Posted by Sten Westgard, MS
I came across a very valuable paper in Clinica Chimica Acta on the stability of hematology controls for MCV. (Some of you are already guessing what this is going to be about...)
If you take a hematology control, how many SDs should you expect to see it shift at week 5 of use versus the first week of use? In other words, what is the SDI you should expect?
- 1 SDI
- 2 SDI
- 3 SDI
- 4 SDI
- 5 SDI
- higher than 5 SDI
The answer, after the jump...
The answer is, of course, all of the above. In the study, SDI values from 1.41 all the way up to 5.71 were observed. That's a lot of drift!
The paper which documented these findings:
The instability of commercial control materials in quality control of mean corpuscular volume, Kim SJ, Lee EY, Song YJ, Song J, Clinica Chimica Acta 434 (2014) 11-15.
Not only did the authors document the shift in mean of the controls, they also counted the number of false rejections and false negatives caused by this drifting control:
"Under the 1:2s rule, false positive rates ranged from 9.73% to 40.44% depending on control materials during 5 weeks. False negative rates were 0.89 to 3.56% for the same five-week use under the same rule."
Up to 40% false rejections! That's a laboratory drowning in noise. It's no wonder that a lot of work-arounds have been recommended, such as widening out to 4s or more, or arbitrarily shifting the control mean as the control ages and deterioriates, and ignoring all trend rules, such as the 2:2s, 4:1s, and 10:x. This is one reason why "Westgard Rules" aren't as popular in hematology - the performance of the control isn't stable enough to support the use of many of the rules.
Widening limits can certainly decrease the number of false rejections (but false negatives will creep up), the bigger question is whether or not those wide limits will decrease true error detection. If the Ricos goal for MCV is used to judge the methods (2.42% allowable total error and 1.26% allowable maximum bias), some of these methods just don't perform well enough to be in operation.
The RCPA has a different goal, which is more generous at 10% allowable total error past the level of 50. If that is true, we could make Sigma-metric calculations that would indeed justify widening the control limits to 3.5s or 4s.
Of course, the root cause of the problem is the stability of these controls. A logical solution would be to build better more stable controls that don't deteriorate, or to continually re-establish the proper mean with more data points. But either of those solutions would increase the cost of the control and/or the effort required by the laboratory. Our current mania with cutting costs while crossing fingers (hoping that quality isn't compromised) means that improving hematology controls won't come anytime soon.
I'm curious to know if other labs out there have seen similar problems