USP's Gary Ritchie with an Update on PAT Standards Development

Feb. 6, 2007
Emil Ciurczak met with Gary Ritchie at the 2007 IFPAC conference to discuss the latest in standards development for process analytical technologies (PAT) at USP. Listen to the interview or read the transcript here.

At IFPAC, Emil Ciurczak interviewed Gary Ritchie, Scientific Fellow for Process Analytical Technologies (PAT) at USP, for a review of PAT standards development within USP, with a special focus on 1119, the general chapter for NIR spectroscopy. The interview transcript follows. To hear the audio file of this interview, click on the Download Now button below the transcript.

EC: So, how many chapters have you worked on and gotten for the USP?

GR: Let’s start with 1119. I didn’t bring that one in — it was pre-existing — but it needed some revision and so we can count the committee work that ensued when I came on board with 1119, for NIR spectroscopy. Then there were the chapters that I did bring in for Raman spectroscopy, acoustic emission and effusivity.

And now, we are looking at revising some of the waters chapters. 643 and 645 were originally devised as lab off-line tests, but were revised to reflect the fact that you can do on-line monitoring.

One of the things that I recognized early when I joined the USP, with respect to the PAT initiative, was the fact that the USP actually had the first on-line general chapter for process monitoring for pharmaceutical waters, published around 1997.

Other chapters now under consideration by USP’s general chapters committee focus on chemical imaging and chemometrics.

EC: I think you’re being a little modest about the work on 1119. After all it was back in 1998-99 when you held the meeting and brought people from industry, academia and the USP together to propose a chapter, and if I remember, you browbeat them into coming up with a usable document for submission in nine months. Compared to some of the committees that work on these things through several ice ages, I think that was rather commendable.

GR: Other than the science that I’ve always been involved in at the analytical level, I grasped very early the importance of understanding how people use a technique and why they need to use a technique. One of the things that 1119 had suffered from for many years was the fact that, while it was in the Compendium, it didn’t receive much attention.

That’s because it focused mainly on system suitability or system verification. It existed as a way for someone who wanted to implement NIR measurement, to know how to calibrate the wavelength for photometric accuracy and photometric precision.

That would have been great if a lot of people were using NIR. But when the chapter existed [in that form], not that many people were using NIR.

What they really needed to know, from an analytical point of view, especially if they were going to use the technique for pharmaceutical work, was, "How am I going to generate and validate a method?" And the chapter really didn’t answer that question. It instead answered the question, "I have an instrument. Does it measure something?"

EC: So the new chapter basically has some hints on how to do quantitative and qualitative validations?

GR: Identification and all the things that you, in your early days, actually taught the industry about what the techniques were useful for. And my work at Purdue, and yours, answered the questions: “Okay, we know that NIR can be used to take this measurement. How do we validate it so that FDA and other regulators actually accept it?”

EC: The other thing was that you had been instrumental in getting the new NIR standards for USP because NIST had given up that standard. They no longer make it, yet it is required to have some traceable standard for any technique. The FDA really likes to have everything required for IQ, OQ, PQ, for method validation. For instance, in chromatography, you need a USP standard to run against. Now for spectroscopy, you have a USP standard to run against.

GR: Absolutely. When I joined USP it became obvious that we needed to make the standards more useful, so that they went beyond merely answering questions that regulatory agencies could have regarding a submission. How could regulatory authorities independently verify that a person was using the chapter and verify suitability?

How, for example, could FDA verify the fact that the suitability values reported back in some lab notebook stating that “on a certain date, a given wavelength was tested.” How could they verify that that the testing actually was done?

EC: They couldn’t. It’s pretty much like standardized thermometers or anything else. “Traceable” is the key word.

GR: That’s correct. And the void that was left by SRM 1920, the previous NIST NIR standard that you referred to, needed to be filled by an improved 1119 Chapter. It really was a no brainer.

EC: So you’re doing this with multiple testing of standards on grating and Fourier transform instruments?

GR: The specification has not changed, it’s actually been expanded. When the original 1920A standard spec was originally tested, during a "round robin" of collaborative studies involving six equipment vendors, it specified three wavelengths, and established 1 nm and 1.5 nm wavelength specifications.

What we’ve done in the certification of the new standard that will be coming out shortly, is that we’re actually certifying standardizing seven wavelengths. The user, depending on which portion of the spectrum that he or she wishes to operate in, or even the whole spectrum, can choose any three wavelengths for qualifying any instrument for its intended use in measuring a process or product.

EC: You’ve added a peak above 2000 nm?

GR: Some say that this may not extend the standard’s usefulness. I’ve heard opinions, both pros and cons on the additional band above 2000, but its usefulness remains to be seen. I know that, for fiber optic work, you generally don’t want to read anything above 2100 nm or so because of internal attenuation you get with the energy on the fiber optic material. If you go ahead and use this area of the spectrum for calibration sometimes you’ll get a lot of noise and scatter. But that’s only a problem with fiber optic work.

If someone is using a model for even, say, qualitative work, but using a direct hemispherical detection system, that end of the spectrum is useful up to 2500nm. I think they get an additional advantage with this extra peak.

EC: Thank you. I appreciate it.