
Investigators are sometimes taken aback after discovering that most clinical laboratory testing is not performed in reference to an absolute standard. That is, plasma sodium measured by instrument manufacturer A may be slightly different from that by manufacturer B. Each method performs precisely and reproducibly, but the results and reference ranges are slightly different. Among common analytes, only cholesterol and hemoglobin A1c are related to an absolute standard so that we can say that an LDL in the same way. It may be slightly low or normal depending on the method.
Immunoassays are particularly susceptible to differences between methods. For example, two different FDA-approved immunoassays for free T4 run in CLIA-certified laboratories within the Washington University community have reference ranges of 0.59-1.17 ng/dL and 0.9-1.8 ng/dL, respectively. This disparity is large and meaningful.
CAP, the College of American Pathologists, sends proficiency testing samples to large numbers of clinical laboratories that provide insulin measurements. The following survey from 2011 illustrates the variation in reported insulin results between test methods. All are immunoassays but differ in antibodies used and details of the procedure. The data show consistent and large differences between methods:
In 2011, Core Lab evaluated several methods currently used to measure insulin. Samples were collected, diluted and frozen in aliquots. Insulin was measured by RIA, Elecsys, Immulite, Millipore ELISA and Singulex (fluorescent digital single molecule counting.) Elecsys and Immulite are automated immunoanalyzers. All of the methods, except RIA, use two antibodies which recognize insulin. The results show that RIA and Elecsys give higher values whereas Immulite results are consistently low. Millipore ELISA did not dilute linearly and gave unexpectedly low values at higher concentrations. These experiments were conducted using Immulite reagents manufactured before June, 2012. Immulite changed its method at that time and now claims that the low bias has been corrected. However, we still find Immulite values to be low when compared directly to Elecsys. Users should note that Immulite insulin reported before September, 2012 is not directly comparable to current Immulite insulin.
Core Lab currently recommends the Elecsys insulin assay, for several reasons. First, its values are midrange among immunoanalyzers and only slightly lower than RIA. Second, the analytical measurement range extends to less than 0.5 µU/ml so that unmeasurably low values are not usually seen in insulin-sensitive normal subjects as sometimes occurs with RIA and Immulite. Finally, Elecsys has good performance characteristics and is much less labor-intensive. We anticipate that it will be available for the foreseeable future so that it is suitable for ongoing studies. We will continue to offer the other methods for well-defined purposes after consultation with the investigators.