Total testosterone and DHEAS levels as predictors of androgen-secreting neoplasms: a populational study.

Androgen excess affects between 2% and 10% of women. While the majority of these patients suffer from polycystic ovary syndrome, a few present with an androgen-secreting neoplasm. An elevated circulating total testosterone level and dehydroepiandiosterone sulfate (DHEAS) level have been proposed as screening methods for detecting ovarian and adrenal androgen-secreting neoplasms, respectively. To determine the predictive value of these tests for androgen-secreting tumors in a population of consecutive hyperandrogenic patients, we studied 478 consecutive untreated hyperandrogenic patients presenting over a ten-year period (1987-97). All had at least two of the following features: (1) oligomenorrhea (i.e. cycles > 35 days or < 8 cycles/year), (2) hyperandrogenemia (i.e. a total or free testosterone, or DHEAS > 95th percentile of controls), or (3) hirsutism (i.e. a modified Ferriman-Gallwey score > or = 6). None of these patients had a prior diagnosis of an androgen-secreting neoplasm. Basal levels of testosterone and DHEAS were determined in all patients, with transvaginal sonography and an adrenal computed tomography scan in select individuals. Of the 478 patients included, 65% had hirsutism and oligomenorrhea; 20% had hyperandrogenic oligomenorrhea; and 15% had hirsutism and hyperandrogenemia, without overt oligomenorrhea. Overall, 11 (2.3%) patients had a total testosterone > 8.7 nmol/l (250 ng/dl), of which one actually had an androgen-secreting neoplasm (i.e. true-positive). This postmenopausal patient presented with rapidly progressive virilization, and demonstrated an ovarian hilar cell tumor at surgery. The calculated sensitivity of an elevated testosterone level (> 8.67 nmol/l) for a neoplasm was 100% (1/1), the specificity was 98% (467/477), and the negative predictive value was 100% (467/467), but the positive predictive value was only 9% (1/11). Ten subjects had DHEAS levels > 16.3 mumol/l (6000 ng/ml), and none was diagnosed with an adrenocortical tumor. Although the sensitivity and positive predictive value of a high DHEAS for a neoplasm could not be calculated due to the absence of a test case, the specificity was 98% (468/478) and the negative predictive value was 100% (468/468). These data suggest that the measurement of testosterone and DHEAS is not a cost-effective method of screening for these tumors, due to the low frequency of the disorder and the fact that clinical evaluation alone is often sufficient screening.