Serum testosterone levels are positively associated with serum anti-mullerian hormone levels in infertile women

Anti-Mullerian hormone (AMH) and testosterone (T) both play distinct roles in the early stages of folliculogenesis. However, the relationship between serum T and AMH levels is poorly understood. This study aimed to investigate the association between serum T and AMH levels in infertile women. A total of 1935 infertile women aged 20–46 years were included in the cross-sectional study and divided into four quartile groups (Q1 to Q4) based on serum T levels. Compared to the subjects in the highest T quartile (Q4), those in the lowest T quartile (Q1) showed significantly lower AMH levels. After adjustment for age, body weight, body mass index and FSH, increasing T quartile categories were associated with higher AMH levels. Binary logistic regression analyses revealed that the odds for the risk of diminished ovarian reserve (DOR) were 11.44-fold higher in Q1 than in Q4 and the odds for the risk of excess ovarian reserve (EOR) were 10.41-fold higher in Q4 than in Q1. Our data show that serum T levels are positively associated with serum AMH levels and suggest that androgen insufficiency may be a potential risk factor for DOR; androgen excess may lead to EOR in infertile women.


Results
The characteristics of the 1935 infertile women included in the study are shown in Table 1. The average age was 35.1 ± 4.7 years (range 21-46 years); the average body mass index (BMI) was 22.4 ± 3.8 kg/m 2 (range 14.7-40.8 kg/ m 2 ). The mean serum T level was 0.33 ± 0.35 ng/mL (range 0.05-4.81 ng/mL) and the mean serum AMH level was 3.6 ± 2.8 ng/mL (range 0.03-22.08 ng/mL). Infertility causes included tubal factor (12.1%), male factor (11.6%), diminished ovarian reserve (13.0%), PCOS (9.6%), endometriosis (13.8%), uterine factor (8.0%), multiple factors (21.7%) and unexplained infertility (10.1%). In the group of women < 35 years (n = 887), the mean serum T level was 0.35 ± 0.31 ng/mL and the mean serum AMH level was 4.7 ± 3.1 ng/mL. In women ≥ 35 years (n = 1048), the average serum T level was 0.31 ± 0.38 ng/mL, and the average serum AMH level was 2.6 ± 2.0 ng/mL. The subjects were then categorized into four quartile groups (Q1 to Q4) based on serum T concentrations (Table 2). AMH, DHEA-S, body weight and BMI were positively associated with the T quartile category, whereas age linearly decreased as the T quartile category rose from Q1 to Q4 (all p for tend < 0.001). Body height, TSH, prolactin, 25-OH-vitamin D and FSH were not significantly different among the T quartile categories. Furthermore, compared to the AMH levels in the subjects in the highest T quartile (Q4), those in the lower T quartile (Q1, Q2 and Q3) demonstrated significantly lower AMH levels (p < 0.05).
Generalized linear models were used to assess the independent association of serum T quartile categories with AMH levels after adjusting for potential confounding factors, including age, body weight, BMI and FSH. Regardless of all women or different age groups (< 35 or ≥ 35 years) or different AMH groups (< 1.2, 1.2-5.0, or ≥ 5.0 ng/mL), AMH levels significantly increased in a dose-dependent fashion across increasing T quartile categories in the multivariate adjustment model as shown in Fig. 1.
The age-dependent distribution of serum T levels in all patients (n = 1885) is shown in

Discussion
To the best of our knowledge, the present study is the largest clinical study to investigate the association between serum T and AMH levels in infertile women. In this large retrospective cross-sectional study of 1935 infertile women, higher serum T concentrations were associated with higher serum AMH levels after adjustment for potential confounders. Consistently, infertile women in the lowest T quartile had a 11.44-fold higher odds for the risk of DOR than those in the highest T quartile. The odds for the risk of EOR were 10.41-fold higher in infertile women in the highest T quartile than in those in in the lowest T quartile. Androgens play important roles in the regulation of ovarian function. AR, expressed in oocytes, GCs and theca cells, is pivotal for normal follicular development 10,11 . AR is most highly expressed in the GCs of preantral and early antral follicles, and its expression decreases as the follicles grow 13 . Via the AR, androgens increase the FSH receptor and synergize with FSH to enhance follicle growth [19][20][21] . Moreover, androgens support follicle health Figure 2. Prevalence of (a) diminished ovarian reserve (DOR) and (b) excess ovarian reserve (EOR) according to serum T quartile categories. DOR was defined as serum AMH levels < 1.2 ng/mL; serum AMH levels ≥ 5.0 ng/ mL were defined as EOR. Q quartile.  www.nature.com/scientificreports/ by decreasing follicle atresia and GC apoptosis, and stimulating the proliferation and differentiation of GC 20,22,23 .
Although AR is not expressed in primordial follicles, androgens promote primordial follicle initiation 24,25 via indirect mechanisms, such as upregulation of insulin-like growth factor 1 expression 25 . The above information supports our results that lower T levels were associated with a higher risk of DOR. Studies have shown that women with DOR or POI demonstrated significantly lower serum T levels than controls 26,27 . On the other hand, androgen excess may lead to impaired ovarian function and dysregulated follicle development, displaying irregular cycles, oligo-ovulation and polycystic ovaries 28,29 . These findings agree with our results that infertile women with higher T levels had a higher risk of EOR. Thus, an optimal balance in androgenic actions is necessary for maintaining normal ovarian function. Serum T concentrations decline with age 14 . Thus, our study demonstrated an age-specific normal reference range for serum T levels to aid in identifying women who suffer from androgen insufficiency or excess ( Table 4). As mentioned above, androgens enhance FSH activity through increased FSH receptor expression 20,21 . FSH stimulates AMH expression 30,31 , which could inhibit the sensitivity of preantral follicles to FSH to avoid premature selection by FSH in the gonadotrophin-independent stage 32,33 . Therefore, Dewailly et al. proposed that androgens may promote AMH generation via enhancement of FSH-stimulated AMH expression 34 . Elevated AMH could attenuate FSH-induced aromatase activity, leading to an increase in androgen levels 33 . Moreover, via AMH receptor type 2 on the hypothalamus and pituitary, elevated AMH may boost GnRH-dependent LH pulsatility and secretion which stimulates androgen production in theca cells 35,36 . Taken together, it seems that androgens and AMH mutually stimulate each other. These results support our results that serum T concentrations positively correlated with serum AMH levels. Some studies also showed a positive correlation between serum androgens and AMH [37][38][39] . However, some studies revealed contradictory results, which indicated that androgens or FSH may have an inhibitory effect on AMH expression [40][41][42] . Thus, the accurate relationship between androgens and AMH remains unclear. Further studies with ideal experimental models are needed to clarify the relationship.
Serum T levels have been suggested to be positively associated with ovarian response 43,44 and even pregnancy outcomes 44,45 in women undergoing IVF cycles. Although some conflicting studies have shown that serum T levels do not predict IVF outcomes 43,46 , available data have indicated that T supplementation may improve ovarian response and IVF outcomes in PORs 47,48 . In a randomized controlled trial of 110 PORs undergoing IVF cycles, Kim et al. reported that pretreatment with transdermal T gel significantly increased AFC and reduced the day of stimulation and total dosage of gonadotropins. In addition, the numbers of oocytes retrieved, mature oocytes, fertilized oocytes, and good-quality embryos were significantly higher in the T pretreatment group than in the control group 16 . A meta-analysis of 7 randomized controlled trials conducted by Noventa and colleagues revealed that PORs receiving T therapy demonstrated higher numbers of total oocytes, MII oocytes and total embryos, as well as a higher clinical pregnancy rate and live birth rate than controls 47 . On the other hand, the addition of insulin sensitizing agents to suppress insulin resistance and excess androgen may ameliorate the results of ovulation induction in PCOS patients 49,50 . In the present study, we provided an age-specific normal reference range for serum T levels to help determine whether infertile women require agents for androgen enhancement or suppression (Table 4). Further large-scale, well-defined randomized controlled trials are still mandatory to confirm the effectiveness of androgen supplementation and androgen suppression by agents.
Several potential limitations should be taken into consideration when interpreting the data. First, the retrospective design of this study presented the major limitation. Second, since this is a cross-sectional study, a causal relationship could not be determined between serum T and AMH levels. Third, our study population only consisted of infertile women. We cannot be sure that our results would be applicable to the general population. Fourth, direct testosterone immunoassays have limitations for clinical use, particularly for low concentrations found in women and children 51 . This would be a relevant source of bias. Although high correlation (r ≥ 0.95) of serum T levels was observed between the ARCHITECT 2nd Generation Testosterone assay and the LCMS, the bias could not be totally excluded. Fifth, serum concentrations of free T, androstenedione, and sex hormonebinding globulin (SHBG) were not measured in our routine infertility evaluation. Thus, accurate androgen status would be uncertain in this study.
In conclusion, our data reveal an obvious positive association between serum T and AMH levels in infertile women. Additionally, the risk of DOR was significantly increased in a dose-dependent manner across decreasing T quartile categories; the risk of EOR dose-dependently increased across increasing T quartile categories. Long-term longitudinal studies are required to confirm our results.

Methods
Study design and participants. This was a retrospective cross-sectional study. Infertile women were first identified based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), code 628, from the clinical database in Kaohsiung Veterans General Hospital. To avoid any potential misclassifications, among the infertile women identified by the ICD-9-CM code, only subjects who received a complete infertility survey in the reproductive center of Kaohsiung Veterans General Hospital were selected. A total of 2476 infertile women were identified from May 2013 through March 2020. Then, we performed the chart review of these 2476 infertile women and selected the women who truly met the definition of infertility among them. Infertility was defined by the failure to achieve a successful pregnancy after 12 months or more of regular, unprotected sexual intercourse 52 . Moreover, we excluded the following subjects based on chart review: (1) subjected who experienced repeated surveys; (2) subjects who had extreme age (< 20 or > 46 years) (3) subjected who was diagnosed as primary ovarian insufficiency; (4) subjects who ever underwent ovarian surgery; (5) subjected who had a history of exposure to cytotoxic agents or pelvic irradiation for malignancy; (6) subjected who had androgen-secreting tumors; (7) subjected who was diagnosed as congenital adrenal hyperplasia (8)  www.nature.com/scientificreports/ women were finally included in the study. The study protocol was approved by the institutional review board in Kaohsiung Veterans General Hospital, with the identifier KSVGH20-CT11-03, and conformed to the "Declaration of Helsinki for Medical Research involving Human Subjects. " Need of informed consent was waived by the institutional review board in Kaohsiung Veterans General Hospital due to the retrospective design.
Biochemical measurements. For the infertility survey, we checked blood examinations including AMH, T, DHEA-S, FSH, luteinizing hormone (LH), estradiol, thyroid-stimulating hormone (TSH), prolactin and 25-OH-vitamin D levels. Serum AMH levels were measured by chemiluminescent immunoassay using the Access Immunoassay Systems, the Beckman Coulter enzyme-linked immunoassay (Beckman Coulter, Marseille, France). The analytical range of the lower limit of detection was 0.02 ng/mL. The intra-assay coefficient of variation (CV) was 3.0%, and the interassay CV was 7.0%. DOR was defined as serum AMH levels < 1.2 ng/mL based on the POSEIDON criteria 53 ; serum AMH levels ≥ 5.0 ng/mL, modified from the revised Rotterdam criteria 54 , were considered EOR in this study. Serum T was measured by chemiluminescent microparticle immunoassay using the ARCHITECT 2nd Generation Testosterone assay (Abbott GmbH, Max-Planck-Ring 2, Wiesbaden, Germany). The range was 0.04 ng/mL to 18.62 ng/mL. The assay had a limit of quantitation of ≤ 0.04 ng/mL and had a within-laboratory imprecision of ≤ 10% CV. Potential interference in the ARCHITECT 2nd Generation Testosterone assay from hemoglobin, bilirubin, triglycerides, protein and biotin was evaluated to be ≤ 10%. This assay had a correlation coefficient (r) of ≥ 0.95 for samples with testosterone concentrations ranging from 0.04 ng/mL to 10.09 ng/mL when compared to Liquid Chromatography-Tandem Mass Spectrometry (LCMS).
Continuous variables were presented as the mean ± standard deviation. The subjects were categorized into four quartile groups (Q1 to Q4) based on serum T concentrations. Quantitative variables were evaluated using the analysis of variance (ANOVA) and linear regression analysis among T quartile categories. Bonferroni's method was used for post hoc pairwise comparison in the ANOVA test. Generalized linear model was performed to examine the correlation between serum AMH levels and T quartile categories after adjusting for potential confounders including age, weight, BMI and FSH. Odds ratios (ORs) and 95% confidence intervals (CIs) for DOR and EOR among T quartile categories were assessed using binary logistic regression after adjustment for potential confounders (age, weight, BMI and FSH). All analyses were conducted using statistical software, Statistical Package for Social Sciences (SPSS) version 20.0 (Chicago, IL, USA). All statistical tests used a two-tailed α of 0.05, and statistical significance was defined as p < 0.05.

Ethics declarations. The study conformed to the ''Declaration of Helsinki for Medical Research involv-
ing Human Subjects'' . Additionally, approval was obtained from the institutional review board at Kaohsiung Veterans General Hospital, with the identifier KSVGH20-CT11-03. The study was performed in accordance with approved guidelines. Need of informed consent was waived by the institutional review board in Kaohsiung Veterans General Hospital.

Data availability
The datasets generated during the current study are available from the corresponding author on reasonable request.