Therapeutic Dose: Time to Extrapolate QbD?

Oct. 4, 2011
Preformulation, formulation and stability testing are all embracing QbD.

I have the opportunity to attend a goodly number of conferences on a full range of pharmaceutical topics. Over the last several years, I have concentrated on QbD/PAT, but now, I see that stability and preformulation are both embracing QbD.

It warms the cockles of my heart to see more and more drug manufacturers trying to implement PAT and QbD in their processes. However, it has taken them a bit longer to realize that the preformulation, formulation, and stability stages of the product’s lifetime are as critical to a living, growing, improving product as the actual production stage.

Why is that? Simply put, QbD (Quality by Design) depends on understanding all the ingredients, how they interact, and how the product comes together. It also should take into consideration how the product behaves over time: its stability. So, how do we use the QbD concept at both ends—i.e., how do we extrapolate our knowledge?

Let’s start with preformulation or, as I used to call it, “monkey work.” It used to (and often still does) consist of mixing the API with an equal amount of excipient listed on the GRAS (generally regarded as safe) list. This 50:50 mixture is then subjected to various conditions of temperature and humidity and tested for degradation at various time points. The analysis is usually done by TLC (thin layer chromatography) or HPLC and is time consuming. Not only slow, the 1:1 mixes tell nothing about interactions between several excipients and the API.

A better way would be to make ternary and quaternary mixes and look at them more often by a nondestructive method at more time points. Instead of weeks or months between sampling, mixtures can be examined by Raman or NIR weekly or, in the case of potentially labile API, daily. The samples that show changes can be “sacrificed” and analyzed by conventional means (mustn’t upset QA, must we?) for the FDA report.

In addition, we can look at different salts and polymorphs to give the formulators as much information as possible. While the synthetic chemists often settle on the salt or polymorph that gives the greatest yield, this may not always be the salt or form that lends itself to tabletting. So, with alternate information in hand, the formulators can stop a line of development and head in a new direction much sooner, saving money and patent time. Throw in data on particle size distribution from various vendors (oh, did I forget to mention we’re using Raman or NIR to do incoming RM qualification?) and moisture levels and polymorphic forms and crystallinity and we have even more information for the formulators and process people.

As for the stability end of the process, we can do something similar to the front end. Currently, we take bottles of the product and put them in stability “ovens” to be taken out at set time points (which haven’t changed in over 40 years). More to the point, we take ten or twenty dosage forms from a bottle of 50 or 100 for testing. (What happens to the other 80 or 90?) I would (humbly) suggest we do a similar thing to the front end.

If we look at all of the doses in each bottle, we would get a better picture of how the batch is holding up to stability stress. For example, I was involved in a project examining the pellication (where the capsule bulges instead of opening during dissolution) of capsules. We examined all the capsules in a bottle with NIR, placed them back into the bottle, resealed it, and placed it back into the stability oven.

When the spectrum showed something changing, those capsules were assayed (run dissolution) and compared with ones that did not show any changes. Over time, different lots could be seen to be different, although not why. However, when we couple this data with the well-kept data from a real QbD process, there is a lot to be learned. If we look at the percent of tablets or capsules from a particular batch that change (say 15% at three years for batch XYZ) and compare that with the results at the same time period for batch XXY, we can say which set of operational parameters, raw material suppliers, process equipment, etc. gave a better stability profile.

Since the combination of materials is fixed by the time a batch is prepared, we cannot (or will not) do much with current stability testing data beyond deciding to continue selling the lot or taking it off the market. But the (updated) stability testing can be used to look back at the parameters used for that lot and, perhaps, we can eventually redefine the design space within which we produce the product. Thus, we can extrapolate the QbD concept throughout the lifetime of the product.

A good idea, whose time has come.

About the Author

Emil W. Ciurczak | Contributing Editor