Evaluation of the Bias of Serum Magnesium Measurements and the Commutability of Processed Materials.

Abstract

Measuring serum magnesium plays an important role in the diagnosis of diseases. The performance of serum magnesium measurements is most conveniently monitored by external quality assessment (EQA) or proficiency testing (PT) schemes. However, the commutability of EQA samples and calibrators is often unknown, and the effectiveness of EQA schemes is limited. We designed the present study to evaluate the matrix effects of processed materials and the accuracy of routine methods. Inductively coupled plasma mass spectrometry was applied as a comparative method, and 12 routine methods composed of 12 kits, their supporting calibrators, and a Hitachi 7180 automatic analyzer were chosen as the test methods. Serum from 48 single patients and 24 processed materials were quantified by the comparative and routine methods. The 95% prediction intervals and relative bias were calculated at three medical decision levels (0.60 mmol/L, 1.00 mmol/L, and 2.50 mmol/L) according to EP9-A2 and EP14-A2. The commutability of the materials was evaluated by comparing the values of the processed materials with the limit bias, and the accuracy of the routine methods was evaluated by the relative bias. The precision of all of the assays was good (total CV < 2.01%), and the decision coefficients were high, most were greater than 0.99. The 2013 and 2014 EQA samples showed positive matrix effects in some of the assays; the human serum pools were commutable in all of the assays; the certified reference materials were commutable in most of the assays; the calibrators showed matrix effects in some of the assays; the aqueous reference materials showed matrix effects in most of the assays and could only be used in the reference methods. For all of the assays, the range of the mean bias was -0.04 - 0.06 mmol/L (-4.52% - 7.20%), and the range of the expected bias at the three medical decision levels was -0.03 - 0.06 mmol/L (-4.27% - 9.30%) at 0.6 mmol/L, -0.05 - 0.06 mmol/L (-4.79% - 6.31%) at 1.00 mmol/L, and -0.364 - 0.10 mmol/L (-13.66% - 4.05%) at 2.50 mmol/L. For most of the assays, the biases were in an acceptable range, whereas the accuracy of some of the assays needs improvement. Human serum pools prepared from patient samples were commutable, and other materials still exhibited matrix effects.