This message from Randy Trussell at Tenet Health arrived January 4. Randy, sorry for the delay, and I hope this is still helpful.
(There is a direct response from Dr. Olson appended to the end of George’s response.)
Our lab established the unfractionated heparin (UFH) therapeutic range last year using the Brill-Edwards approach. This year we are evaluating our partial thromboplastin time (PTT) reagent lot change using the cumulative summation (CUSUM) of differences method described by Dr. John Olson in the October, 2004 CAP Today. Dr. Olson suggests the computing the CUSUM of the mean values to correlate old and new lots.
An instructor I had at instrument training said to use the CV% instead of the mean to compare the differences. Is one better that the other, or are they both adequate? Also, should the mean PTT values come only from patients receiving UFH only, or from all patients used in the lot-to-lot?
Thanks for you help in this matter,
Randy, thank you for your question. I’m turning this one over to our statistics experts out there for help.
Follow-up late Monday, 1/25/2010:
Randy, here is a direct response from Dr. Olson:
“Keep in mind that the goal is to determine if the reagents are different. Mean and CV provide different information when comparing two populations. It is possible for the CV of the datasets from two reagents to be the same, but for the means to be very different. CV is a measure of the variability of the data and can tell you if the imprecision of the assay using the two reagents is similar, but you need to compare the mean or median to see if they are reflecting the same values on the same dataset. Both CV and mean/median are useful but measure very different things. Remember, the purpose of the exercise is to stabilize the method for monitoring heparin and only specimens from patient who are anticoagulated with heparin should be used.”
Thanks to Dr. Olson for clarifying this. Geo.
No comments here.