectra may NEVER resemble the PLS model enough to be declared at an end. It could very well be mixed, but it will not appear so. This same difficulty would also prevent a low amount of active from being noticed. It would first be necessary to prove that a particular process does not lose significant portions of the drug substance. In a similar vein, a series of chemical and physical parameters could appear to give the product, say, the correct dissolution profile, but the final product may NOT perform up to specs.
As much as I am smitten with NIR, I always tell my clients to ingratiate comparisons with the reference method into the SOPs for the analysis. There is a much better chance that the total amount of drug substance is within limits than that the dosage form will perform in the manner designed. Why? There are seldom weighing errors in either the raw materials or tablets, but, we do not have good control of hardness, and therefore, dissolution. [Hence, you see the need for PAT, in the first place.]
And then theres the U.S. legal system Just imagine the look on a jurys face when the lawyer representing a person who possibly suffered damage due to a commercial product* says, What? You didnt TEST your product before you released it? Is it worth the effort and potential costs to suffer this way? [Of course, this same lawyer would be the first to attack a company who didnt use PAT with, What? You had the ability to test thousands of tablets, but only chose to test 10 at the end?]
As we have been told, a single lot of product may take over six months to produce. If we look at the benefits of a full-blown PAT system, cutting the time of production from months to days, is it really necessary to stop the final release testing? It would be a validation of the PAT process, after all. We say that, under PAT, every batch is a validation batch, because PAT is not GMP. If we believe this, then the final tests are merely an extension of re-calibration of the on-line tests we do, anyway.
Just as we do vendor validation so we can just do an ID of the raw materials we accept, then we can consider PAT the vendor validation of the process and the compendial tests we run as checks we run, akin to ID tests of raw materials. The two or three days that a product undergoes final tests is miniscule when compared with the months saved by PAT. I vote that we keep them just make the darn specs real, OK?
* Remember, we can sue anyone, any time, for any reason. I, personally, think we have the best pharmaceutical products in the world, but
About the Author
Emil W. Ciurczak is Chief Technical Officer of the Cadrai Group. He has more than 35 years experience in pharmaceutical manufacturing, analytical R&D and regulatory compliance.