Assay-dependent reclassification of biochemical hyperandrogenism in female androgen testing: CLIA versus LC-MS/MS under common clinical decision thresholds.

OBJECTIVES: Female testosterone and androstenedione results are commonly interpreted as either above or below a reporting threshold. This binary reading is practical, but it also means that assay differences may become classification differences. We investigated whether routine chemiluminescent immunoassay (CLIA) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) give the same biochemical hyperandrogenism classification when commonly used decision limits are applied to paired female androgen results. METHODS: We reviewed 495 female reproductive endocrine samples with paired CLIA and LC-MS/MS results for testosterone and androstenedione. The routine thresholds used in reporting, testosterone >2.5 nmol/L and androstenedione >10 nmol/L, were applied to both platforms. Biochemical hyperandrogenism was classified by testosterone alone, androstenedione alone, or either analyte being positive. Discordance referred to the situation where two platforms assigned different classifications to an identical sample. Sensitivity analyses were performed by excluding samples located within ±10%, ±20%, or ± 30% of the cut-off thresholds. RESULTS: The two analytes did not show the same assay pattern. For testosterone, positivity was low with CLIA, involving 13 of 495 samples (2.6%), but increased to 60 samples (12.1%) with LC-MS/MS. Androstenedione was in the opposite direction: 245 samples (49.5%) were positive by CLIA, compared with 183 samples (37.0%) by LC-MS/MS. Because androstenedione contributed most of the positive classifications, the combined definition also gave a higher positivity rate with CLIA than with LC-MS/MS [246/495 (49.7%) vs. 189/495 (38.2%)]. The resulting mismatch was 53 samples (10.7%) for testosterone-based classification, 162 (32.7%) for androstenedione-based classification, and 161 (32.5%) when either analyte was considered. Removing samples close to the thresholds reduced the mismatch, but some discordance remained even after the ±30% exclusion: 2.8%, 11.3%, and 8.1%, respectively. CONCLUSIONS: In this female reproductive endocrine cohort, applying common decision thresholds to CLIA and LC-MS/MS results did not produce the same biochemical hyperandrogenism classification. The direction of reclassification depended on the analyte: LC-MS/MS increased testosterone-based positivity, whereas CLIA increased androstenedione-based and combined positivity. The findings do not define LC-MS/MS-specific cut-offs, but they show why female androgen results near decision limits-or results that conflict with the clinical picture-should be interpreted with the assay method in mind.