Use of medians and "average of normals" of patients' data for assessment of long-term analytical stability.

Using results from patients, we studied several methods for long-term (years) monitoring of the analytical stability of various chemistry tests. For each test, we obtained on a bimonthly basis histograms of > or = 200 to approximately 2000 data points and determined the median, mean, and mean of those data points in the reference range ("average of normal" or AON). Examining approximately 4 years' data, we found that the medians and AON are extremely stable for most tests, and that all observed shifts in the medians or in the AON could be explained by expected or unexpected changes in the laboratory. The means and peak values of the Erlang (gamma) distribution were less useful, owing to shifts caused by extreme outliers, especially for enzymes in serum. We assumed that the underlying patient population and testing patterns were stable. We recommend the use of medians or the AON of patients' data under defined conditions for monitoring the long-term analytical stability of certain tests involving serum or whole blood.