Estimation of the minimal preanalytical uncertainty for 15 clinical chemistry serum analytes

We sought a model to estimate preanalytical uncertainty of blood samples collected and processed by using optimal procedures.

Optimal preanalytical handling of blood samples included use of a loosely fastened tourniquet, wide bore needles, recommended clotting time and centrifugation speed, and minimal storage before analysis. Blood was collected from each arm of 20 volunteers into 2 rapid-serum tubes and 2 serum-separation tubes. Linear mixed-effects models were used to estimate the between-venipuncture SD, the preanalytical SD (excluding venipuncture), the measurement repeatability SD, and systematic differences between the tubes and between venipunctures.

No significant systematic differences were found between successive venipunctures. However, statistically significant mean differences were seen between serum-separation tubes and rapid-serum tubes for 7 of the 15 analytes. The preanalytical SD (excluding venipuncture) for lactate dehydrogenase (3.2 U/L, 95% CI 2.8-3.7) was significantly higher than the SD for measurement repeatability (1.9 U/L, 95% CI 1.7-2.1). For potassium both the preanalytical SD (excluding venipuncture) (0.092 mmol/L, 95% CI 0.080-0.11) and the between-venipuncture SD (0.075 mmol/L, 95% CI 0.048-0.12) were significantly higher than the measurement-repeatability SD (0.031 mmol/L, 95% CI 0.028-0.035). For glucose the between-venipuncture SD (0.20 mmol/L, 95% CI 0.14-0.27) was significantly higher than the preanalytical SD (excluding venipuncture) (0.07 mmol/L, 95% CI 0.06-0.08), and the measurement repeatability SD (0.057 mmol/L, 95% CI 0.051-0.064).

By applying linear mixed-effects models we have estimated the minimal preanalytical uncertainty that will influence all patient results.