Zooming in on the Low End: Functional Sensitivity of Automated Testosterone Immunoassays

Presentation Number: THR-120
Date of Presentation: March 5th, 2015

Sonia L. Laulu1 and Joely A. Straseski*2
1ARUP Institute for Clinical and Experimental Pathology, Salt Lake City, UT, 2University of Utah & ARUP Laboratories, Salt Lake City, UT


Background: Measurement of testosterone provides an overall assessment of androgen status and aids in diagnosis of several endocrinopathies in men, women, and children.  Testosterone concentrations are commonly determined using commercial immunoassays.  However, there are well-documented concerns regarding the reliability of these assays at low testosterone concentrations.  This is particularly troublesome for patient populations where low testosterone concentrations are expected, such as women and children.  While mass spectrometry methods have proven to provide more accurate results, adopting these methods is not feasible for all laboratories.  The objective of this study was to assess the functional sensitivity (FS) of 2 automated testosterone immunoassays.

Methods: Residual serum samples were obtained following measurement of testosterone using LC-MS/MS.  Samples with similar results were pooled together to prepare 12 human serum pools (pool concentrations 1.2 – 962.5 ng/dL by LC-MS/MS) which were aliquoted and stored at -70°C until use.  All pools were tested for total testosterone using the Roche MODULAR E170 and Abbott ARCHITECT i2000SRusing 2 reagent lots and at least 2 calibrations.  Each pool was assayed once per day, 2 days per week for 5 weeks, totaling 10 replicates per pool. FS was estimated by fitting a power function to the imprecision data and calculating the testosterone concentration that gave a CV of 20% using Excel.

Results: Both manufacturers define limit of quantitation (LoQ) as the lowest analyte concentration that can be reproducibly measured with an imprecision of ≤ 20%.  The FS for the E170 was 15.8 ng/dL, which did not meet the manufacturer’s LoQ claim of 12.0 ng/dL. The true FS of the ARCHITECT could not be adequately assessed because none of the pools tested had CVs >20%.  However, the lowest pool measured on the ARCHITECT had a testosterone concentration of 4.1 ng/dL (1.2 ng/dL by LC-MS/MS) with a CV of 2.6%, which confirmed the manufacturer’s LoQ claim of ≤ 4.3 ng/dL.  This same pool was below the analytical measurement range of the E170 (< 2.5 ng/dL), therefore no data was generated.  Overall, the second lot of E170 reagent demonstrated significantly lower testosterone concentrations (p values ≤ 0.009); contributing to the final CV used in FS calculations.  With few exceptions, the E170 and ARCHITECT over-recovered with an average % recovery of 107 and 119, respectively.

Conclusions:  The E170 FS did not meet the LoQ claim provided by the manufacturer while the ARCHITECT demonstrated acceptable performance at low testosterone concentrations and met manufacturer’s claim.  In comparison to LC-MS/MS, both methods had a greater tendency to over-recover.  It is critical to understand the performance of testosterone immunoassays in populations with low testosterone concentrations, therefore patient-specific comparisons would prove useful to further assess assay sensitivity.


Nothing to Disclose: SLL, JAS