Quality Standards
MU goals - can anything hit them?
Now that new performance specifications for permissible measurement uncertainty (pU) are here, it's time to put them to the test. Can today's instruments hit these targets?
What's the possibility of hitting pU?
Sten Westgard, MS
September 2021
- 6 of 8 assays of a Hitachi 7600 cannot meet pU goals.
- 8 of 9 assays of a Cobas c702 cannot meet pU goals.
- 6 of 8 assays of 2 ARCHITECT c16000s cannot meet pU goals.
- 6 of 7 assays of an AU 680 cannot meet pU goals
- MU pU goals, can anything hit them?
The CCLM article by Braga and Panteghini https://pubmed.ncbi.nlm.nih.gov/33725754/ set forth, after decades of debate, a set of performance specifications for measurement uncertainty. It's time to test the instrument performance to see if today's methods meet today's permissible uncertainty.
Conveniently, an article was just published in the Scandinavian Journal of Clinical and Laboratory Investigation which actually calculated the imprecision and measurement uncertainty:
Evaluation of 20 clinical chemistry and 12 immunoassay analytes in terms of total analytical error and measurement uncertainty. Ahmet Rifat Balik and Funda Gucel, https://doi.org/10.1080/00365513.2021.1955294
While the study evaluated 32 analytes, remember that Braga and Panteghini only published 13 performance specifications. The For the purposes of this study, we're evaluating one of the "big 5" instruments to see if it meets the MU APS.
The Braga and Panteghini pU and the performance
The imprecision was estimated from an entire year (2020) of IQC. Bias was estimated by participation in EQAS, an external quality assessment program, over the same time period.
Measurand | Milan Model | APS for standard MU, % | APS for desirable lab CV% | Level 1 CV | Level 2 CV | Bias |
glucose | outcome-based (best) | 2.00% | 1.00% | 2.93% | 2.77% | 1.53% |
sodium | Biological Variation (2nd best) | 0.27% | 0.14% | 1.22% | 1.2% | 0.55% |
potassium | Biological Variation (2nd best) | 1.96% | 0.98% | 1.41% | 1.61% | 0.6% |
chloride | Biological Variation (2nd best) | 0.49% | 0.25% | 1.45% | 1.51% | 0.47% |
total calcium | Biological Variation (2nd best) | 0.91% | 0.46% | 2.26% | 2.41% | 1.04% |
urea | Biological Variation (2nd best) | 7.05% | 3.03% | 4.19% | 3.89% | 2.05% |
alanine aminotransferase | Biological Variation (2nd best) | 4.65% | 2.38% | 3.55% | 3.2% | 1.87% |
Does pU Pass or Fail?
When the performance specification is applied to the imprecision measured on this instrument, what is the verdict? Note that the MU and pU are specifications that mostly ignore bias. Measurement Uncertainty can't be combined across all the levels if bias exists. So typically the approaches assume (1) either bias is so small (unspecified) that it can be ignored or (2) the bias varies over the long term, so it can be incorporated as just like another imprecision, or (3) the bias must be eliminated before any of the measurement uncertainty approaches can be applied.
Measurand | Milan Model | APS for standard MU, % | APS for desirable lab CV% | Level 1 CV | Level 2 CV | Level 1 CV pU? | Level 2 CV pU? |
glucose | outcome-based (best) | 2.00% | 1.00% | 2.93% | 2.77% | FAILS | FAILS |
sodium | Biological Variation (2nd best) | 0.27% | 0.14% | 1.22% | 1.2% | FAILS | FAILS |
potassium | Biological Variation (2nd best) | 1.96% | 0.98% | 1.41% | 1.61% | FAILS | FAILS |
chloride | Biological Variation (2nd best) | 0.49% | 0.25% | 1.45% | 1.51% | FAILS | FAILS |
total calcium | Biological Variation (2nd best) | 0.91% | 0.46% | 2.26% | 2.41% | FAILS | FAILS |
urea | Biological Variation (2nd best) | 7.05% | 3.03% | 4.19% | 3.89% | FAILS | FAILS |
alanine aminotransferase | Biological Variation (2nd best) | 4.65% | 2.38% | 3.55% | 3.2% | FAILS | FAILS |
As it happens, the paper authors calculated their own version of permissible measurement uncertainty, drawing heavily from Haeckel's 2015 formula [Haekcel R, Wosniok W, Gurr E, et al. Permissible limits for uncertainty of measurement in laboratory medicine. Clin Chem Lab Med, 2015;53(8):1161-1171]. These pU's are significant different than Braga and Panteghini's.
Measurand | pUEQAS % | pU% | Level 1 MU | Level 2 MU | Level 1 MU verdict? | Level 2 MU verdict? |
glucose | 14.5% | 7.41% | 6.76 | 6.5 | PASSES | PASSES |
sodium | 4.7% | 1.01% | 1.22 | 1.2 | FAILS | FAILS |
potassium | 12.4% | 6.33% | 2.85 | 2.8 | PASSES | PASSES |
chloride | 6.2% | 3.16% | 3.13 | 3.23 | PASSES | FAILS |
total calcium | 8.40% | 4.29% | 5.13 | 5.38 | FAILS | FAILS |
urea | 24.90% | 12.68% | 9.47 | 8.96 | PASSES | PASSES |
alanine aminotransferase | 34.70% | 17.69% | 8.24 | 7.68 | PASSES | PASSES |
It's not a scandal that different approaches to determine measurement uncertainty and permissible measurement uncertainty exist. Nor is it a surprise that different pU's will generate different judgments of method performance. In the world of total error, total allowable error, and Sigma-metrics, there have been different performance specifications for decades. And different verdicts can be reached depending on which TEa goal one chooses.
What can be said is the new pU's from Braga and Panteghini are very demanding, and further research is needed to see if other instruments and methods can achieve these goals. If no one can hit these targets, they are essentially a game without players.