Analytical bias for cholesterol and the percent of the population deemed at risk for coronary heart disease

The effect of methodological bias on the population at risk is dependent on the location of the reference value in the distribution of the population. We fitted the cumulative distribution for cholesterol to a rational function and calculated the apparent reference values for four biased methods: Technicon SMAC (2.6%), DuPont aca (4.0 to 4.8%), Kodak DT-60 (-2.0 to -5.5%), and BMD Reflotron (-7.4 to -7.8%). With the true and apparent reference values for cholesterol and the rational function, we determined the percentage increase or decrease in the population deemed at risk for coronary heart disease. The population at risk increased by as much as 48% for methods with positive bias, and decreased by as much as 54% for methods with negative bias. If we restrict the percentage of the population incorrectly diagnosed to 3% and use reference values (cut points) recommended by the National Cholesterol Education Program, the maximum allowable methodological bias would be 1.6% for positive bias and -1.55% for negative bias. Therefore, an absolute methodological bias of 3% (as recommended by the Laboratory Standardization Panel) may be too liberal.