When I started doing NIR, I was confronted with some amazing “truths” about the technique, many of which turned out to be urban legends. Since it had evolved, instead of being developed at university level (i.e., mass spec, NMR), there were a number of misconceptions (e.g., samples “have” to be ground, since there is no way to account for particle size) that held back method development (in Pharma, for sure) for years. Since I am a physical chemist, taught physics and optics, most of the “urban legends” about NIR sounded strange … and took me years to disprove; now NIR is widely used by most Pharma companies.
In a similar manner, there are so many misconceptions about PAT and QbD, it is not entirely surprising that neither are universally practiced. The two “truths” I wish to address here are:
1. PAT is expensive and a luxury;
2. We already design our products, so the operational parameters on our MMF are the design space.
The first is interesting, if only because it contains a grain of truth: PAT does require an initial fiscal outlay. But if we look at PAT as a part of operational costs, it takes on a slightly different pale. Traditional Pharma production is often designed to account for (at least) 10% loss of materials. Add in reworked products and destroyed products and the “cost” of not having the process understood and under control can be 25% of the product cost.
When process instrumentation, equipped with proper computerized controls, is installed in a process, other than maintenance costs, you only pay for it once. With uncontrolled losses, we pay for the problem throughout the lifetime of the product. When you add to the “costs” the time-to-market of a multi-month production (per lot), the so-called “luxury” of PAT suddenly becomes as superfluous as health or car insurance.
The second “truth” is harder to dispute, because of definitions of words. The word “design” comes to mind immediately. Of course, everything that exists comes from some sort of design. Human beings are the result of several millions of years of “hunt-and-peck” experimentation by Nature. Like software, new versions come along constantly (with as many bugs as Microsoft has) and, despite numerous revisions, we still have flaws (e.g., male pattern baldness and snoring).
The defenders of the status quo will argue that every process has a “design space,” consisting of how long we mix, how long we dry, how fast and hard we tablet, etc. What is not spoken is that the “operational parameters” for most products are determined by one-at-a-time experimentation and tweaking of parameters. [Keep in mind human DNA has about a 7% contribution from viruses — so much for “intelligent design.”] So much of the “design” is done univariantly or one-parameter-at-a-time. While this can lead to a reasonable product, it seldom shows the interactions between parameters. And, despite claims of a designed product, we have to ask: “Why do we still have a ±10% spec, if we are under control and know what we’re doing?”
The blending and granulating steps in solid dosage production has been compared with detergents, paints, polymers and cake mixes. The question is, why can’t we learn from consumer products? In the industry’s defense, we not only need to have the ingredients blended well, but also the solid dosage form needs to have characteristics such that it releases the active at the proper time and rate into the bloodstream.
However, this makes “the way we’ve always done it” approach even more problematic. With the increase in poorly soluble APIs, prolonged release dosage forms and lower dosage (higher activity) APIs, the matrix becomes more and more important.
The QbD approach to a design space is multivariate; that is, it takes all the parameters into account in the design of a drug delivery system. There are many inexpensive algorithms available (with training courses) for Design of Experiment, so there should be no longer any reason for modern drug producers to rely on “we’ve always done it this way” as a design paradigm.
Quality by Design means using the suggestions in ICH Q8 and 9 to intelligently design, using good science, the operational parameters for producing a proper dosage form. Now can we kill these two “urban legends,” at least and move on to some others?