George, Our lab participates in both CAP CGE and NASCOLA ECAT proficiency challenges. One of those assays, von Willebrand factor ristocetin cofactor (VWF:RCo) activity by the agglutination method, is difficult to judge in terms of performance. When looking at results, what is acceptable for criteria? I usually look for ±2 SD for other tests. Sometimes CAP publishes evaluation criteria such as ±3 SD for antithrombin, but I cannot find criteria for VWF:RCo. Any help with this one? Thanks, love your site!
Michele L. Drejka, Lead Technologist
Isermann Special Coagulation Laboratory
Frederick B. Cohen MD Comprehensive Cancer and Blood Disorders Center
Hi, Michele, and thank you for your question, which arrived November 15; I’ve been a little slow in responding as I was looking for some expert assistance, which I received from Dean Willett and Dr. Jon Geske at Precision BioLogic, and Dr. Emmanuel Favaloro, Department of Haematology, Institute of Clinical Pathology and Medical Research (ICPMR), Westmead Hospital, New South Wales, Australia, who provided this comprehensive response:
Hi all, apologies for delay in reply; I was away on leave;
Not sure that I understand the question correctly; is this in relation to acceptable performance in an external quality assurance (EQA) process (what the Americans call ‘proficiency testing,’ which abbreviates as PT and then causes confusion in coagulation labs in Australia!!) or in relation to in-house variability for internal quality control (QC)?
For EQA, every provider uses a different terminology and different acceptance criteria; not sure about CAP & NASCOLA ECAT, but NEQAS (UK), for example, uses the lowest and highest 10% of participant results to express ‘undesirable’ performance (I believe for all numerically based tests). In Australia, the RCPA uses ‘allowable limits of performance’ which is essentially an expert committee guideline which is test specific; for VWF:RCo, allowable limits are ±4 for values up to 10%, and ±40% for values >10%. These are fairly wide to take into consideration the test’s well-known variation. In practice, as working examples:
A group median result of 5% VWF:RCo will have an acceptable allowable limit range of 1–9% (i.e. 5 ±4), and any values reported below or above this will be identified as ‘undesirably low’ or ‘undesirably high’ respectively.
A group median result of 50% VWF:RCo will have an acceptable allowable limit range of 50 ±(40% of 50%) = 50 ± 20% = 30–70%, and any values reported below or above this will be identified as ‘undesirably low’ or ‘undesirably high,’ respectively.
Yes, very wide, but the RCPA will still get around 20% of participants falling outside the acceptable ranges in any given survey sample! As a comparator, allowable limits for VWF:Ag are ±3 for values up to 10%, and ±30% for values >10%, and allowable limits for FVIII are ±3 for values up to 10%, and ±25% for values >10%.
In terms of SD, however, this per-test variability is probably already taken into consideration, since SDs reflect a proportional variation; thus, a test result for FVIII including ±2SD will be narrower than a test result for VWF:RCo including ±2SD, and so ±2SD might be valid as an allowable range for both tests. ±2SDs basically covers about 95% of expected test results, and ±3SDs basically covers about 99% of expected test results.
In quality control, we basically advise to follow established rules such as Westgard. For this, I would refer you to the Westgard website. Hope this addresses the question, otherwise advise!
Many thanks to Dr. F for this excellent response. In the spirit of shameless self-promotion, I also direct Michele to Chapter 5, Quality Assurance in Hematology and Hemostasis Testing, in Rodak BF, Fritsma GA, Keohane EM. Hematology Clinical Principles and Applications, 4th Edition, Elsevier, 2012. My chapter illustrates standard deviation and the Westgard rules, especially those used in Hematology and Hemostasis. Geo.