Effects of Weight Loss on Testosterone, Sex Hormone Binding Globulin, Adiposity, and Insulin Sensitivity in Women and Men

Abstract

Objective:

Testosterone and glucose disposal (Rd) are positively associated in adult men; the opposite is reported in women. Sex-specific relationships between testosterone or sex-hormone binding globulin (SHBG) and Rd in men and women with and without obesity and following weight loss were examined.

Methods:

Adult men and women (n=27/28; BMI=20–41kg/m²) underwent measurements of body composition, Rd, SHBG, bioavailable (BioA) and total testosterone. Men and women (n=17/15) with obesity completed a 16-week dietary weight loss program with repeat testing.

Results:

BioA testosterone was lower in men with obesity and was related to Rd positively in men and negatively in women (P<0.05). Across subjects, weight loss increased Rd and SHBG (P<0.01). BioA testosterone was unchanged in men, yet individual changes were independently related to Rd (P=0.031). In women, BioA testosterone declined (P<0.009) but was not related to Rd.

Conclusions:

BioA testosterone was associated with Rd; positively in men and negatively in women. Weight loss reduced BioA testosterone in women, yet individual changes were associated with improved Rd in men. SHBG was a better correlate of improved Rd in women. Additional studies should identify mechanisms which drive sex differences and interventions which modify testosterone, reduce adiposity, and improve Rd across both sexes.

INTRODUCTION

Obesity affects over 250 million people worldwide and is strongly related to the development of type 2 diabetes (T2D) and insulin resistance (IR). Elevated adiposity may directly contribute to the development of these conditions through oversupply of fatty acids to peripheral tissues with specific fat depots (e.g., abdominal visceral adipose tissue, AVAT) conferring greater risk than others (e.g., abdominal subcutaneous AT, ASAT; gluteal-femoral AT) ¹. The relationship between testosterone, obesity, and IR is complex and bidirectional. Sex-specific differences in the distribution of AT during puberty¹ demonstrate the importance of sex hormones in AT development². Indeed, testosterone can directly impact fat metabolism through the stimulation of lipolysis³, although this is depot-dependent, and the process may be altered in dysfunctional adipocytes⁴.

Epidemiological studies have shown that higher testosterone levels in women are associated with increased T2D risk, obesity, and IR, while in men, higher testosterone is associated with reduced risk^5–10. Androgen deprivation in men with prostate cancer increased adiposity and induced IR¹¹ while testosterone treatment in hypogonadal men with obesity and T2D reversed these processes in some¹², but not all studies¹³. In women, an antiandrogen, flutamide, decreased subcutaneous AT and AVAT when added to a hypocaloric diet¹⁴. Alternatively, the anabolic/androgen steroid nandrolone increased AVAT and reduced subcutaneous AT in healthy postmenopausal women with obesity¹⁵.

Sex hormone binding globulin (SHBG) is central to the relationship between testosterone, obesity, and IR in both men and women^16,17. SHBG production by hepatocytes is decreased in individuals with IR partly through suppression of the transcription factor hepatocyte nuclear factor 4 (HNF4α by insulin and inflammatory cytokines which are elevated in people with obesity, hepatic steatosis, and metabolic syndrome¹⁶. SHBG is lower in men and women with T2D compared to subjects without T2D⁵, and this relationship appears to be independent of changes in other sex hormones¹⁸. However, men with obesity have lower total testosterone levels in part because SHBG is reduced¹⁹. Central obesity, low SHBG, and increased androgens are characteristics of women with polycystic ovarian syndrome (PCOS) and predicted incident obesity in perimenopausal women²⁰. These relationships persist in women with obesity without a concurrent PCOS diagnosis^6,21,22, potentially driven by the inhibitory effects of insulin on SHBG²³.

Many studies have examined the relationships between IR and testosterone and/or SHBG cross-sectionally in individuals with obesity^2,17, yet fewer studies have examined how lifestyle interventions impact these relationships in a sex-dependent manner, or tested how changes in adiposity and body fat distribution mediate these relationships. Of these studies, the interventions include hormone replacement therapy or other pharmaceuticals^2,24, lifestyle interventions^22,25–30, GLP-1 receptor analogs³¹, or surgical weight loss^25,30,32. The current study extends previously published analyses^33–35 and was designed to examine the sex-specific associations between testosterone, SHBG, and IR across a range of body weights (phase 1) and the impact of diet-induced weight loss on the same relationships in healthy adult women and men with obesity (phase 2). We hypothesized that in men, increased body weight, abdominal fat, and peripheral IR would be related to lower testosterone levels while in women the opposite would be true; greater body weight, abdominal fat, and IR would be related to greater testosterone levels. In phase 2, we hypothesized that weight loss would increase SHBG in both sexes while testosterone levels would increase in men but decrease in women, and these changes would be associated with reductions in AVAT and peripheral IR.

METHODS

Subjects and protocol.

The details of this study have been published previously^33,34. Healthy men and premenopausal women with normal/overweight (BMI <28.9kg/m²; M/F: n=7/8) or obesity (BMI ≥29kg/m²; M/F: n=20/20) were recruited through public advertisement near the University of Pittsburg campus. Exclusion criteria included diabetes, hyperlipidemia (plasma triacylglycerols ≥350mg/dL or total cholesterol ≥300mg/dL), coronary heart disease, vascular disease, or using anti-hypertensive or oral contraceptive medications. All women reported regular mensuration. Body composition, peripheral IR, plasma total testosterone, and SHBG were measured and non-SHBG testosterone (bioavailable [BioA] testosterone) was calculated (figure 1)³⁶. A subset of the subjects with obesity (M/F: n=15/17) completed a supervised, 16-week energy restriction weight loss program. Subjects provided written informed consent, and the protocol was approved by the University of Pittsburgh Institutional Review Board.

Figure 1.

Experimental design

(A) Following baseline anthropometric and cardiorespiratory fitness testing, subjects consumed a weight maintenance diet containing ≥200g CHO for three days before the impatient visit. Additionally, subjects refrained from strenuous activity/exercise for two days prior to admittance to the General Clinical Research Center (GCRC) at the University of Pittsburg. Subjects’ consumed a standardized dinner consisting of 10 kcal/kg body weight (50% energy [%E] from carbohydrates, 30%E fat, 20%E protein). The following morning a primed (2 μCi), continuous (0.20 μCi/min) infusion of labeled glucose was initiated. Insulin was infused at 40 mU/m²/min and 20% dextrose at variable infusion rates to maintain euglycemia which was confirmed every five minutes during the hyperinsulinemic-euglycemic clamp. (B) A subset of subjects continued onto phase two and consumed a very low-energy diet (800 kcal/day) for 10 weeks which was followed by a 4-week gradual reintroduction phase (to 1,200 kcal/day) and then two weeks of weight maintenance before the same series of test shown in 1A were repeated.

Body Composition.

Dual energy X-ray absorptiometry (DEXA) and computed tomography (CT) were completed as previously described³³. In brief, DEXA (Lunar model DPX-L, Madison, WI) was used to quantify whole body fat (FM) and fat free mass (FFM). Cross-sectional AVAT and ASAT areas were measured by CT (9800 CT scanner, GE, Milwaukee, WI).

Insulin Sensitivity.

The hyperinsulinemic-euglycemic clamp for measurements of IR was described before³³. In brief, subjects consumed a weight-maintaining diet with ≥200 grams carbohydrates/day for three days before the study and refrained from exercise for two days. Subjects consumed a standardized dinner (10 kcal/kg; 50% energy [%E] from carbohydrates; 30%E fat; 20%E protein) and the following morning a primed (20 μCi), continuous infusion (0.20 μCi/min) of [3-³H] glucose (New England Nuclear, Boston, MA) was initiated. Blood samples were collected from a second IV placed in a radial vein of the contralateral arm. Following 100 minutes of isotope infusion, the insulin infusion (40 mU/m²/min; Humulin; Eli Lilly, Indianapolis, IN) began and continued for three hours. Euglycemia was maintained with a variable infusion of 20% dextrose and plasma glucose concentrations were monitored every five minutes with an automated glucose oxidase reaction analyzer (YSI 2300 Glucose Analyzer; Yellow Springs, OH). Subjects remained fasted until the completion of the clamp and glucose disposal was measured during steady state. Indirect calorimetry (DeltaTrac, Anaheim, CA) was performed during basal and insulin-stimulated steady states for estimations of glucose oxidation (GluOx).

Energy restriction-induced weight loss.

Following the baseline measurements, subjects with obesity were invited to participate in a 16-week supervised weight loss program. The subjects consumed a very-low energy diet (VLED; 800kcal/day) for 10 weeks followed by a gradual increase to 1,200kcal/day until week 14, after which, a weight maintenance diet was implemented. Subjects completed the same series of tests described above in week 17. Diet composition was described previously³³. Subjects were provided mineral and vitamin supplementation and safety monitoring throughout the VLED.

Analytical Methods.

Total testosterone, SHBG, and dehydroepiandrosterone (DHEA)-sulfate were measured as previously described^33,35,37. Briefly, Coat-A Count solid phase radioimmunoassay (RIA) was used to quantify total testosterone (Diagnostic Products), and SHBG was measured with the active SHBG two-site immunoradiometric assay (Diagnostic System Laboratories). BioA testosterone was calculated with the Vermeulen equation from the levels of testosterone and SHBG³⁶. Plasma insulin was measured with a commercial RIA kit (Pharmacia, Uppsala, Sweden; Linco Research, St. Louis, MO). Liquid scintillation was used to measure glucose specific activity.

Calculations.

Glucose disposal (Rd) was calculated according to the equations of Steele³⁸. GluOx was calculated from the rates of gas exchange³⁹ and nonoxidative glucose disposal (NOGD) represents the difference between Rd and GluOx. A power analysis completed based on data from Ken-Dror et al²⁹ revealed 14 subjects/group were needed for 80% power to detect a significant difference in total testosterone following weight loss in men (19/group required for 90% power). While few studies have assessed similar outcomes in women, data from Leenen and colleagues²² suggested only nine subjects were needed to have 90% power to detect a decrease in total testosterone following weight loss.

Statistical Analyses.

Data are presented as mean±SE and P<0.05 was considered significant while P<0.10 was considered a trending relationship. Baseline data were compared with two-way ANOVA to determine the effects of sex, obesity, and the interaction. Mixed model ANOVAs were used to compare effects of weight loss between men and women with time and sex as fixed effects and subjects as random effects. Pearson regression analysis was applied to determine the relationships between sex hormones and Rd. To assess potential spurious correlations, the conservative Bonferroni-adjusted α-value of 0.0125 (within sex and phase correction) was considered. Statistical analyses were performed using R and R studio (lmer:lme package) and figures were prepared using GraphPad.

RESULTS

A portion of the results were reported previously^33,34 and the current results are in line with the earlier findings. This analysis expands upon those studies by investigating the relationship between sex, testosterone, SHBG, body weight, adiposity, and IR across individuals with and without obesity and following weight loss.

Phase 1

Healthy men and women (n=27/28; table 1) with and without obesity were included. Subjects with obesity tended to be older (~3.5y) and had greater BMI (31%), body weight (32%), FM (54%), FFM (20%), AVAT (52%), and ASAT (60%) than subjects without obesity. As reported previously³³, blood glucose was similar across subjects. Insulin concentrations were greater in subjects with obesity (weight effect: P=0.001), while glucose disposal (P<0.001), insulin-stimulated GluOx (P=0.003), and NOGD (P<0.001) were lower.

Table 1.

Phase 1: Characteristics of the subjects at baseline

	Women		Men		ANOVA
	N/OW n=8	OB n=20	N/OW n=7	OB n=20	P-value Weight x Sex
Age (y)	34±3	38±1	34±2	37±1	0.687╪
BMI (kg/m2)	23.4±1.0	33.8±0.7	23.2±0.4	34.0±0.9	0.855*
Body weight (kg)	65±3	92±3	71±3	108±3	0.191*,#
Fat mass (kg)	20±3	39±1	13±1	34±3	0.697*,#
Fat Mass (%)	30±3	43±1	18±1	31±1	0.914*,#
Fat free mass (kg)	41±1a	49±2b	55±2a	70±2b	0.069*,#
AVAT (cm2)	59±10	140±13	86±16	156±15	0.745*
ASAT (cm2)	225±30	518±22	154±22	452±32	0.944*,#
Glucose (mg/dL)	82±3	83±1	83±2	87±2	0.680
Insulin (mU/L)	7.3±1.4	17.5±2.6	6.0±0.9	14.9±1.9	0.825*
Rd (mg/FFM/min)	9.8±1.0	6.3±0.5	8.8±0.5	5.4±0.5	0.982*
GluOx (mg/FFM/min)	4.3±0.3	3.2±0.2	3.4±0.3	2.6±0.3	0.573*,#
NOGD (mg/FFM/min)	5.5±1.0	3.2±0.4	5.3±0.6	2.7±0.3	0.890*
DHEA-SO4 (μmol/L)	2.9±0.5	3.7±0.4	5.6±0.6	4.7±0.5	0.163#

Data are presented as mean±SE and were analyzed by two-way ANOVA using R. The P-values represent the interaction term between weight and sex (*P<0.05 weight effect; ^#P<0.05 sex effect; ^╪P<0.10 weight effect; ^§P<0.10 sex effect). Significant interactions were followed by Tukey-adjusted pairwise posthoc testing using least squares means (emmeans). Within women or men, different superscripted letters indicate significant differences (P<0.10) between subjects with normal/overweight (N/OW) or obesity (OB). BMI, body mass index; AVAT, abdominal visceral adipose tissue; ASAT, abdominal subcutaneous adipose tissue; Rd, glucose disposal; GluOx, glucose oxidation; NOGD, non-oxidative glucose disposal; BioA, bioavailable; SHBG, sex hormone binding globulin; DHEA-SO₄, dehydroepiandrosterone-sulfate.

Missing data: GluOx and Rd (men-NW n=6); NOGD (men-NW n=5)

Women tended to have lower DHEA-sulfate than men, with no relationship to body weight (table 1). In men, DHEA-sulfate was related to body weight (r²=0.248, P=0.007; not shown) and ASAT (r²=0.165, P=0.032) with trending relationships in AVAT (r²=0.135, P=0.054) and Rd (r²=0.110, P=0.085). Women with obesity had 9% higher total testosterone levels compared to the women with normal/overweight, although this difference was not significant (P=0.950; fig. 2A). Two women had total testosterone levels slightly greater than 1.7 nmol/L but denied hirsutism and had regular menstruation. By contrast, men with obesity had 41% lower total testosterone levels than men with normal/overweight (P<0.001). Four men had total testosterone levels <10.4 nmol/L but exhibited no other symptoms of hypogonadism. SHBG was significantly lower in women (43%) and men (33%) with obesity compared to subjects without obesity (sex effect: P<0.001, fig. 2B). Calculated BioA testosterone, which excludes testosterone bound to SHBG³⁵, was greater in women with obesity (42%) and lower in men with obesity (24%) compared to subjects with normal/overweight with a significant sex by weight interaction (P=0.022; fig. 2C). Post-hoc group comparisons revealed this difference was significant in men (P=0.017). Body weight and total testosterone were negatively correlated in men, while in men and women, greater body weight was associated with lower SHBG (fig. S1). Both AVAT and ASAT were negatively related to total testosterone in men (fig. S1F–G) and SHBG in women (fig. S1J–K). Only ASAT correlated with SHBG in men (fig. S1N–O). BioA testosterone was positively related to body weight, AVAT, and ASAT in women (fig. 2D–F), while in men, BioA testosterone did not correlate significantly with total body weight but was negatively related to AVAT and ASAT (fig. 2H–J).

Figure 2.

Phase 1: Total testosterone, SHBG, bioavailable testosterone, and the relationships with body weight, AVAT, ASAT, and glucose disposal at baseline. (A) Total testosterone, (B) sex-hormone binding globulin (SHBG), and (C) Bioavailable (BioA) testosterone are presented as mean±SE (normal/overweight [N/OW]: women n=8; men n=6; obesity [OB]: women n=20 (n-1 for BioA); men n=20). Relationships between BioA testosterone and (D, H) body weight, (E, I) AVAT, (F, J) ASAT, and (G, K) glucose disposal are shown for women and men, respectively. Data were analyzed by mixed model ANOVAs (A-C; ^¥P<0.05 weight effect; ^#P<0.05 sex effect) or Pearson correlation coefficient (D-K). Interaction terms with P<0.10 were followed by Tukey-corrected post-hoc analysis (within a group: *P<0.05, **P<0.01, ***P<0.001).

Trending relationships were observed between SHBG and Rd in women (fig. S1L) and between total testosterone and Rd in men (fig. S1H). SHBG was not related to Rd in men (fig. S1P). Conversely, BioA testosterone was negatively related to insulin sensitivity in women, while in men it was positively related to insulin sensitivity (fig. 2G&K). These relationships were attenuated when AVAT or ASAT were included as covariates suggesting abdominal adiposity may influence the relationship between BioA testosterone and insulin sensitivity across a range of body weights. Upon consideration of a Bonferroni-corrected P-value (0.0125), BioA testosterone and ASAT in men and women and Rd only in men no longer met the threshold of significance. Examination of the components of Rd (GluOx and NOGD) revealed that women exhibited a significant interaction between BioA testosterone and GluOx (r²=0.415, P<0.001, not shown), which was independent of AVAT (P=0.002), ASAT (P=0.003), and FM (P=0.004), and only a trending relationship with NOGD (r²=0.123, 0.073, not shown). In men, GluOx was not related to total testosterone (r²=0.018, P=0.517, not shown), but NOGD was strongly related (r²=0.273, P=0.007, not shown).

Phase 2

Men and women (n=15/17) with obesity underwent a VLED weight loss intervention (fig. 1B). Subjects were similar in age and pre-intervention BMI (table 2). As shown before³³, subjects significantly reduced body weight (mean of all subjects: −15±1%), FM (−30±2%), AVAT (−41±4%), ASAT (−33±4%) and FFM (−5±1%), while percent FFM was increased (+11±1%; P<0.001). Relative cardiorespiratory fitness was unchanged following the VLED³³ and was not related to changes in testosterone or SHBG (not shown). Blood glucose remained under 100mg/dL following weight loss (table 2). Plasma insulin decreased significantly (−48±10%), GluOx did not change, and both NOGD and Rd increased³³. Men (+40±15%, P=0.0001) had a two-fold greater increase in Rd than women (+20±6%, P=0.091).

Table 2.

Phase 2: Changes in subject characteristics with weight loss

	Women		Men		ANOVA
	Pre n=17	Post n=17	Pre n=15	Post n=15	P-value Time x Sex
Age (y)	39±1	40±1	38±2	38±2	0.258¥
BMI (kg/m2)	34.0±0.7	29.5±0.8	34.5±0.9	29.0±0.7	0.165¥
Body weight (kg)	92±3a	80±3b	109±4a	91±3b	0.020¥,#
Fat mass (kg)	39±1	30±2	35±3	22±2	0.121¥,#
Fat Mass (%)	42±1	37±1	31±2	24±1	0.166¥,#
AVAT (cm2)	147±14a	106±11b	167±19a	89±8b	0.036¥
ASAT (cm2)	520±22a	404±23b	465±32a	307±26b	0.058¥,#
Fat free mass (kg)	49±2a	47±2b	70±2a	66±2b	0.013¥,#
Glucose (mg/dL)	84±2	82±1	87±2	85±2	0.938
Insulin (mU/L)	18.2±3.0	9.8±1.3	15.5±2.4	7.4±0.8	0.992¥
Rd (mg/FFM/min)	6.2±0.5a	7.3±0.6b	5.5±0.6a	7.8±1.0b	0.041¥
GluOx (mg/FFM/min)	3.1±0.3	3.4±0.3	2.6±0.3	2.7±0.2	0.837
NOGD (mg/FFM/min)	3.1±0.5a	3.9±0.5a	2.9±0.4a	5.1±0.9b	0.077¥
DHEA-SO4 (μmol/L)	3.8±0.5	3.5±0.4	4.7±0.6	4.0±0.6	0.437¥

Data are presented as mean±SE and were analyzed by mixed model repeated measures ANOVA using R with significant interactions (P<0.10) followed by Tukey-adjusted pairwise posthoc testing using least squares means (emmeans). The P-values represent the interaction term between time and sex (^¥P<0.05 time effect; ^#P<0.05 sex effect; ^╪P<0.10 time effect; ^§P<0.10 sex effect). Within women or men, different superscripted letters indicate significant posthoc differences (P<0.10) between pre- and post-testing. BMI, body mass index; Rd, glucose disposal; GluOx, glucose oxidation; NOGD, non-oxidative glucose disposal; BioA, bioavailable; SHBG, sex hormone binding globulin; DHEA-SO₄, dehydroepiandrosterone-sulfate.

Missing data: AVAT and ASAT (women-post n=16; men-post n=14); glucose (women-post n=15; men-post n=10); insulin (women-post n=14; men-post n=10); Rd, GluOx, and NOGD (men-post n=14).

DHEA-sulfate was reduced (table 2; time effect) and SHBG increased following weight loss (+27±7%, time effect: P=0.003). When analyzed within a sex, neither total nor BioA testosterone were modified by the intervention in men (fig. 3A–C), while total (P=0.052) and BioA testosterone (P=0.009) declined in women. Following VLED, changes in SHBG (fig. S2), DHEA-sulfate (not shown), and BioA testosterone (fig. 3D) were not related to changes in body weight in women. By contrast, increased total testosterone in men, but not SHBG, DHEA-sulfate (not shown), or BioA testosterone, correlated to body weight loss (fig. S2E,M,H). Reduced AVAT was a strong predictor of BioA (fig. 3E–F&I–J) and total testosterone changes (fig. S2F–G) in men but not women. Absolute FFM change was not related to BioA testosterone in men (r²=0.133, P=0.181, not shown) or women (r²=0.010, P=0.715, not shown).

Figure 3.

Phase 2: Total testosterone, SHBG, bioavailable testosterone, and the relationships with body weight, AVAT, ASAT, and glucose disposal following weight loss. (A) Total testosterone, (B) sex-hormone binding globulin, and (C) bioavailable (BioA) testosterone at pre- and post-testing (women n=15; men n=17) are presented as mean±SE. Relationships between changes in BioA testosterone and (D, H) body weight, (E, I) AVAT, (F, J) ASAT, and (G, K) glucose disposal are shown for women and men, respectively. Data were analyzed by mixed model ANOVAs (A-C; ^¥P<0.05 time effect; ^#P<0.05 sex effect) or Pearson correlation coefficient (D-K). Interaction terms with P<0.10 were followed by Tukey-corrected post-hoc analysis (within a group: ^#P<0.05, *P<0.05, **P<0.01, ***P<0.001).

Rd and BioA testosterone were unrelated in women following weight loss (fig. 3G), yet increased Rd strongly predicted an increase in SHBG (fig. S2L) independent of AVAT (P=0.027) and ASAT (P=0.033) but not fasting insulin (P=0.063). Despite no significant increase in BioA testosterone in men following the VLED, individual changes were positively related to Rd (fig. 3K) independent of FM (P<0.001), AVAT (P=0.021), ASAT (P=0.038), SHBG (P=0.012), and fasting insulin (P=0.002).

DISCUSSION

This study demonstrates that the associations between testosterone, peripheral IR, and fat distribution across a range of body weights are sex-specific and determines how diet-induced weight loss impacts testosterone and its potential role in improving insulin sensitivity. In women, greater BioA testosterone was related to lower Rd. In men, greater BioA testosterone was associated with higher Rd. As shown previously^33,34, subjects who completed the weight loss program had substantial improvements in body composition and Rd. Following the VLED, BioA testosterone declined in women with no consistent change in men. This analysis revealed for the first time a significant, independent relationship between the change in total and BioA testosterone and improved peripheral IR in men. Changes in total and BioA testosterone were also associated with a reduction in AVAT in men. In women, increased SHBG was a stronger correlate of improved Rd, however this relationship was partly dependent upon insulin concentrations. The results from this study contribute to our understanding of the sexually-dimorphic effects of testosterone and SHBG and extend the knowledge into settings of VLED-induced weight loss.

Testosterone and metabolic health

Low testosterone occurs in 20–40% of men with obesity compared to 4–5% of the general population⁴⁰, and T2D further increases this risk^5–9. In men, increasing obesity suppresses SHBG and gonadotropins resulting in reduced testosterone secretion²⁵. Among women, hyperandrogenism is generally due to PCOS which affects at least 10% of the population worldwide. PCOS is characterized by increased obesity risk, especially visceral adiposity⁴¹, increased luteinizing hormone pulsatility, hyperinsulinemia, and reduced SHBG which synergize to stimulate ovarian androgen synthesis⁴². Thus, in men and women, dysregulated testosterone^5,7–9 and SHBG^16,17 concentrations are often associated with poor metabolic health, adiposity, and increased disease risk including T2D, obesity, and cardiovascular disease. However, the relationship between these factors is bidirectional and complex.

Our data tend to agree with previous publications; in women, BioA testosterone was positively related to body weight, AVAT, and ASAT. In men, BioA testosterone was negatively correlated to AVAT and ASAT and weakly with body weight. Mendelian randomization studies have confirmed the dimorphic relationships between BioA testosterone and obesity in population-based studies²¹. Importantly, BioA testosterone, which represents circulating testosterone not bound to SHBG, has been independently connected to obesity and metabolic dysfunction¹⁸. We found that BioA testosterone was significantly related to Rd in women and men. In line, Lutz et al⁶ reported higher insulin sensitivity, measured by an oral glucose tolerance test, was associated with lower and higher total testosterone in women and men, respectively. Similarly, higher serum testosterone was positively related to homeostatic model for IR (HOMA-IR; i.e., greater IR) in women and negatively in men with waist circumferences as a mediating factor¹⁰. Taken together, these data support an interaction between testosterone and IR. Indeed, Haffner and colleagues⁸ reported a positive correlation between total testosterone and whole-body insulin sensitivity, measured by hyperinsulinemic-euglycemic clamp, in healthy men. This relationship was driven by NOGD, and the current data show similar trends in men. Early work in male rats supports glycogen synthesis as a key dysregulated factor in hypoandrogenism-induced IR⁴³. Thus, increasing testosterone in males may directly improve the ability of skeletal muscle to consume glucose for glycogen synthesis. While a relationship between markers of skeletal muscle mitochondrial function and low testosterone⁹ would support this hypothesis, VO₂max and BioA testosterone were not related. Importantly, the relationship between BioA testosterone and IR did not meet the significance cutoff in men after corrections for multiple correlations. This relationship should be interpreted with caution and may reflect an early connection between BioA testosterone and Rd in men with obesity but not without hypogonadism.

In women, GluOx appeared to be the primary driver of the relationship between BioA testosterone and Rd. This is in opposition to evidence in oophorectomized rats in which testosterone administration induced skeletal muscle IR mediated by the insulin receptor glycogen synthase system⁴⁴. Based on these data, we expected NOGD to be a stronger correlate in women. These sexually-dimorphic relationships may indicate differing effects of testosterone on skeletal muscle nutrient partitioning. The prior study⁴⁴ was conducted in rats with low estrogen and treated with high doses of testosterone, suggesting a plateau effect. Indeed, the sex-specific differences in body composition and IR observed here may be due to vast differences in testosterone concentrations; men are maximally androgenized while women have low androgenic activity⁷. Finally, much of the available literature reports on total testosterone levels which complicates the ability to identify the SHBG-independent effects of testosterone on IR since IR leads to a decrease in SHBG^19,45. The current data support BioA testosterone, over total testosterone as a stronger correlate of Rd in women and men.

Effects of weight loss on testosterone and metabolic health: Impact of sex

This study’s intervention phase sought to establish a deeper understanding of the biology of sex differences in testosterone and IR. Many studies have investigated the effects of weight loss on testosterone concentrations in men^{22,26–29,32}. In women without PCOS, fewer studies are available^22,32, and no studies have compared the sex-specific changes in testosterone to improvements in insulin sensitivity and body composition following weight loss. In men, many studies^{26–28,30,32,46}, but not all²², have shown that diet-induced weight loss or bariatric surgery significantly increased total testosterone concentrations. In men with T2D in the Look AHEAD trial, one year of weight loss elicited a 14% increase in total testosterone and an 18% increase in SHBG⁴⁶. Similarly, a meta-analysis²⁹ reported total testosterone increased 2.5nmol/L in men who underwent low-energy diets whereas bariatric surgery increased total testosterone by 7.2nmol/L. The increment of total testosterone following weight loss was related to higher baseline BMI²⁹ and the degree of weight reduction²⁵. Despite a greater weight loss in the current study (15.8% vs. 8.6% via low-energy diet²⁹), the men achieved only a 0.5 nmol/L increase in total testosterone (table 2). In fact, as noted in the meta-analysis²⁹, both total and BioA testosterone levels declined after dieting in some men. It is possible that the level of energy restriction impacted testosterone production through altered luteinizing hormone levels and pulsatility⁴⁷. However, examination of the studies analyzed in Ken-Dror et al. ²⁹ revealed that as the level of energy restriction (1–12 month intervention) increased, (range: 400–1700kcal/d) total testosterone rose, in line with the established relationship between degree of weight loss and increasing testosterone in men²⁵. Considering those data, we find it unlikely that 800kcal/day for 10 weeks caused testosterone levels to decrease due to low energy availability which has been shown in previous fasting and time restricted eating trials^47,48. Instead, men in previous studies were generally older with greater obesity, and many had diabetes, and only four of the men in our weight loss study had total testosterone levels at baseline that were below the reference range. We believe that differences in the study population is the most likely cause for different testosterone outcomes with weight loss. Nevertheless, the variable changes among our subjects remains surprising and more research is clearly needed to understand the mechanisms which underpin changing hormones following weight loss.

Despite minimal changes in mean testosterone levels, an independent positive correlation persisted between peripheral Rd and BioA testosterone following weight loss in men suggesting that small changes in BioA testosterone may exert beneficial effects on glucose flux independent of changes in body composition and SHBG. Some^27,28,49, but not all^26,28, weight loss studies have found similar relationships between glucose tolerance and testosterone, however these studies only included HOMA-IR or blood glucose. For example, HOMA-IR was reduced and testosterone increased following 12-weeks of energy restriction by either a low- or high-carbohydrate diet²⁸. Another study implementing a VLED found HOMA-IR was the strongest correlate of increased testosterone levels²⁷. Finally, bariatric surgery induced robust testosterone increases in men which was strongly related to improved HOMA-IR⁴⁹. Applying gold standard measurements of body composition and IR, our study extends previous findings to show that improvements in AVAT and glucose disposal strongly predict the rise in total and BioA testosterone. While cytokines were not measured in this study, cytokines linked to AVAT are one proposed mechanism for testosterone deficiency in men with IR⁵⁰. Furthermore, we cannot rule out other factors (e.g., AT function, aromatase activity) that may mediate this relationship. SHBG is also strongly related to and has been implicated as a contributing factor to the development of obesity and IR¹⁶. Despite a rise in circulating SHBG, this change was not predictive of improved Rd, and the relationships between BioA testosterone and IR remained independent of SHBG. Although our findings align with previous reports, the relationship between improved glucose disposal and BioA testosterone did not meet significance when analyzed with a Bonferroni corrected P-value. Thus, this relationship should be confirmed in a larger sample size of men and women across the spectrum of androgen dysregulation.

The relationship between IR and testosterone following weight loss in women without PCOS is not well established. Some studies found weight loss reduced total and free testosterone^22,30 and increased SHBG²², yet other studies in women with T2D found minimal changes in SHBG⁵¹, potentially due to estrogen use⁵². In contrast to data from men, fasting reduced testosterone and increased SHBG in women even with moderate to no weight loss⁴⁸, potentially due to changes in luteinizing hormone levels and pulse frequency or circulating insulin levels. In the study by Escobar-Morreale et al. ³², testosterone was reduced and SHBG increased following bariatric surgery although most women in that study had PCOS. In line with data from the Diabetes Prevention Program in premenopausal women with T2D and not using estrogen⁵², we found SHBG increased with weight loss. In our study, the rise in SHBG was also a better correlate of Rd improvements than was BioA testosterone. Reductions in circulating insulin, hepatic steatosis and improved Rd that accompanied weight loss may explain the increase in SHBG and the decrease in testosterone^45,53. More research is needed to understand these complicated relationships in women with obesity with and without PCOS who are undergoing weight loss or lifestyle interventions. DHEA-sulfate, which has previously been associated with insulin resistance in young, healthy women⁵⁴, was not related to improved body weight, composition, or insulin sensitivity in men or women in this study. Yet, changes in other androgens, SHBG, and other sex hormones may be important in response variability and long-term success to lifestyle interventions⁵⁵. The effects of various dietary interventions, including fasting, impact the control of testosterone production differently in men and women highlighting the need to examine both sexes in studies of testosterone and weight loss.

Limitations

The conclusions of this study reflect detailed findings in a small population of subjects which can limit generalizability. Further, when applying a Bonferroni correction within a sex for each phase (α=0.0125), the relationships between BioA testosterone and Rd in men are attenuated to nonsignificance, potentially due to a limited sample size, and should be interpreted with caution. Second, testosterone was measured using RIA which is less accurate for the low levels observed in women. Liquid chromatography-mass spectrometry is the current gold-standard method for androgen quantification at low levels. Third, other hormones, which play important roles in metabolic health, were not measured. Specifically, measurement of luteinizing hormone, estradiol, and cortisol should be included in future studies considering VLED⁵⁶, as well as fasting⁴⁸, may result in suppression of testosterone levels due to changes in these hormones. The repeated measures design was a strength; however, the study lacked a control group to evaluate how no intervention might impact the results. Finally, an exercise program was not included; thus, we cannot know what the impact of adjunct exercise might have on these associations. Nonetheless, this study was strengthened by including both cross-sectional and interventional phases.

Conclusion

Circulating testosterone levels correlate with IR in a sex-specific manner and these relationships are differentially mediated by body composition and adipose depots. Higher BioA testosterone among women was inversely related to Rd, while in men, IR is associated with low testosterone. Following VLED-induced weight loss, BioA testosterone was unchanged across men and was reduced in women, despite robust reductions in body weight, AVAT, and ASAT. SHBG was increased across all subjects, underscoring sex difference in the metabolic control of testosterone production. Changes in BioA testosterone and IR were independently related in men while improved IR was related to increased SHBG in women. These data corroborate cross-sectional studies and expand our knowledge about how testosterone and metabolic health are impacted by weight loss. Future work examining sex-specific effects of weight loss on IR should consider the role of sex hormones. Additional studies should investigate the role of sex hormones in the loss of skeletal muscle mass that occurs with weight loss. Finally, these results support the need to identify optimal interventions to alter testosterone, reduce adiposity, and improve IR in women and men.

STUDY IMPORTANCE.

1). What is already known about this subject?

1. There is a sexual dimorphism in testosterone concentrations in men and women with obesity. Men with obesity often exhibit reduced testosterone concentrations while women with obesity have higher levels.

2. Dysregulated testosterone and sex hormone binding globulin (SHBG) concentrations are often associated with poor metabolic function, adiposity, and increased disease risk including type 2 diabetes, obesity, and cardiovascular.

2). What are the new findings in your manuscript?

1. This study demonstrated the mediating effect of abdominal visceral and subcutaneous adipose tissue on the relationship between bioavailable testosterone and insulin resistance in men and women with and without obesity.

2. Following diet-induced weight loss over 16 weeks, bioavailable testosterone concentrations were not significantly changed across subjects, while SHBG significantly increased.

3. Despite minimal average changes in bioavailable testosterone, individual changes were related to enhanced peripheral insulin sensitivity in men only. This correlation was independent of changes in body composition, insulin, and sex-hormone binding globulin. In women, SHBG was a stronger correlate of improved insulin sensitivity.

3). How might your results change the direction of research or the focus of clinical practice?

1. Future research should investigate how sex differences in testosterone differentially mediate metabolic improvements following weight loss. For example, the localized effects of testosterone on skeletal muscle insulin sensitivity may be a key next step in understanding sex differences of testosterone-mediated effects on insulin resistance (e.g., nutrient partitioning to glucose oxidation or glycogen storage).

2. This work supports the need to better understand the duration in which changes in circulating testosterone and SHBG concentrations occur, or the level of weight loss needed.