To the Editor.—The National Cholesterol Education Program (NCEP) was launched in 1985 by the National Heart, Lung, and Blood Institute. Their Web site states, “The goal of the NCEP is to contribute to reducing illness and death from coronary heart disease (CHD) in the United States by reducing the percent of Americans with high blood cholesterol.” To know if high blood cholesterol is present, one must, of course, measure it. The NCEP provides guidelines for the analytical performance of cholesterol assays.1 These guidelines inform us about the required quality of cholesterol assays. This implies that if guidelines are met, the likelihood of incorrect treatment decisions due to analytical laboratory error will be low. These guidelines are used by manufacturers to design and verify cholesterol assays.
I pointed out during the 21st College of American Pathologists Conference (in 1991) that the basis for the NCEP analytical performance guidelines is flawed and described how it should be modified. I discussed these ideas with one of the NCEP guideline authors and published my talk in this journal.2 However, the NCEP guidelines, which have since been revised twice, still contain the same basic flaw. Since then, I have restated my original arguments3 and chaired a National Committee for Clinical Laboratory Standards (NCCLS) subcommittee, which advocates a different procedure.4
Miller et al5 provide an example for low-density lipoprotein cholesterol that illustrates the problem. In their example, 3 of 4 assays studied passed the NCEP performance goals using the NCEP method to estimate performance (total analytical error = 3.8%, 5.9%, and 5.8%). However these assays failed the NCEP goals if estimation methods were used that accounted for random interferences (total analytical error = 12.6%, 16.5%, and 41.6%), with the following percentage of observations exceeding the 12% NCEP recommended guideline: 8%, 11%, and 10%. These results could adversely affect patient classifications. Moreover, it is reasonable to assume that users would like to know that these assays fail the NCEP guidelines when estimation was performed correctly. Given that the estimation method would change as recommended, one could ask how should the NCEP analytical performance guidelines change? Perhaps this is an issue for the NCEP.
Isn't it time to reconcile the NCEP method with other approaches? I recommend that the estimation method in the NCEP guideline be changed to the direct estimation method of total analytical error, such as the method that follows NCCLS EP21. If a modeling method is used, it must be more complete than the one proposed by NCEP.