Abstract
Abstract
Background
Many clinical guidelines include laboratory measurements, which require accuracy across laboratories. This study aimed to determine whether testosterone measurements by a small subset of labs using commutable specimens could be used as a surrogate marker of accuracy across widely used testing platforms. It also assessed the frequency of categorization errors that might occur near the male hypogonadal threshold and whether performance on traditional surveys predicted performance using commutable specimens.
Methods
Three presumed commutable samples were created using human serum; Sample A was provided in a routine mailing of a traditional (non-commutable) proficiency testing survey (reference value 234 ng/dL, n = 1431 laboratories), Samples B and C were part of a routine mailing of an accuracy-based survey (reference values 277 ng/dL and 196 ng/dL, n = 118 and 138 laboratories).
Results
The overall mean bias from the reference method value was less than 2% for all 3 samples, but individual peer (method) group biases ranged −16% to +15%. Five percent of reported values for Sample A were above the hypogonadal threshold (264 ng/dL), falsely indicating “healthy state” (peer group miscategorization ranged 29% to 0%). Conversely, 44% of results for Sample B were below the hypogonadal threshold, falsely indicating hypogonadism (peer group miscategorization ranged 100% to 0%). Comparison of results between non-commutable and commutable specimens indicated that bias was different in direction and magnitude for 3 of 6 method groups.
Conclusions
Assessing performance using commutable specimens reported by smaller numbers of users may serve as a reasonable approximation for performance of method groups in general.