Introduction

Obesity in women is frequently associated with increased circulating androgen levels which can cause hirsutism and adversely affect fertility [1, 2]. In conjunction with menstrual cycle abnormalities and cysts of the ovaries, hyperandrogenemia constitutes the diagnosis of the polycystic ovarian syndrome (PCOS) [3], which additionally carries an increased risk for type 2 diabetes and cardiovascular diseases [1, 4, 5]. A reduction of circulating sex hormone-binding globulin (SHBG) levels, which results from a suppressive influence of obesity-associated insulin resistance and compensatory hyperinsulinemia on the hepatic release of this binding protein [6–10], further aggravates hyperandrogenemia since it increases the fraction of unbound, i.e., free and bioavailable, testosterone [11].

Bariatric surgery represents the most, if not the only, effective therapy for severe obesity [12–16]. Of note, two recent studies showed a normalization of preoperatively elevated serum androgen levels after various types of bariatric surgery along with a marked improvement of hirsutism in obese women who preoperatively displayed the diagnosis of the PCOS [17, 18]. Another study that tested 42 women undergoing gastric banding found an increase in SHBG levels and a reduction of testosterone levels as well as the free androgen index at 1 year after the operation [19]. However, all of these studies exclusively included premenopausal women so that it remains unclear whether the observed reduction of androgen levels after surgically induced weight loss is also present in postmenopausal women. Also, since previous studies almost exclusively included women with preoperatively elevated androgen levels, it remains unknown whether bariatric surgery also reduces circulating androgen levels that are within the normal range preoperatively.

To further elucidate changes in circulating androgen levels after massive weight loss induced by bariatric surgery, we assessed the serum levels of total testosterone, dehydroepiandresterone sulfate (DHEA-S), and SHBG in 36 severely obese women, six of whom were in a clinically defined postmenopausal state, before and at ~1 year after gastric bypass surgery. Circulating levels of glucose, insulin, and high sensitive C-reactive protein (hsCRP) were also measured in parallel to evaluate how putative changes in androgen levels relate to changes in glucose/insulin metabolism and the inflammation marker hsCRP.

Materials and Methods

Study Population

Thirty-six severely obese patients (body mass index (BMI) range, 38.7–55.1 kg/m2) with a mean ± SEM age of 41.2 ± 1.6 years (range, 20–58 years) were tested before and at ~1 year after a Roux-en Y gastric bypass operation. Only patients not taking drugs known to affect androgen levels, e.g., estrogens or metformin [20], were included in the study. Based on clinical criteria, i.e., no menses for >2 years and age >50 years, six women were classified as being in a postmenopausal state. All patients gave written informed consent and the study was approved by the local ethic committee.

Bariatric Procedure

After careful preoperative evaluation, eight women underwent a proximal and 28 women underwent a distal gastric bypass operation. In both the proximal and distal Roux-en-Y gastric bypass (RYGB) procedure, the largest part of the stomach was transected, thereby creating a small gastric pouch of 30 ml which was anastomized to the proximal jejunum. The diameter of this pouch-jejunal anastomosis was standardized to be 10 mm. In the proximal RYGB procedure, the biliopancreatic limb (duodenum and upper part of the proximal jejunum) was side-to-side anatomized to the jejunum 150 cm distal from the pouch-jejunal anastomosis, thereby creating a Roux-en Y or alimentary limb. In the distal RYGB procedure, the biliopancreatic limb was side-to-side anatomized to the ileum 60 to 100 cm proximal from the Bauhin's valve, thereby establishing a rather short common channel. The length to the biliopancreatic limb as measured from the ligament of Treitz was approximately 60 cm in the proximal and 60 to 100 cm in the distal RYGB procedure.

Laboratory Assessment

Androgen levels were determined in fresh serum samples that were obtained in the morning after an overnight fast of at least 8 h and were prospectively collected. Serum total testosterone and DHEA-S were measured by chemiluminescence immunoassay (both Beckman Coulter International S.A., Nyon, Switzerland) with an interassay coefficient of variance (CV) of 3.93 and 8.3 %, respectively. SHBG was determined by enzyme-linked immunosorbent assay (ELISA) on an Immulite analyzer (Siemens Medical Solution Diagnostics, Gwynned, UK) with an interassay CV of <4.9 %. Non-SHBG-bound, i.e., bioavailable, testosterone and non-SHBG-non-albumin-bound, i.e., free, testosterone were calculated by using the method of Vermeulen et al. [21]. In addition, the free androgen index (FAI) was calculated as the total testosterone/SHBG ratio × 100 [22]. Normal levels for androgens were defined as follows according to the reference ranges provided by the assay manufacturers: total testosterone (0.52–2.43 nmol/l), DHEA-S (premenopausal, <10.3 μmol/l; postmenopausal, <5.1 μmol/l), and SHBG (16–120 nmol/l). For free and bioavailable testosterone, we used the reference ranges (0.003–0.037 and 0.07–0.86 nmol/l, respectively) provided by Vermeulen et al. [21]. Insulin was measured by chemiluminescence immunoassay (Beckman Coulter International S.A., Nyon, Switzerland) with an interassay CV of 4.2 %. hsCRP was measured by ELISA on an Immulite analyzer (Siemens Medical Solution Diagnostics, Gwynned, UK) with an interassay CV of 2.3 %. Homeostasis model assessment (HOMA) was calculated as an index of insulin resistance (IR) by using the formula [23]: \( \mathrm{HOMA}-\mathrm{IR}={{{\left( {\mathrm{fasting}\,\mathrm{insulin}\,\left( {{{\mathrm{mU}} \left/ {\mathrm{I}} \right.}} \right)\times \mathrm{fasting}\,\mathrm{glucose}\,\left( {{{\mathrm{mmol}} \left/ {\mathrm{l}} \right.}} \right)} \right)}} \left/ {22.5 } \right.} \).

Body Weight

Height and weight were measured with the patients wearing light clothing but no shoes. BMI was defined as weight (kg) divided by height-square (m2). Waist circumference was measured to the nearest 0.5 cm midway between the lowest rib and the iliac crest. Hip circumference was measured horizontally at the level of the largest lateral extension of the hip. Percent excess weight loss was calculated by the formula \( \left( {\left( {{{{\left( {\mathrm{preoperative}\ \mathrm{weight}-\mathrm{current}\ \mathrm{weight}} \right)}} \left/ {{\left( {\mathrm{preoperative}\ \mathrm{weight}-\mathrm{ideal}\ \mathrm{weight}} \right)}} \right.}} \right)\times 100} \right) \) and percent excess BMI loss was defined by the formula \( \left( {\left( {{{{\left( {\mathrm{preoperative}\ \mathrm{BMI}-\mathrm{current}\ \mathrm{BMI}} \right)}} \left/ {{\left( {\mathrm{preoperative}\ \mathrm{BMI}-25} \right)}} \right.}} \right)\times 100} \right) \).

Statistical Analysis

Data are provided as mean ± SEM. Statistical analysis was based on paired Student's t-test. Relationships between variables were evaluated by Pearson's correlation coefficient and partial correlations. All data were tested for normal distribution by Kolmogorov–Smirnov and log-transformed when required. A P-value <0.05 was considered as significant.

Results

Table 1 summarizes the anthropometric data obtained before and at ~1 year after the gastric bypass operation in all patients as well as separately for pre- and postmenopausal women. At 1 year after surgery, patients had lost on average 43.1 ± 1.8 kg, which translates to 37.1 % reduction of their initial body weight, a reduction of their BMI by 16.6 ± 0.7 kg/m2 as well as 79.7 ± 15.0 % EWL and 86.4 ± 16.8 % EBL. Before the surgery, three women (all premenopausal) displayed type 2 diabetes which was remitted in all cases at the time of the follow-up assessment.

Table 1 Anthropometric measures before and ~1 year after gastric bypass surgery in pre- and postmenopausal women
Full size table

Before surgery, only two (5.6 %) and four (11.1 %) women displayed elevated levels of DHEA-S and total testosterone, respectively, but due to relatively low SHBG levels, 12 (33.3 %) displayed elevated levels of free testosterone and 14 (38.9 %) of bioavailable testosterone. However, the majority of women did not have elevated androgen levels.

There were neither correlations between the preoperative BMI and any of the androgenic measures nor with insulin and glucose levels (all P > 0.25). FAI was correlated with serum insulin levels (r = 0.38; P = 0.02) as well with log-transformed HOMA-IR values (r = 0.36; P = 0.03). Bioavailable testosterone levels were correlated with serum insulin levels (r = 0.34; P = 0.044) and tended also to correlate with log HOMA-IR (r = 0.38; P = 0.06). Total and free testosterone tended to correlate with serum insulin levels (r = 0.28; P = 0.09 and r = 0.33; P = 0.053) and log HOMA-IR (both r = 0.30; P = 0.08), while SHBG and DHEA-S levels showed no correlation with insulin and log HOMA-IR levels (all P > 0.12). Also, there were no correlations between any of the androgenic measures and glucose as well as hsCRP levels (all P > 0.29).

After the operation, serum levels of total testosterone, bioavailable testosterone, free testosterone, FAI, and DHEA-S markedly decreased, while SHBG levels significantly increased (all P < 0.01; Fig. 1). Changes in total testosterone (r = −0.70; P < 0.001), free testosterone (r = −0.93; P < 0.001), bioavailable testosterone (r = −0.93; P < 0.001), FAI (r = −0.957; P < 0.001), and DHEA-S (r = −0.48; P = 0.003) levels (calculated as 1-year postoperative value minus preoperative value) were inversely correlated with the respective preoperative hormone levels, indicating that patients with high preoperative levels displayed the greatest postoperative reduction. However, when only women with preoperatively normal androgen levels were analyzed, the remaining patients still showed a significant decrease in total testosterone (from 1.5 ± 0.1 to 0.9 ± 0.1 nmol/l; P < 0.001), free testosterone (from 0.025 ± 0.002 to 0.010 ± 0.001; P < 0.001), bioavailable testosterone (from 0.543 ± 0.036 to 0.208 ± 0.027 nmol/l; P < 0.001), FAI (from 6.6 ± 0.8 to 1.7 ± 0.2; P < 0.001), and DHEA-S (from 3.1 ± 0.3 to 2.6 ± 0.3 μmol/l; P = 0.001). It is noteworthy that the number of patients showing above-normal total, free, and bioavailable testosterone was reduced to zero at ~1 year after the operation and only one woman still showed an elevated DHEA-S level at ~1 year after the operation.

Fig. 1
figure 1

Individual as well as mean ± SEM circulating levels of total testosterone (a), free testosterone (b), bioavailable testosterone (c), free androgen index (d), DHEA-S (e), and SHBG (f) before and ~1 year after a gastric bypass operation in 28 severely obese women. Dashed gray line in a–c and e indicates the upper limit of the normal range. In e, the upper dashed gray line refers to the premenopausal and the lower dashed gray line refers to the postmenopausal upper limit of the normal range. Solid lines indicate premenopausal women and dotted lines indicate postmenopausal women. **P < 0.01; ***P < 0.001

Full size image

Pre- and postmenopausal women showed no differences in androgen levels before the operation (all P > 0.15). A separate analysis of pre- and postmenopausal women revealed that both groups showed a significant reduction in total testosterone, free testosterone, bioavailable testosterone, and FAI (Table 2). DHEA-S was the only hormone that did not decrease in postmenopausal women (P = 0.68).

Table 2 Laboratory data before and ~1 year after gastric bypass surgery in pre- and postmenopausal women
Full size table

Circulating levels of glucose, insulin, hsCRP as well as HOMA-IR were also found to be distinctly reduced at 1 year after the operation, and this was the case in premenopausal as well as in postmenopausal women (all P < 0.05). Correlation analysis revealed that relative changes—expressed as percentage change from baseline values—in free testosterone levels and FAI were positively related to relative changes in BMI (r = 0.39; P = 0.02 and r = 0.40; P = 0.02, respectively). Changes in SHBG were inversely correlated with absolute changes in insulin (r = −0.37; P = 0.029), and this correlation remained significant after adjusting for percentage BMI changes (r = −0.36; P = 0.049). Also, changes in free testosterone and bioavailable testosterone were correlated with changes in insulin (r = 0.42; P = 0.01 and r = 0.44; P = 0.008; respectively), and these correlations also remained significant after adjusting for relative BMI loss (r = 0.40; P = 0.03 and r = 0.42; P = 0.02, respectively) and for relative weight loss (r = 0.40; P = 0.03 and r = 0.42; P = 0.02, respectively). Lastly, absolute changes in FAI were correlated with changes in insulin levels (r = 0.45; P = 0.006) and changes in HOMA-IR (r = 0.36; P = 0.03). Again, the correlation with absolute insulin changes remained significant after adjusting for percentage BMI loss (r = 0.43; P = 0.02) and relative weight loss (r = 0.43; P = 0.02), while the correlation with absolute HOMA-IR changes failed significance after adjusting for these weight loss variables. None of the changes in androgenic levels as well as in insulin and glucose levels correlated with the absolute changes in BMI and hsCRP (all P > 0.17).

Discussion

In line with foregoing studies [17, 18], our data indicate that circulating androgen levels in severely obese women markedly decrease after a surgically induced weight loss. The present results extend previous findings by showing that this reduction in androgen levels also occurs in postmenopausal women as well as in women with preoperative normal androgen levels. Furthermore, the present data show that the reduction in free and bioavailable testosterone levels as well as that of the FAI correlates with postoperative changes in serum insulin levels. This finding suggests a close relationship between the regulation of androgen and glucose/insulin metabolism.

Preoperative SHBG inversely correlated with serum levels of insulin, supporting previous results [24] on a suppressive influence of insulin on hepatic secretion of the binding protein. Accordingly, the increase in SHBG levels after the operation was directly, i.e., independently of BMI changes, and inversely correlated with changes in insulin levels. Of note, serum SHBG concentration represents a major determinant of free and bioavailable testosterone as well as of FAI so that the presently found correlation between changes in these androgenic markers and insulin levels appears to be very reasonable. However, no correlations with insulin were found for changes in total testosterone and DHEA-S so that other factors than insulin appear to mediate the reduction in levels of these hormones after a surgically induced weight loss.

The clinical consequence of the reduction of androgen levels that are already within the normal range before gastric bypass surgery remains to be explored. On the one hand, increased SHBG levels and reduced testosterone levels can be assumed to reduce the risk of diabetes [25] and cardiovascular diseases [26]. On the other hand, however, it should be noted that low levels of testosterone and DHEA-S have been implicated to be associated with an increased risk for cognitive dysfunction, reduced mood, sexual drive, and frailty particularly in older women [27–29]. Therefore, it appears possible, in principle, that the reduction of androgen levels exerts some adverse effects after the surgery.

It should be noted that we neither systematically tested the women participating in our study for the presence of a PCOS nor systematically assessed menstrual cycle characteristics. These points clearly represent the limitations of our study, which should be overcome in further investigations in this field of research.

In conclusion, our results show that circulating androgen levels markedly decrease in severely obese women after gastric bypass surgery independently of their pre- or postmenopausal state and also in women with preoperatively normal androgenic levels. The correlations found of changes in SHBG, free, and bioavailable testosterone levels and FAI with postoperative changes in insulin levels suggest that a reduction in insulin levels may contribute to the normalization of androgenic hormone activity after weight loss in obese women. The consequences of the reduction of preoperative hormonal androgen levels in obese women are clearly in need for further investigations.