Evaluation of three first-generation ion-selective electrode analyzers for lithium: systematic errors, frequency of random interferences, and recommendations based on comparison with flame atomic emission spectrometry.

Ion-selective electrode analyzers for measuring lithium (Li/ISE) in serum became available in early 1987. We compared results for patients' samples from three of them vs results from flame atomic emission spectrometry (FAES). Within-run and day-to-day imprecision ranged from 0.01 to 0.03 mmol/L and 0.01 to 0.04 mmol/L, respectively. Comparing Li/ISE results (y) with the FAES results (x) gave the following equations: y = 1.063x - 0.035 for AMDEV's Lytening 2, y = 1.020x + 0.038 for NOVA's Model 11, and y = 1.030x - 0.027 for AVL's Model 985. Unexplained positive errors greater than 0.2 mmol/L were observed for two of the 90 patients' samples, but only a few additional excessively high values were seen in 3000 patients' samples run subsequently (Lytening 2). Causes of error in clinical Li/ISE measurements are still unclear; simply characterizing them as "matrix effects" does not correct the underlying analytical problem. An increase in pH from loss of CO2 gave low results on two of the three Li/ISE analyzers but did not change FAES results. Trimethylammonium bicarbonate used in a reconstitution solution caused extremely high Li/ISE results but did not change FAES results. Performance specifications to help reduce and correct these errors are recommended.