Ovarian Hyperandrogenism and Response to Gonadotropin-releasing Hormone Analogues in Primary Severe Insulin Resistance

Abstract Context Insulin resistance (IR) is associated with polycystic ovaries and hyperandrogenism, but underpinning mechanisms are poorly understood and therapeutic options are limited. Objective To characterize hyperandrogenemia and ovarian pathology in primary severe IR (SIR), using IR of defined molecular etiology to interrogate disease mechanism. To extend evaluation of gonadotropin-releasing hormone (GnRH) analogue therapy in SIR. Methods Retrospective case note review in 2 SIR national referral centers. Female patients with SIR with documented serum total testosterone (TT) concentration. Results Among 185 patients with lipodystrophy, 65 with primary insulin signaling disorders, and 29 with idiopathic SIR, serum TT ranged from undetectable to 1562 ng/dL (54.2 nmol/L; median 40.3 ng/dL [1.40 nmol/L]; n = 279) and free testosterone (FT) from undetectable to 18.0 ng/dL (0.625 nmol/L; median 0.705 ng/dL [0.0244 nmol/L]; n = 233). Higher TT but not FT in the insulin signaling subgroup was attributable to higher serum sex hormone–binding globulin (SHBG) concentration. Insulin correlated positively with SHBG in the insulin signaling subgroup, but negatively in lipodystrophy. In 8/9 patients with available ovarian tissue, histology was consistent with polycystic ovary syndrome (PCOS). In 6/6 patients treated with GnRH analogue therapy, gonadotropin suppression improved hyperandrogenic symptoms and reduced serum TT irrespective of SIR etiology. Conclusion SIR causes severe hyperandrogenemia and PCOS-like ovarian changes whether due to proximal insulin signaling or adipose development defects. A distinct relationship between IR and FT between the groups is mediated by SHBG. GnRH analogues are beneficial in a range of SIR subphenotypes.

Polycystic ovary syndrome (PCOS) affects 5% to 20% of women and is a major cause of morbidity and subfertility. As captured in various diagnostic criteria, PCOS is characterized by clinical or biochemical evidence of androgen excess, chronic ovulatory dysfunction (manifesting as menstrual irregularity), and polycystic ovarian morphology, typically demonstrated sonographically by an increased number of peripheral preantral-like follicles (1).
The association of PCOS with diabetes was first recognized a century ago (2), and in recent decades the association between insulin resistance (IR) and hyperandrogenism has been widely reported (3). From the 1970s onwards, severe hyperandrogenism was observed in particularly extreme forms of IR, including those due to genetic insulin receptor deficiency and anti-insulin receptor autoantibodies (4)(5)(6), and those associated with lipodystrophy (7)(8)(9). Before onset of diabetes, IR is associated with compensatory hyperinsulinemia, which can be extreme. The effects of excessive insulin action on the ovary, which appear to be independent of the insulin receptor (INSR), are suggested to drive ovarian hyperandrogenism in common PCOS (10).
A systematic assessment of ovarian pathology in severe insulin resistance (SIR) has not been conducted, although scattered reports have described massive, bilateral ovarian enlargement (11,12) and ovarian neoplasia (13) associated with severe genetic insulin receptor dysfunction (Donohue syndrome). We now describe the spectrum of hyperandrogenemia and ovarian pathology in a large cohort of patients with SIR, including many with defined monogenic or autoimmune etiologies. We show that primary IR, whether caused by proximal defects in insulin signaling or lipodystrophy, is sufficient to cause ovarian hyperandrogenism with ovarian morphology indistinguishable from common PCOS. IR and sex hormone-binding globulin (SHBG) concentration show a distinct positive relationship in the insulin signaling subgroup, consistent with a direct role for liver insulin signaling in suppressing hepatic SHBG production in other forms of IR. Finally, we evaluate the efficacy of GnRH analogue therapy in lowering testosterone levels and improving androgenic symptoms in women with SIR, extending prior evidence that pulsatile gonadotropins play a permissive role in IR-associated ovarian hyperandrogenism.

Patients
We performed a retrospective review of case records from 2 national referral centers for SIR at the National Institutes of Health, USA (since 1977), and University of Cambridge, UK (since 1992). All patients were referred based on clinical and/or biochemical evidence of SIR and/or lipodystrophy and were studied using procedures approved by the institutional review board of the National Institute of Diabetes and Digestive and Kidney Diseases or by the UK Research Ethics Committee. Each participant or their legal guardian provided written, informed consent, and minors provided verbal or written consent in accordance with local regulations. Studies were conducted in accordance with the principles of the Declaration of Helsinki.

Inclusion Criteria
Patients (all female) met 1 or more of the following criteria: (1) body mass index (BMI) <30 kg/m 2 with normal glucose tolerance and fasting hyperinsulinemia >150 pmol/L, and/or peak insulin >1 500 pmol/L after a 75 g oral glucose challenge; (2) BMI <30kg/m 2 and diabetes with insulin requirements >3 units/kg/day; (3) a disorder known to cause SIR (eg, lipodystrophy). All patients were evaluated by an endocrinologist and appropriate endocrine testing was performed to exclude secondary causes of IR and/ or lipodystrophy where clinical suspicion was raised. Dysthyroidism and hyperprolactinemia were excluded prior to referral. Similarly, where indicated, adrenal sources of hyperandrogenemia were excluded by the reviewing endocrinologist. Patients of all ages were included in this study; however, given the focus on the interaction between hyperinsulinemia and the hypothalamic-pituitary-gonadal (HPG) axis, some analyses were restricted to life stages during which the HPG axis is likely to be active. Pubertal status was assessed formally in a subset of patients (Tanner breast stage II-IV categorized as "midpubertal"; Tanner breast stage V or postmenarche categorized as "postpubertal"). If clinical assessment of pubertal status was not documented, we pragmatically used a threshold of 10 years and above to define the group with likely HPG axis activation. In a further subset of patients, menopausal status was determined from chart review (ie, menopause documented in chart, age >60 years, or luteinizing hormone (LH)/follicle-stimulating hormone (FSH) in the menopausal range).

Biochemical Assays
All patients underwent biochemical evaluation as part of clinical care. Analyses, including serum total testosterone (TT), sex hormone-binding globulin (SHBG) and fasting insulin, were performed in accredited clinical diagnostic laboratories of a range of referring hospitals, using various platforms over a 40-year period. The majority of TT levels were determined by immunoassay, with a subset of around 10% to 20% measured by mass spectrometry. For all individuals, results reported as below or above the assay range were treated as being at the lower or upper limit of the assay range, respectively, for the purposes of graphical representation and statistical analysis. In patients with a documented SHBG level, free testosterone (FT) was estimated using the Vermeulen method, assuming an albumin concentration of 45 g/L. (14) Insulin Resistance Subphenotyping Causal genetic mutations, where reported, were identified by Sanger sequencing or exome/genome sequencing as part of clinical care or programs of research into the genetic basis of SIR. Type B IR was diagnosed by the presence of anti-insulin receptor autoantibody detected by immunoprecipitation, as previously described (15). Patients were classified into 3 groups based on clinical, biochemical and genetic evaluation: (1) lipodystrophy, either generalized or partial (including congenital and acquired syndromes); (2) primary disorders of insulin signaling, due to either (a) loss-of-function mutations in INSR (encoding the insulin receptor), PIK3R1 (encoding the p85α regulatory subunit of PI3-kinase), AKT2 (encoding protein kinase B), or TBC1D4 (encoding AS160), or (b) type B IR; or (3) SIR without lipodystrophy of unknown but suspected genetic etiology ("idiopathic SIR").

Ovarian Pathology
Ovarian tissue samples were obtained as part of clinical care, processed in accredited histopathology laboratories and reported by local histopathologists. Available hematoxylin and eosin-stained sections were further reviewed by an independent gynecological histopathologist blinded to clinical presentation and diagnosis but not to the age of the patient.

Statistics
Summary statistics presented are median and interquartile range (IQR), unless indicated otherwise. TT, FT, fasting insulin or SHBG levels were compared among multiple groups using a 2-tailed, nonparametric Kruskal-Wallis test followed by post hoc pairwise Wilcoxon rank sum test with Bonferroni correction. Association between fasting insulin and testosterone or SHBG was assessed using Spearman's test. Statistical significance was declared at P < .05. All analyses were performed in R version 3.5.2 (2018-12-20).

Prevalence and Severity of Hyperandrogenemia in Primary Severe Insulin Resistance
A total of 279 patients (aged between 4 months and 70 years) with confirmed or likely primary SIR had a serum TT level available for review. Of these, 185 (66%) had partial or generalized lipodystrophy (114 and 71 patients, respectively) and 65 (23%) had a primary disorder of insulin signaling, affecting the insulin receptor or components of the downstream insulin signaling cascade ( Fig. 1A and 1B). These included 9 patients with severe biallelic variants causing Donohue or Rabson-Mendenhall syndrome, while the remaining 24 had less extreme IR manifesting peripubertally, caused by monoallelic INSR pathogenic variants (historically known as "type A" IR). Four patients had IR due to pathogenic variants in PIK3R1, encoding the p85α regulatory subunit of PI3-kinase (reported in (16)), 1 had a pathogenic variant in downstream protein kinase B (AKT2, reported in (17)), and 1 had a pathogenic variant in TBC1D4, encoding AKT substrate of 160 kDa (AS160), which is implicated in insulin-stimulated GLUT4 translocation (18,19). A further 29 patients (10%) had SIR without lipodystrophy, insulin receptor autoantibodies, or a confirmed genetic diagnosis, but suspected to be of monogenic etiology ("idiopathic SIR"). Of 152 patients in the cohort whose medication records were readily available, the proportions taking metformin, insulin, or other oral hypoglycemic agents were 50%, 48% and 20%, respectively. Only 8% were on a hormonal contraceptive preparation, and 2% were taking spironolactone.
Serum TT concentrations across the study cohort, measured using different assays in multiple laboratories, ranged from undetectably low to 1562 ng/dL (54.2 nmol/L) (Fig. 1C). The median testosterone level was 40.3 ng/dL (1.4 nmol/L; IQR 20-94.5 ng/dL or 0.694-3.279 nmol/L). One-third of patients (94/279, 34%) had a testosterone level above 70 ng/dL (2.4 nmol/L), which approximates the upper limit of normal in adult females in most . The p85 α regulatory subunit of PI3-kinase interacts with phosphorylated IRS proteins, activating and approximating the p110 catalytic subunit to the plasma membrane, resulting in appearance of the phospholipid second messenger phosphatidylinositol (3-5)-trisphosphate (PIP3). This in turn acts as a docking site for a range of effector proteins including Protein Kinase B (AKT), activated by PIP3-dependent phosphorylation by phosphoinositidedependent kinase 1 (PDK1). One target of AKT2 in adipose tissue and skeletal muscle is the AKT substrate of 160 kDa (AS160, encoded by the TBC1D4 gene), phosphorylation of which leads to translocation of GLUT4 glucose transporters to the plasma membrane. Loss-of-function mutations in INSR, PIK3R1 (encoding the p85ɑ subunit of PI3K), AKT2 and TBC1D4 (encoding AS160) have all been associated with monogenic insulin resistance. (B) Composition of severe insulin resistance subgroups. The lipodystrophy group included patients with generalized or partial lipodystrophy (genetic or acquired). The insulin signaling defect group included patients with pathogenic variants in the insulin receptor (Donohue syndrome (DS), Rabson-Mendenhall syndrome (RMS) as well as less severe "type A" insulin resistance), pathogenic variants in genes encoding downstream insulin signaling components (p85 α, AKT2, and AS160) and insulin receptor dysfunction due to acquired autoantibodies ("type B" insulin resistance). The idiopathic severe insulin resistance group consisted of a highly selected group of patients with severe insulin resistance of presumed monogenic etiology but no known genetic defects and no insulin receptor autoantibodies. (C) Total serum testosterone levels in female patients aged 4 months to 70 years with a range of genetic and acquired syndromes of insulin resistance, measured in various referring hospitals over a 40 year period (n = 279). Total testosterone of 70 ng/dL, approximating the upper limit of normal in most clinical immunoassays, is indicated for reference. (D) Calculated free testosterone levels in 233 patients who had a documented sex hormone-binding globulin (SHBG) concentration. Upper limit of normal (2.4 ng/dL) indicated for reference. clinical immunoassays. Thirty-six of 279 (13%) had a serum testosterone in excess of 150 ng/dL (~5 nmol/L), approximately twice the upper limit of normal. Among 233 patients with a documented SHBG level, calculated serum FT concentration ranged from undetectably low to 18.0 ng/dL (0.625 nmol/L) (Fig. 1D). The median FT level was 0.705 ng/dL (0.0244 nmol/L; IQR 0.340-1.781 ng/dL or 0.0118-0.0618 nmol/L). Forty of 233 (17%) had a calculated FT level above the upper limit of normal (2.4 ng/dL or 0.083 nmol/L) (Fig. 1D).

Severity of Hyperandrogenemia in Different Insulin Resistance Subphenotypes
Seventeen patients were under 10 years old and therefore deemed unlikely to have a pubertal HPG axis. TT and FT, SHBG, and fasting insulin were evaluated in the remaining 262 patients aged 10 or above, grouped by clinical subphenotype ( Fig. 2A and Table 1). Within this cohort, median LH and FSH concentrations were 5.0 U/L (IQR 2.5-9.7) and 5.0 U/L (IQR 3.3-9.7), respectively. Median LH/FSH ratio was 0.84 (IQR 0.54-1.3) and did not differ significantly between patients with lipodystrophy and those with defects in proximal insulin signaling (data not shown). Among 173 individuals with lipodystrophy, TT ranged from below the assay limit of detectability to 1239 ng/dL (43.0 nmol/L) (median 24.0 ng/dL, IQR 20.0-59.0 ng/dL) while FT ranged from below detectability to 7.93 ng/dL (0.275 nmol/L) (median 0.619 ng/dL, IQR 0.338-1.35 ng/ dL). TT and FT levels in generalized lipodystrophy (whether genetic, acquired and autoimmune, or of unknown cause) were no different from those in partial lipodystrophy (genetic, acquired, or idiopathic). Moreover no significant differences were observed in TT or FT concentrations among genetic subgroups in generalized or partial lipodystrophy. Among 61 patients with defects in insulin signaling, TT ranged from below detectability to 1561 ng/ dL (54.1 nmol/L) (median 95.1 ng/dL, IQR 43.0-239 ng/ dL) and FT ranged from below detectability to 18.0 ng/dL (0.613 nmol/L) (median 1.03 ng/dL, IQR 0.37-2.57 ng/dL). No significant differences were observed in TT or FT levels among patients with pathogenic variants in insulin receptor or proximal insulin signaling components, or type B IR.
TT was significantly higher in the insulin signaling group and idiopathic SIR group than the lipodystrophy group (Kruskal-Wallis test followed by pairwise Wilcoxon rank sum test with Bonferroni correction, P < .005) (Fig. 2B). Patients with insulin signaling disorders did not have significantly different TT levels from those with idiopathic SIR (Fig. 2B). In contrast, patients in the idiopathic SIR group had significantly higher FT than both the lipodystrophy and insulin signaling groups (Fig. 2C), while there was no significant difference in FT between the lipodystrophy and insulin signaling groups.

Association Between Hyperinsulinemia, Sex Hormone-binding Globulin and Hyperandrogenemia in Primary Severe Insulin Resistance
To assess whether differences in TT and FT concentrations between clinical subgroups are related to degree of hyperinsulinemia and SHBG levels, we compared fasting insulin and SHBG, where available, in patients aged >10 years with lipodystrophy, insulin signaling defects or idiopathic SIR. Fasting insulin concentration ranged from 21 pmol/L to over 49 000 pmol/L (n = 245, median 335 pmol/L, IQR 166-875 pmol/L; Table 1), and was above the upper limit of normal (60 pmol/L) in 97% of all patients. In those on insulin analogue therapy, or with impaired beta-call function, fasting insulin concentration is likely to have underestimated the severity of IR, while metformin use will have reduced resistance. Fasting insulin was significantly higher in patients with insulin signaling defects than in those with lipodystrophy or idiopathic SIR (Fig. 2D). A comparable observation was made using homeostatic model assessment of IR (HOMA-IR) in lieu of fasting insulin, although this model is not validated in monogenic IR (data not shown).

Hypothalamic-Pituitary-Gonadal Axis Activity and Hyperandrogenemia in Severe Insulin Resistance
To determine the extent to which androgen excess in primary SIR is affected by HPG axis activity, we analyzed 194 patients of all ages whose pubertal/menopausal status was documented in clinical records. They included 65 patients with generalized lipodystrophy, 87 with partial lipodystrophy and 42 with defective insulin signaling, of whom 16 had a genetic INSR pathogenic variant, 25 had type B IR and 1 had a TBC1D4 defect. Individuals were classified as prepubertal (n = 13), midpubertal (n = 37), postpubertal (n = 112), and postmenopausal (n = 32) ( Table 2 and, Fig. 3A). None of the 13 prepubertal individuals (ie, lacking clinical evidence of HPG activation) had an elevated TT concentration (Fig. 3B), and as a group they had significantly lower TT levels compared to all other groups (Kruskal-Wallis test followed by pairwise Wilcoxon rank sum test with Bonferroni correction, P < .005) (Fig. 3B). No differences were observed between midpubertal or postpubertal patients, nor was lower TT observed after menopause (Fig. 3B). Within the midpubertal and post-pubertal subgroups, patients with insulin signaling disorders had significantly higher TT levels than patients with lipodystrophy (Wilcoxon rank sum test, P < .05), as observed in Fig. 2D, but no statistically significant difference was observed in the postmenopausal group (Fig. 3B).

Ovarian Pathology Associated With Severe Insulin Resistance
To assess ovarian morphology associated with hyperandrogenemia in primary SIR, we reviewed the case notes of 9 patients who underwent either oophorectomy or ovarian wedge resection as part of clinical care, and for whom ovarian tissue or histological reports were available. Clinical, biochemical, and histological data are presented in Table 3. Representative histological images, where available, are provided in Fig. 4. Of the 9 patients, 4 had a proven INSR variant, 4 had partial lipodystrophy (3 associated with known genetic variants), and 1 had a heterozygous pathogenic

Utility of Gonadotropin Receptor Agonists in Disparate Etiologies of Severe Insulin Resistance
We previously reported that long-acting GnRH analogue therapy can lower testosterone levels and ameliorate hyperandrogenism in patients with primary SIR resulting from insulin receptor autoantibodies (P13; (30)). We now describe 6 patients (including P13) with SIR (2 with type B IR, 2 with pathogenic INSR variants, 1 with a pathogenic variant in TBC1D4, and 1 with acquired partial lipodystrophy associated with juvenile dermatomyositis) who received GnRH analogue therapy for management of hyperandrogenism (Table 4). One patient (P4) also underwent oophorectomy ( Table 3). Four of the patients were postpubertal, 1 had primary amenorrhea and 1 was postmenopausal. All 6 patients experienced amelioration in hyperandrogenic symptoms and a reduction in TT to normal or near-normal levels in the absence of any change in their insulin sensitivity or glycemia control. In 5 patients, oophorectomy was avoided; 4 of these women were premenopausal with a desire for fertility preservation (P10-13), and 1 was post-menopausal but at high operative risk due to comorbidities (P15). One patient (P4), having previously undergone unilateral oophorectomy, underwent completion oophorectomy after GnRH analogue therapy further improved her symptoms.

Discussion
There is compelling evidence that IR, with consequent hyperinsulinemia, drives hyperandrogenism and ovarian dysfunction. Adult-onset insulin receptor blockade by autoantibodies (type B IR) leads to acute, often severe hyperandrogenism and ovarian enlargement which reverses when antibodies, and hence IR, are cleared or remit (31,32). Severe hyperandrogenism in several types of congenital SIR has also been described (16,22,33,34). Here, we offer comprehensive assessment of androgen excess in primary SIR and confirm that hyperandrogenemia is common in female patients with SIR (34% of those studied had TT levels above the adult female reference range) and that it can be extreme, sometimes exceeding the adult male range.  Hyperandrogenemia occurs irrespective of IR etiology, being seen in insulin signaling disorders and lipodystrophy alike, and ovarian histopathology is indistinguishable from PCOS in each of these groups. Of biochemical note, higher TT concentrations in the insulin signaling subgroup did not correspond to higher FT concentrations, in keeping with a positive correlation of SHBG with an index of IR in this group, quite distinct from the inverse relationship between IR and SHBG seen in lipodystrophy or common forms of IR. This agrees with our previous finding that preserved or increased SHBG concentration in SIR is a discriminating marker of insulin receptor dysfunction (35). Although detailed evaluation of clinical hyperandrogenism in our cohort was beyond the scope of this study, our findings support a causal relationship between IR/hyperinsulinemia and PCOS, with ovarian hyperandrogenism resulting from excessive insulin action on the ovary. We did not find evidence for elevated circulating LH/FSH ratio among pubertal and postpubertal individuals, despite this being a recognized feature of PCOS. It is noteworthy that not all subjects with IR syndromes had elevated testosterone, and the correlation between serum insulin and testosterone was not strong (r s = 0.211), suggesting that ovarian sensitivity to insulin may vary considerably among individuals.
"Insulin resistance" is usually defined in terms of attenuated ability of insulin to stimulate glucose uptake (36,37). Until beta cell failure occurs, this is compensated for by hyperinsulinemia, with plasma insulin concentration raised by 1 or 2 orders of magnitude in severe cases. "Insulin resistance" does not apply equally to all of insulin's actions, however, as evidenced by differences between simple insulin deficiency and IR. Some consequences of IR, including acanthosis nigricans and ovarian enlargement, require hyperinsulinemia, and presumably reflect increased insulin action (38)(39)(40). The ovary may thus be viewed as an "innocent bystander" in IR, suffering adverse trophic actions of the high insulin concentration required to maintain euglycemia. This is supported by hyperandrogenism and PCOS described in patients with insulinoma (41,42), and in hyperinsulinemic patients with glycogen storage disease (43), and by increased prevalence of hyperandrogenism and PCOS-like ovarian appearances in patients with type 1 diabetes, attributed to supraphysiological exogenous insulin in peripheral circulation (44,45).
Several mechanisms have been invoked to explain the mixed profile of insulin resistant and insulin sensitive phenomena seen in "insulin resistance." One possibility is that different arms of the insulin signaling pathway are differentially affected by IR. For example, it has been suggested based on studies of muscle insulin signaling in PCOS that "metabolic" insulin signaling through the PI3K pathway is selectively impaired, leaving "mitogenic" signaling intact (46). Another model holds that high concentrations of insulin activate mitogenic insulin-like growth factor 1 (IGF1) receptors, which may be compounded by alterations in IGF binding proteins in IR (47).
If a selective defect in metabolic actions of insulin were critical to pathogenesis of IR-associated ovarian For 194 patients, hypothalamic-pituitary-gonadal axis (HPG) status was documented in, or inferred from, their clinical charts. Individuals were categorized as prepuberty, midpuberty, postpuberty or postmenopause. Each group was further subdivided into 3 clinical subgroups (generalized LD, partial LD, insulin signaling defect). (B) Total serum testosterone by HPG axis activity and clinical subgroup. Testosterone levels in patients with different HPG status were compared using the Kruskal-Wallis test followed by pairwise Wilcoxon rank sum test with Bonferroni correction (***P < .005). Within each HPG status subgroup, patients with lipodystrophy and insulin signaling disorders were compared using the Wilcoxon rank sum test (*P < .05). Abbreviations: IR, insulin resistance; LD, lipodystrophy.
a Age corresponding to ovarian histology.
b Measured prior to oophorectomy or ovarian biopsy. Insulin was measured in the fasting state.
c Also reported in Table 4. Normal ovarian volume 5 mL. Table 3. Continued phenotypes, then defects in the insulin receptor, affecting all insulin signaling equally, would be associated with a less severe ovarian phenotype than defects in the PI3K/AKT2 arm of the signaling pathway. This was not observed. The ovarian phenotype we describe in an infant with Donohue syndrome, characterized by minimal residual insulin receptor function, suggests moreover that ovarian effects of extreme hyperinsulinemia are not mediated through the insulin receptor. These findings favor the hypothesis that insulin drives hyperandrogenism and PCOS via action on ovarian IGF1 receptors.
In keeping with this, the ovarian histopathology we describe in diverse forms of SIR demonstrates commonality in appearances irrespective of age/pubertal stage and monogenic defect. All 4 patients with pathogenic variants in INSR showed features consistent with PCOS, including numerous follicular cysts within an extensively luteinized stroma associated with dense stromal proliferation. Other features, such as larger cysts or massive ovarian enlargement, were more variable. Features of PCOS were similarly observed in 3 patients with partial lipodystrophy, including 1 patient with digenic IR (P2) and 1 patient with heterozygous pathogenic variants in PPARG (P8).
Most published series suggest that a testosterone concentration of around 150 ng/dL (~5 nmol/L) is a reasonable trigger for screening for virilizing tumors (48,49). Serum testosterone concentration in excess of this level was observed in 13% of women/girls in our cohort. Our case series demonstrates that SIR is sufficient to induce hyperandrogenemia sometimes in excess of this threshold, which is usually not of tumoral origin. On the other hand, 5 patients with SIR of different etiologies developed ovarian tumors associated with virilization. Added to reports of neoplasia arising in the context of sustained severe   (30) hyperinsulinemia, ovarian hyperthecosis and SIR (13), this suggests that IR-related hyperandrogenemia may not be entirely benign in the long term. We suggest that long-term hyperinsulinemia may increase risk of virilizing ovarian tumors. Screening strategies for autonomous tumors need to be evaluated, possibly using intermittent GnRH suppression testing to look for evidence of autonomous androgen secretion. Finally, we report the treatment of 6 patients with hyperandrogenism due to SIR with gonadotropin suppression, adding to existing literature (30). All exhibited testosterone lowering into the normal female range or frank suppression, and experienced marked symptomatic improvement, supporting a requirement of pulsatile gonadotropins for the adverse actions of hyperinsulinemia to cause ovarian hyperandrogenism. This is in keeping with the synergic action of insulin and luteinizing hormone on ovarian thecal cells in vitro to drive thecal cell hyperplasia and androgen synthesis (50)(51)(52). Suppression of gonadotropins with longer-acting GnRH analogues may thus be a useful therapeutic option for hyperandrogenic patients in a broad range of SIR subphenotypes, and warrants further study in several settings. In infants with Donohue syndrome and marked ovarian enlargement, GnRH analogues may be considered to prevent ovarian enlargement that in some cases is severe, while in young women with extreme IR and hyperandrogenism they may be used in combination with hormone replacement in a "block and replace" strategy when future fertility is desired. This may be an undesirable strategy in dyslipidemia, for example due to lipodystrophy, where estrogens may exacerbate hypertriglyceridemia, however (53). In postmenopausal women, GnRH agonist therapy may allow oophorectomy to be avoided when surgical risk is high, as reported in idiopathic post-menopausal hyperthecosis (54)(55)(56). A note of caution is sounded by a recent report of progressive ovarian enlargement in a woman with SIR treated with long-term GnRH analogue therapy despite effective treatment of hyperandrogenemia. This was attributed to gonadotropinindependent mitogenic actions of insulin on granulosa cells (57).
The retrospective, historical nature of our data is a limitation of this study. In particular, we report TT concentrations determined using various platforms in different referring centers over several decades. Nevertheless, most of these were immunoassay-based measurements and thus broadly comparable. A small minority were determined by mass spectrometry; given the increased specificity of this technique, we would anticipate even higher levels of hyperandrogenemia had they been measured by immunoassay instead. FT was not measured directly, but estimated from TT and SHBG, assuming normal albumin, using Time (in specified units) since onset of GnRH analogue therapy at re-evaluation.
c Also reported in Table 3. Insulin was measured in the fasting state. Table 4. Continued conventional methods. Medication (including insulin) history was recorded in only a subset of individuals and not incorporated into analyses. Finally, whilst we did not programmatically exclude potentially confounding pathophysiology (such as dysthyroidism, hyperprolactinemia, or liver/adrenal disease) in all study participants, this was undertaken in referring centers prior to referral. Overall, the unique size and rarity of this patient cohort allows important, generalizable lessons to be drawn in spite of these drawbacks. IR is reported in up to 70% of cases of "common" PCOS (58, 59) but the direction of causality is debated. Some evidence suggests that primary hyperandrogenemia may cause IR. For example, nonclassical congenital adrenal hyperplasia (prior to glucocorticoid replacement) is associated with decreased insulin sensitivity (60) as is androgen use by healthy women (61,62). The effects of pharmacological androgen reduction or androgen receptor antagonists on insulin sensitivity have been less consistent (63); however, Mendelian randomization in a population-based study has supported a causal link between hyperandrogenemia and both PCOS and type 2 diabetes (64). In keeping with these human observations, rodents exposed to exogenous androgens exhibit many features of PCOS (65). More recently, adiposederived androgens, which are increased in PCOS, have been proposed to drive metabolic abnormalities in PCOS (66). Our data do not provide additional support for this hypothesis, as androgen reduction after GnRH analogue administration or oophorectomy did not discernibly alter insulin sensitivity in severe, primary IR. This does not exclude a bidirectional relationship between hyperandrogenemia and IR, although we suggest that the effect of IR on androgen secretion is likely far greater than the effect of hyperandrogenemia on insulin sensitivity, based on our findings.