From Heather DeVries, Indiana University: I’ll make this one short and sweet…how many samples should be used in a cumulative summation study?
Heather, I’ve never used nor taught CUSUM, so I’m posting this in hopes of attracting comments. I did find a helpful reference, Westgard JO,Groth T, Aronsson T, de Verdler C. Combined Shewhart-cusum control chart for improved quality control in clinical chemistry. Clin Chem 1977; 23: 1881-1887. Looking for comments from our statistics experts.
Thanks for that information,
Thanks for that information, Dr. Riley! We used approximately 100 samples to establish our heparin therapeutic range, and 20 in each lot change to evaluate the cumsum. Our smaller sites must ask others for samples to accomplish this, which puts a burden on our larger sites, but we felt 10 samples wasn’t enough. I was really hoping with the revision to CAP question HEM.37880 would help alleviate this process for the smaller sites:
Revision states: “Laboratories in a local care network or system using the same instrument and same lot of an aPTT reagent, can share their nomogram of Heparin. However, a verification study using one laboratory as the reference laboratory to show that their results are comparable to each other must be performed.”
Conversations with CAP did not clear up my question as to whether one site could maintain the cumsum for a system if all sites use the same model analyzer and share PTT reagent lot numbers, as our system does (after initial establishment of the HTR).
This is for aPTT-based
This is for aPTT-based heparin therapeutic range (HTR) determination for a new lot of aPTT reagent, correct? In Olson JD, et al. Arch Pathol Lab Med. 1998;122:782-798, which is cited in the CAP Hematology and Coagulation Checklist, they do not stipulate the exact number of samples needed for the cumsum. They also don’t state the number of samples for the HTR either, for that matter, at least in that Olson et al. or the checklist. They do state in the Olson et al. paper that for the example data provided in the appendix “…..more than 30 specimens were used.” In Marlar RA, Clement B, Gausman J. Semin Thromb Hemost. 2017; 43: 253-60., it is stated that a minimum of 20 samples is needed for the aPTT vs. anti-Xa HTR (aka Brill-Edwards method). The Marlar et al. paper states that the cumsum is aPTT-based only, and no anti-Xa testing performed, so I assume from reading the Marlar paper that the cumsum uses the same amount of samples as the full HTR determination, but only anti-Xa is used. Hopefully that is clear as mud now? Maybe ask CAP directly?