Performance criteria for testosterone measurements based on biological variation in adult males: recommendations from the Partnership for the Accurate Testing of Hormones.

BACKGROUND

Testosterone measurements that are accurate, reliable, and comparable across methodologies are crucial to improving public health. Current US Food and Drug Administration-cleared testosterone assays have important limitations. We sought to develop assay performance requirements on the basis of biological variation that allow physiologic changes to be distinguished from assay analytical errors.

METHODS

From literature review, the technical advisory subcommittee of the Partnership for the Accurate Testing of Hormones compiled a database of articles regarding analytical and biological variability of testosterone. These data, mostly from direct immunoassay-based methodologies, were used to specify analytical performance goals derived from within- and between-person variability of testosterone.

RESULTS

The allowable limits of desirable imprecision and bias on the basis of currently available biological variation data were 5.3% and 6.4%, respectively. The total error goal was 16.7%. From recent College of American Pathologists proficiency survey data, most currently available testosterone assays missed these analytical performance goals by wide margins. Data from the recently established CDC Hormone Standardization program showed that although the overall mean bias of selected certified assays was within 6.4%, individual sample measurements could show large variability in terms of precision, bias, and total error.

CONCLUSIONS

Because accurate measurement of testosterone across a wide range of concentrations [approximately 2-2000 ng/dL (0.069-69.4 nmol/L)] is important, we recommend using available data on biological variation to calculate performance criteria across the full range of expected values. Additional studies should be conducted to obtain biological variation data on testosterone from women and children, and revisions should be made to the analytical goals for these patient populations.