Abstract
Performance evaluations of ion-selective field-effect transistors (ISFETs) have been carried out using both quality-control materials and whole arterial blood samples. Comparison of the results from these evaluations suggests that the whole blood evaluation may be more useful when assessing the value of particular sensors for clinical applications. The effect of outliers on the imprecision estimates is demonstrated for ISFETs and ISEs, both graphically and in the calculation of the estimates with an without the outliers present. Estimates of constant and proportional bias, against an alternative sensor, determined from the intercept and slope of the linear regression vary according to the regression method used. The bias estimates obtained for the K+ISFET against the Radiometer KNA1 using ordinary least squares regression are compared with the Deming/Mandel method and the three-group resistant line method of Tukey. The Thorn EMI ISFETs are demonstrated to have acceptable imprecision and only a small bias compared with direct ISE instruments for whole blood assay and can be considered suitable for incorporation into clinical instrumentation.