RE: inter assay controls and SPC

["John McConnell" <wysowl@msn.com.au>]

Shawn Way wrote,

One of the questions that was recently given to me was if SPC is applicable
to intra and inter assay controls for biologics testing...It is my
contention that the variability and repeatability of the assay should be
determined via SPC on the control sample.

Does this application truly fall under SPC or traditional statistical
methods?

If I understand you correctly, what you outlined is very common in the
biological element of the pharmaceutical industry, particularly where
chromatography is involved.  Biosynthetic Human Insulin manufacture and
testing is a good example.

If that is the case, I can assure you that much more information can be
gained from intelligent use of SPC approaches than by any other I have seen
used.  In the pharma labs in which I have worked, separate SPC charts are
made for control and production samples.  I have examples where production
data trended significantly and yet this trend was eventually traced to the
test itself.  The controls had the same trend, but no one had been alarmed
because the data from the control samples fell within specifications.

I urge you to try this.  Take the production and control data for one set of
controls (if they are duplicate or triplicate samples, convert them to an
average and then plot this average as a Single Point and Moving Range
chart).  Be prepared to sort your data by technician and by
instrument/column.

Never has a new client lab had controls that were stable...or even close to
stable.  This includes pharmaceutical and aerospace components.  Every case
showed special cause points as well as significant shifts in the mean.

And these are the controls, not the production samples.  Scary stuff, if you
use pharmaceutical products.  In one case nearly all failures of final
release tests were traced to excessive variability in the labs.

All these labs were able to significantly improve precision, reduce or
eliminate repeat tests, reduce deviation reports and speed up turnaround
time.  Be aware that all the charts in the world change nothing.  We must
actually work hard at reducing variation as opposed to meeting standards.

I have a newsletter on reducing variation in labs if anyone is interested.

Have Fun!

John McConnell