The effect of the immunoassay curve fitting routine on bias in troponin

Unlike many dose-response curves used in clinical chemistry, the immunoassay curve used to quantitate measurands is often sigmoidal rather than linear. Consequently, a more complex curve fitting model is required. Various models are available, but they can introduce bias, and there can be little awareness of why this error can be introduced.

These curve-fitting models include those based on the law of mass-action, empirical models such as splines or linearization models such as the log/logit function. All these models involve assumptions, which can introduce bias as the dose-response curve is 'forced' to fit or minimize the distance between the standard concentration points to the theoretical curve. The most common curve fitting model is the four or five parameter model, which uses four or five parameters to fit a sigmoidal curve to a set of standard points.

Measurement of cardiac troponin is an important element in establishing a diagnosis of acute myocardial infarction. We use troponin, a cardiac biomarker, to demonstrate the potential effect of the bias that the curve fit could introduce. Troponin is used for both rule-in and rule-out decisions at different concentrations and at either end of the dose-response curve. The curve fitting process can cause lot-to-lot reagent (and calibrator) variation in immunoassay. However, laboratory staff need to be aware of this potential source of error and why it occurs. Understanding how the error occurs leads to a greater awareness of the importance of validating new reagent/calibrator assessment using patient samples with concentrations at crucial decision points.