French and US endocrine societies recommend using GC-MS or RIA after purification (extraction + chromatography) to assess blood levels of testosterone in women. However, most of laboratories use automatized methods that have to be reserved to measure testosterone levels in men. The aim of this study was to show the consequences of analytical discrepancies of some immunological methods on the diagnostics values of testosterone levels assayed in women. Compared to GC-MS the correlations of the assayed levels varied (Spearman's rank correlation coefficients: 0.935; 0.793; 0.841; 0.852 respectively for RIA Immunotech™ with extraction and chromatographic purification; Testosterone Access-DxI800®; Testosterone Immulite 2000®; Testosterone II Cobas E601®). The testosterone levels allowed an accurate conclusion in 95.2 %; 75.8 %; 77.4 %; 89.8 % of patients, respectively. The agreement with GC-MS results was very good for RIA method (κ=0,840), moderate for DxI800® method (κ=0,414), moderate for Immulite® method (κ=0,467), good for Cobas® method (κ=0,667). Most of discordances are false hypertestosteronemia. The use of non recommended methods may leads to nosological errors (misclassification rates of 10 to 25% with automatized methods) that causes loss of chance in part of female patients.