Linearity
The approximate linear range of each analyte is generally specified by the manufacturer or in the literature, but should be confirmed by each laboratory. If a linear relationship between analyte concentration and method response is not obtained, it will be impossible to report quantitative patient results by that method. If the linear range does not span the range of analyte concentrations generally expected in patient samples, the need for frequent dilutions may make the method too expensive and inconvenient.
The linear range is determined by testing specimens with defined concentrations. Linearity standards can be purchased from the instrument manufacturer, Maine Standards Company LLC, or the College of American Pathology (CAP). Ideally, standards should be suspended in the same matrix as patient samples. Typically, five samples with concentrations extending from 0 to 1.5 times the range of routine analytic interest should be used. The five levels should be spaced evenly throughout the range. At least one of these levels should be at the medical decision point.
Alternatively, dilutions of a high standard or patient specimen that is above the expected reportable range can be used to obtain a series of samples within the stated working range of the instrument. Each sample should be prepared by direct dilution of the original high sample rather than by serial dilutions. The latter should not be used because pipetting errors will be propagated through subsequent samples.
Triplicate measurements should be made on each sample and the mean of the triplicates calculated. The means of each measurement are plotted in an XY plot with the actual data points on the Y axis and expected values on the X axis. The graph is visually examined for linearity and precision. If the five data points fall on a straight line, the method is linear.
If the method is perfectly linear a straight line will be obtained that will have a slope of 1.00 and Y intercept of zero. The slope is an indicator of the % recovery; if the slope is 0.94 then recovery is 94%.Linearity can be accepted if the slope is 1.00 +/- 0.03 and the Y intercept is 0 +/- the within run precision.
A general rule of thumb is that a method can be considered linear if there is less than 10% variance between observed and expected values at each level. Nonlinearity should be suspected if the percent differences for the highest and lowest concentrations have one sign (e.g. positive number) and the intermediate levels have the opposite sign (e.g. negative number).
