PCOS and Insulin Resistance: the Engine Behind the Androgens

You are sitting in the consulting room, the lab printout is in front of you, and the numbers are green. Fasting glucose 91. HbA1c 5.3. All within range. And still something is off. There is more hair on your chin. The skin along your jaw has not settled down for years. Your period comes when it feels like it. And the sentence you hear most often is: well, your glucose values look fine.

Many women know this pattern. The glucose is normal and the hormones are not. That is not a contradiction. That is exactly what you can expect with polycystic ovary syndrome. Because the value doing the work in this picture is not on the standard lab sheet at all. It is not the glucose. It is the insulin your body needs in order to keep that glucose normal.

I am writing this article because I have this conversation almost every week. And because two mistakes sit side by side online on this topic. One of them acts as if insulin resistance in PCOS were a footnote. The other acts as if a HOMA index above 2.5 were a verdict. Neither gets it right. Between these two poles lies a chain that is fairly well understood at the cellular level, and a measurement that is considerably shakier than the calculator pages make it look.

What you get here: the mechanics, explained in images. The numbers as they stand in the original papers. And an honest sorting of which test shows you what and where it is blind. Plus the position of both guidelines, including where they contradict each other. Especially there.

My starting point

In PCOS, a raised insulin level is not an incidental finding. It is one of the engines that keep the androgens up. That is why I look at glucose metabolism just as carefully as at the hormone profile. And that is why, in the same consultation, I add that the common insulin measurements are poorly standardised and that the guideline judges them cautiously for good reasons. I use them as orientation. Not as a verdict.

What this article covers

The chain from insulin through the theca cell to the androgens
Why SHBG in the liver falls at the same time
Why lean women can be affected too
Why fasting glucose shows nothing for a long time
Six ways of measuring, with their limits, honestly sorted
Why there is no universal HOMA cut-off
What the 2023 and 2025 guidelines recommend
Effect sizes for weight, training, nutrition and sleep
Metformin with its benefit and its downside
Why I also look at environment and inflammation

Guideline consensus of professional societies Human studies in people, observational or interventional Tissue human tissue outside the body Cell culture cells in a dish, not in a human being

The chain this is all about

Imagine your insulin as a key. It opens the door for glucose at your cells. When the doors get stiff, the body does not turn up the force, it makes copies. It simply sends more keys. At first you do not see the result in your glucose, because the glucose still gets in. What you do not see is the number of keys it now takes.

And this is where the misunderstanding begins. Insulin is not just a glucose hormone. Insulin is a growth and instruction signal, and it has receivers in places nobody thinks of. Two of those places are decisive for this topic. One sits in the ovary. The other in the liver.

The first lever sits in the ovary

In the wall of every maturing follicle lies a layer of cells, the theca cells. Their job is to build androgens, precursors from which the neighbouring cells then make oestrogen. This cell layer has insulin receptors. And it responds to them.

In vitro, cell study The precise core of the chain

A team around Nestler isolated theca cells from the ovaries of women with PCOS and cultured them in a dish. The researchers then added insulin, plus IGF-1 and blocking antibodies against the respective receptors.

Insulin increased testosterone production in these cells, and IGF-1 did so as well. Blocking the insulin receptor abolished the insulin effect. An effective block of the IGF-1 receptor, by contrast, changed nothing about the insulin effect.

For you this means: in this cell culture the insulin effect on androgen production apparently ran through the insulin receptor itself and not through the IGF-1 receptor. IGF-1 also increased testosterone production in the same paper, so it has a route of its own. And the important limitation right alongside it: this is cell culture, not a living human being.

Nestler JE et al., J Clin Endocrinol Metab 1998. DOI: 10.1210/jcem.83.6.4886 · PMID: 9626131 [In vitro]

The second lever sits in the liver

Sex hormones do not swim around freely in the blood. Most of them travel bound, attached to a transport protein with the unwieldy name sex hormone binding globulin, SHBG for short. Bound testosterone is on the move, but it reaches the target cell less readily. According to the common model, it is mainly the unbound fraction and the fraction only loosely bound to albumin that is active at the hair follicle and at the sebaceous gland. That picture is a simplification, and it is genuinely debated in endocrinology, but it describes the relationships well enough to follow what comes next.

Your liver builds this transport protein. And the liver aligns its production with the metabolic situation.

In vitro, cell study Less transport protein when insulin is higher

Loukovaara and colleagues worked with HepG2 cells, a human liver cell line. The actual subject of the paper was cycloheximide, an inhibitor of protein synthesis. Alongside that, the cells were treated with insulin, thyroid hormone and estradiol, and both the SHBG protein and the corresponding mRNA were measured.

Insulin lowered SHBG production, including at the mRNA level. Thyroid hormone and estradiol increased SHBG production; at the mRNA level thyroid hormone raised it moderately, while estradiol showed no clear effect there.

For you this means: when insulin is high, the liver throttles precisely the protein that would otherwise keep your testosterone bound. This too is cell culture for now.

Loukovaara M, Carson M, Adlercreutz H, J Steroid Biochem Mol Biol 1995. DOI: 10.1016/0960-0760(95)00141-l · PMID: 7662587 [In vitro]

The pathway behind this is now reasonably well described. Simple sugars such as glucose and fructose can drive new fat production in the liver. In the process, the amount of a control factor called HNF-4alpha can drop, and it is this factor that switches on the blueprint for SHBG. Less control factor can mean less transport protein. This pathway too has been described in liver cell cultures and summarised in a review article, not measured in the living human body. That is why specialists say SHBG can be a sensitive metabolic marker and not merely a hormone value.

The chain in four steps

One signal, two directions, one result

The doors get stiff. Muscle and fat cells respond less well to insulin. The pancreas compensates by releasing more insulin. Blood glucose stays unremarkable at first.
The ovary listens in. The theca cells carry insulin receptors. In cell culture, more insulin means more androgen production.
The liver closes the gates. Under the influence of insulin, SHBG production falls. The transporter for sex hormones becomes scarce.
The free fraction rises twice over. More production and less binding push in the same direction. And it is mainly the free fraction that is active at hair and skin.

Important for context: steps 2 and 3 are shown in cell cultures. In humans, the chain appears indirectly, for example through the observation that lower insulin goes together with lower androgens and higher SHBG. That is something other than direct proof in the living body.

And now the context that often gets lost

This chain explains a large part of the picture. It does not explain the whole picture. Two independent reviews arrive at the same careful wording: the dysregulation of androgen production appears to be intrinsic to the ovary itself, and insulin as well as LH amplify it rather than create it on their own. In a large review by Rosenfield and Ehrmann, roughly two thirds of cases show this typical ovarian over-response. And in about half of those affected, the same review describes a metabolic syndrome of weight-related and or intrinsic insulin resistance.

That is an important sentence, and I would like you to take it with you. Insulin is an amplifier. Not a sole culprit. Anyone telling you that PCOS is simply a sugar disease of the ovaries is cutting corners.

Reframe

You do not have two problems. You have one axis. Most women experience their diagnosis as a collection of separate building sites. Skin here, cycle there, weight over there, and the glucose supposedly somewhere else entirely. In physiology these points hang on a shared signalling pathway.

That is not bad news. It is ordering news. When several complaints hang on one axis, then a change at that axis can arrive at several places at once. And it explains why treating symptoms only at the surface often stays exhausting.

Red flags that belong in medical hands promptly

Before we go on, the most important part first. PCOS is a medical diagnosis, and the workup belongs in gynaecological or endocrinological hands. The following signs are not a case for internet research but for a prompt appointment:

very heavy or suddenly changed bleeding
any bleeding after the menopause
acute, one-sided lower abdominal pain
lower abdominal pain together with fever
unintentional weight loss
headaches or visual disturbances, especially together with milk discharge from the breast
rapidly progressing virilisation, meaning a deepening voice, marked muscle gain or strong hair growth within a few months

This list stands here early on purpose. Nothing in this article is a reason to postpone or replace a gynaecological workup.

And now you know why blood sugar and hormones are looked at together in this topic. Not because it is fashionable, but because they are physically connected through two receptor pathways. If you still need the basics of the condition, meaning the Rotterdam criteria, the symptom picture and the differential diagnosis, you will find them in Understanding PCOS: causes and symptoms. Here we continue with the engine.

Why this affects lean women too

There is a sentence that often falls in consulting rooms and that unsettles more than it clarifies. Roughly: you are slim, so you do not have insulin resistance. That sentence sounds logical. It is still not covered by the data.

Human study, reference method The historical evidence from 1989

Dunaif and colleagues studied 19 obese and 10 non-obese women with PCOS against 11 obese and 8 non-obese control women. Measurement used the euglycaemic clamp, the demanding reference method in which insulin and glucose are infused under control.

Insulin-stimulated glucose utilisation was markedly reduced in both PCOS groups, regardless of whether it was related to total weight, to fat-free mass or to the insulin level, each with P less than 0.001. Insulin clearance did not differ between the four groups.

For you this means: the reduced insulin action in PCOS is not simply a consequence of weight. And the high insulin level does not arise because the body breaks insulin down more slowly.

Dunaif A et al., Diabetes 1989. DOI: 10.2337/diab.38.9.1165 · PMID: 2670645 [Cohort, n=48]

Twenty-four years later an Australian group measured the same thing with the same method and delivered numbers worth remembering.

75 %of lean women with PCOS were insulin resistant on the clamp

95 %of overweight women with PCOS

62 %of overweight control women without PCOS

Human study, reference method The number behind the phrase "lean as well"

Stepto and colleagues studied 20 overweight and 20 lean women with PCOS against 14 overweight and 19 lean control women with matched BMI. All had taken no insulin sensitisers and no contraceptive pill for at least three months. That was a study protocol requirement and is explicitly not a suggestion to stop anything before a blood test.

The glucose infusion rate was 339 plus minus 76 in lean controls, 270 plus minus 66 in lean women with PCOS, 264 plus minus 66 in overweight controls and 175 plus minus 96 in overweight women with PCOS, in milligrams per minute and square metre. The relationship between BMI and insulin sensitivity was steeper in PCOS.

For you this means: three out of four lean women with PCOS were insulin resistant in this study. And lean women with PCOS sat on average roughly where overweight women without PCOS sat.

Stepto NK et al., Hum Reprod 2013. DOI: 10.1093/humrep/des463 · PMID: 23315061 [Cohort, n=73]

Where does it come from, if not from weight?

One of the most interesting trails leads to the insulin receptor itself. A receptor is like a switch with several contacts. When insulin docks, certain contacts, the tyrosine sites, are supposed to receive a phosphate. Then the signal travels on. If other contacts, the serine sites, are permanently phosphorylated instead, the switch jams.

Human tissue, ex vivo A possible mechanism in a subgroup of those affected

Dunaif and colleagues examined insulin receptors from skin fibroblasts and skeletal muscle of women with PCOS and of control participants for their phosphorylation pattern.

About 50 percent of the women with PCOS showed a 3.7-fold increased, insulin-independent phosphate deposition at serine sites alongside reduced insulin-triggered tyrosine phosphorylation, with P less than 0.05. The same pattern was found in muscle. The remaining women with PCOS were unremarkable. No mutations in the receptor gene were found.

For you this means: in a subgroup of those affected there is an independent signalling fault that has nothing to do with weight. In roughly half of them. Not in all.

Dunaif A et al., J Clin Invest 1995. DOI: 10.1172/JCI118126 · PMID: 7635975 [Cohort, human tissue]

And what about visceral fat?

Here comes a place where I deliberately write against the current. Online you often read that lean women with PCOS simply have hidden visceral fat that the scales do not show. That sounds plausible. The best available analysis, however, only supports it halfway.

Counter-evidence, cleanly separated

Visceral fat in PCOS: probable, but not as clear-cut as claimed

A meta-analysis pooled 39 studies that compared body fat distribution using imaging between women with PCOS and BMI-matched controls. Compared with controls, more visceral fat was found, with a standardised mean difference of 0.41 and a confidence interval from 0.23 to 0.59. Trunk fat and android fat were higher as well.

And now the part that is rarely quoted along with it: in the subgroup that included only studies using MRI and CT, that is the most accurate methods, no difference remained statistically significant. For visceral fat the difference there was 0.19 with an interval from minus 0.04 to 0.41.

My reading: a connection is plausible and visible with coarser methods. The claim that every lean woman with PCOS has hidden visceral fat goes beyond these data. Fat distribution is one piece of the puzzle, not the explanation.

Zhu S et al., Front Endocrinol 2021. DOI: 10.3389/fendo.2021.697223 · PMID: 34566888 [Meta-analysis, k=39]

Reframe

Your dress size is not a lab value. If you are lean and still have symptoms, you are neither a special case nor an exception to the rule. You are in good company with three out of four lean women from the clamp study.

And if you are not lean, the reverse holds just as much. Your weight is not the cause of your diagnosis, it is one factor among several that can amplify the axis. The difference between those two sentences is the difference between blame and physiology.

If you would like to understand the blood sugar and hormone axis more fundamentally, without the PCOS focus, that is covered in Insulin resistance and hormones in women. And if what preoccupies you most is why the weight stays despite everything, then Losing weight with insulin resistance belongs alongside it. Here I stay with the mechanism.

Why fasting glucose often does not see the problem

Back to that green lab printout from the beginning. Why does it say nothing when something has long been running in the body?

Because your body is a very good bookkeeper. It has one priority that ranks above almost everything: blood glucose should stay in a narrow window. When the cells respond less well to insulin, the body does not lower the standard. It raises the effort. The pancreas sends more insulin, and glucose stays normal. This phase can last for years.

Which means: a normal fasting glucose tells you that the compensation is working. It does not tell you what it costs. It is like a car still managing the incline. The speed is right. But the engine is turning at 5000 revs.

Human study, prospective What the glucose tolerance test finds in addition

Legro and colleagues studied 254 women with PCOS aged 14 to 44 at two centres, one urban and ethnically mixed, one rural. All received a 75 gram glucose load after standardised preparation.

31.1 percent had impaired glucose tolerance, 7.5 percent had diabetes. Among the non-obese women with a BMI under 27 it was 10.3 and 1.5 percent. Compared with controls the odds ratio was 2.76 with an interval from 1.23 to 6.57. And the point at issue here: criteria based on fasting glucose found 3.2 percent with diabetes, the criteria using the glucose tolerance test found 7.5 percent.

For you this means: in this study the fasting value alone would have missed about half of the diabetes cases. And lean women were affected too.

Legro RS et al., J Clin Endocrinol Metab 1999. DOI: 10.1210/jcem.84.1.5393 · PMID: 9920077 [Cohort, n=254]

HbA1c is convenient and still coarse

The long-term value has one unbeatable advantage. You do not have to fast, you do not have to sit around for two hours, and one tube is enough. That is exactly why it gets used so readily. Its resolution for this topic is simply limited.

Human study, prospective What an HbA1c threshold misses

Celik and colleagues studied 252 women with PCOS according to the Rotterdam criteria and 117 control women in a university outpatient clinic and compared HbA1c directly against the glucose tolerance test.

On the glucose tolerance test, 41 of the 252 women had a glucose disorder. With an HbA1c threshold of 5.6 percent, 20 of these 41 women would not have been identified. Sensitivity at this threshold was 52.4 percent, specificity 74.4 percent.

For you this means: almost half of the abnormal findings would have slipped through the net. Not because the value was measured wrongly, but because it answers a different question.

Celik C et al., Hum Reprod 2013. DOI: 10.1093/humrep/det002 · PMID: 23335611 [Cohort, n=252]

Meta-analysis, 2628 women The number behind the guideline recommendation

Belsti and colleagues analysed nine studies with a total of 2628 women with PCOS and tested fasting glucose and HbA1c against the glucose tolerance test as reference. This work directly informed the 2023 international PCOS guideline.

An HbA1c from 6.5 percent reached a pooled sensitivity of 50.00 percent with an interval from 35.53 to 64.47, at a specificity of 99.86 percent. A fasting glucose from 7.0 millimoles per litre reached a pooled sensitivity of 58.14 percent at a specificity of 92.59 percent. The authors write that this HbA1c threshold could lead to misdiagnosis in half of the women with type 2 diabetes in this group.

For you this means: an unremarkable HbA1c rules out little in PCOS. An abnormal one, by contrast, is highly informative, because specificity is high.

Belsti Y et al., Diabetes Metab Syndr 2024. DOI: 10.1016/j.dsx.2024.102970 · PMID: 38442646 [Meta-analysis, k=9, n=2628]

Reframe

A normal value is not an all-clear, it is a snapshot taken under resting conditions. Fasting glucose measures your metabolism at idle. Your symptoms, though, do not arise at idle, they arise under load.

That is why the glucose tolerance test is not an added luxury in this topic. It is the test that asks the question your everyday life asks every day: what happens when sugar arrives?

And now you know why a green lab printout and a restless hormone balance can be true at the same time.

What can be measured how, honestly sorted

Now it gets practical. Six routes are on the table, and all six get muddled online. I sort them by what they show, and above all by where they reach their limit. In this topic the limit is the more important half of the information.

Six ways of measuring in PCOS. This table does not replace a medical assessment, it is meant to show you the logic behind it.
What is measured	What it shows	Where the limit lies
Fasting insulin and HOMA index	How much insulin your body needs at rest. Simple arithmetic, a single tube.	The insulin assays used by laboratories are not harmonised. There is no cut-off that holds across assays. The German guideline advises against it outside of studies.
75 gram glucose tolerance test, optionally with insulin	How your metabolism responds under load. The most accurate routine route for glycaemic status.	Time consuming, two to three blood draws. There are no validated cut-offs for the additional insulin curve.
HbA1c	Average glucose over the last eight to twelve weeks. Convenient, no fasting needed.	Coarse. Pooled sensitivity of 50.00 percent against the glucose tolerance test at a threshold of 6.5 percent.
Triglyceride to HDL ratio	A rough pointer to insulin action, from two values that almost every lipid panel already contains.	Not a diagnostic value. Cut-points differ markedly between groups. In women on the pill and with lipid disorders the picture shifts.
SHBG	An indirect marker. Low SHBG goes together with less favourable glucose metabolism in large cohorts.	Influenced by the pill, thyroid, liver and nutrition. Not interpretable as a single value.
Continuous glucose monitoring, sensor	Your own pattern across days. An observation tool for everyday life, meals, sleep and stress.	Not a diagnosis. For PCOS there are essentially no robust intervention data. Risk of misunderstanding by overrating single spikes.

The costliest blind spot: the HOMA index has no universal cut-off

If you search online for HOMA, you find calculators. And above every calculator sits a number, usually 2.0 or 2.5, occasionally 3.0. These numbers look like natural constants. They are not.

The HOMA index is a simple calculation from fasting insulin and fasting glucose. Glucose is well standardised between laboratories. Insulin is not. And because insulin sits in the numerator of the calculation, every imprecision of the insulin test travels one to one into your result.

Human study, German population cohort Why a single cut-off is not serious

Matli and colleagues analysed the Gutenberg Health Study, a population-based cohort with 15,030 participants aged 35 to 74. Fasting glucose, insulin and C-peptide were available for 10,340. In addition they compared three different insulin immunoassays against each other.

Median HOMA-IR was 1.54 without diabetes, 2.00 with prediabetes and 4.00 with diabetes. In the most strictly selected reference group of 1,065 people, the median was 1.09 and the 97.5th percentile 2.35, with no meaningful difference between men and women. The comparison of the three assays showed unsatisfactory correlation and systematically different HOMA values. The authors conclude: assay-specific reference intervals are needed.

For you this means: a HOMA of 2.6 from lab A and a HOMA of 2.6 from lab B are not the same statement. Anyone who gives you a universal boundary is simplifying something that cannot be simplified.

Matli B et al., Clin Chem Lab Med 2021. DOI: 10.1515/cclm-2021-0643 · PMID: 34380182 [Cohort, n=10,340]

The sentence I keep for this topic

The HOMA index is a pair of binoculars, not a microscope. It can show you a direction. It cannot set you a boundary. If your value is 2.4 and someone else's from another laboratory is 2.8, that is not a medical difference but possibly a difference in the test kit.

Two values that are already on your printout

There is an underrated detour. Triglycerides and HDL cholesterol appear in almost every standard lipid panel. Both are well standardised between laboratories, unlike insulin. And their ratio to each other says more in women with PCOS than you might think.

Human study, cross-sectional A rough pointer from standard values

Blum and colleagues studied 171 premenopausal women with overweight or obesity and without PCOS, and 71 women with PCOS. Insulin action was measured via steady-state plasma glucose in the modified insulin suppression test, meaning with a true measure and not with an estimating formula.

In the PCOS group, the triglyceride to HDL ratio, the ratio of total cholesterol to HDL and HDL alone each reached an AUROC above 0.80. The optimal cut-points were 1.3 for triglyceride to HDL, 3.4 for total cholesterol to HDL and 52 milligrams per decilitre for HDL. In the group without PCOS, the best cut-point for triglyceride to HDL was 1.9 at an AUROC of 0.73.

For you this means: two values you probably already have can give a rough orientation. The cut-point does differ markedly between groups, though, and a different one applies to premenopausal women than to men.

Blum MR et al., J Endocrinol Invest 2021. DOI: 10.1007/s40618-020-01430-2 · PMID: 33687700 [Cohort, n=242]

SHBG as a warning light

Do you remember the second lever from the first section? That one can actually be measured. A low SHBG is therefore not an isolated hormone value but a window onto the liver and the metabolism.

Human study, nested case-control What a low SHBG predicts in large cohorts

Ding and colleagues studied 359 women with newly diagnosed type 2 diabetes against 359 controls within the Women's Health Study, all postmenopausal and without hormone therapy. They also examined two SHBG gene variants in a Mendelian randomisation.

Relative to the lowest quartile, the multivariable odds ratios were 0.16 in the second, 0.04 in the third and 0.09 in the fourth quartile, with a trend P below 0.001. In the Mendelian randomisation, each standard deviation of higher SHBG gave an odds ratio of 0.28 in women.

For you this means: SHBG is more than a footnote. But take care when transferring this, because these women were postmenopausal and did not have PCOS. For a 28 year old with PCOS this is an analogy, not proof.

Ding EL et al., N Engl J Med 2009. DOI: 10.1056/NEJMoa0804381 · PMID: 19657112 [Cohort, n=718]

The sensor: a window, not a diagnosis

Finally, continuous glucose monitoring. A small sensor on the upper arm that measures every few minutes. Curiosity about it is high, and I understand that well. It shows you your own pattern instead of an average.

Only this window needs a frame, otherwise it frightens you for no reason. A study of 153 metabolically healthy people aged 7 to 80 provides that frame: mean glucose 98 to 99 milligrams per decilitre in every age group except those over 60, where it was 104, a median of 96 percent of time between 70 and 140, mean coefficient of variation 17 plus minus 3 percent. And this median: 2.1 percent of time above 140, which is roughly 30 minutes a day. In healthy people too. So a single spike is not a finding.

If you want to know how such a sensor phase runs in practice and what can be drawn from it, that is covered in 14 days with a glucose sensor. And if you are stuck on the question of how to build meals so the curve stays calmer, then Avoiding blood sugar spikes belongs alongside. Both are an addition, not diagnostics.

Reframe

A test that has a limit is not a bad test. It is an honest one. The error does not arise from the measurement but from the expectation that a number could describe a person.

In consultations I often say: we are not collecting proof, we are collecting clues. And then we look at whether they point in the same direction as what you feel in everyday life. When lab results and lived experience diverge, that is a question and not a fault of yours.

What the guidelines say, and why they differ

Now comes the section I actually wrote this article for. Because this is where an experience arises that many women know. You bring a HOMA printout with you. The response is reserved to dismissive. And you go home with the feeling that somebody did not listen.

I would like to show you that this response is, as a rule, not indifference. It is a correct quotation.

The international guideline 2023

The international evidence-based PCOS guideline was updated in 2023, supported among others by the Endocrine Society, ESHRE and ASRM, using GRADE methodology, 58 prioritised clinical questions and involvement of professional societies and patient organisations from 71 countries. The result is 254 recommendations and practice points. The authors describe the quality of evidence themselves as low to moderate. That is remarkably open.

Position of the 2023 international guideline

On glycaemic status in PCOS

The most accurate test: The 75 gram glucose tolerance test, regardless of body weight.
The substitute options: Fasting glucose and HbA1c may be considered, explicitly with a note on markedly reduced accuracy.
The rhythm: Assessment at diagnosis, then reassessment every one to three years depending on individual risk profile.
Trying to conceive and pregnancy: Consider a glucose tolerance test before fertility treatment and again at 24 to 28 weeks of pregnancy.

Source: Teede HJ et al., Recommendations From the 2023 International Evidence-based Guideline. J Clin Endocrinol Metab 2023. DOI: 10.1210/clinem/dgad463 · PMID: 37580314 [Guideline]

The German S2k guideline 2025

Since June 2025 there has been a first national German PCOS guideline, issued by the German Society of Endocrinology under AWMF registry number 089-004. S2k means consensus-based, with a formalised voting procedure and documented consensus strength for every single recommendation.

Two recommendations from it appear to contradict each other, and that is precisely why it is worth reading them alongside a third.

Position of the German S2k guideline 2025

Three recommendations that belong together

Recommendation 2.2.1: In all women with PCOS without known diabetes, carbohydrate metabolism should be evaluated at first diagnosis and thereafter every one to three years depending on the individual risk profile. Possible options are the glucose tolerance test, HbA1c and or fasting glucose. Consensus strength 11 of 11.
Recommendation 1.3.2: In women with PCOS, parameters of insulin resistance should not be measured outside of studies or scientific research questions. Consensus strength 11 of 11.
Recommendation 1.3.1: To assess the individual risk constellation, glucose metabolic status should be checked. Consensus strength 18 of 18.

Source: DGE, S2k guideline on the diagnosis and treatment of polycystic ovary syndrome, AWMF registry number 089-004, version 1.1, June 2025 [Guideline]

At first glance this sounds paradoxical. Measure glucose yes, measure insulin no. At second glance it is highly consistent.

The three reasons behind the no, written out

First, the measurement technique. Insulin immunoassays are not harmonised. The German Gutenberg cohort showed this across three assays, they correlated unsatisfactorily and delivered systematically different HOMA values. A lab value that comes out differently depending on the kit makes a poor criterion.

Second, the missing cut-off. The 97.5th percentile in the strictly selected reference group was 2.35, but that applies to exactly that assay and that population. There is no validated, generally applicable cut-point for insulin resistance.

Third, the missing consequence. And that is the strongest argument. The same guideline records in recommendation 4.2.2 that metformin can be used in women with PCOS even without testing glucose and insulin metabolism, with a consensus strength of 10 of 10. If the treatment decision does not depend on the measurement, then the main reason for measuring falls away.

This logic is clean. A test that is not standardised, has no cut-off and changes no decision does not belong in routine care. That is how a good guideline thinks, and it is right about that.

A guideline protects you from being treated for every number. That is its job, and it does it well.

Shukri Jarmoukli

And now my view, clearly separated

What follows is not evidence. It is a clinical impression, and I mark it explicitly as such.

What I observe clinically

Why I still look at a markedly raised fasting insulin

The mechanistic data are unusually consistent for this field. Insulin increased androgen production in theca cell cultures, it lowered SHBG in liver cell cultures, and in two small intervention studies with roughly a dozen women each, androgens went down as insulin levels fell. A markedly raised fasting insulin is therefore not a pointless finding. It can be a clue to an axis that is contributing to the picture.

The second reason is, to be honest, a human one. Some women describe this moment to me like this: for years they were told the values are normal, and then there is a number that matches what they experience day to day. That can release something. That is my observation from conversations and not a study finding. It can explain why the body behaves the way it does. That is a benefit in understanding, not diagnostic proof, and I say so every time.

The third point is the limit I set myself. I do not treat a number. I do not treat a HOMA index. If nothing can be derived from a value that I would not also do without it, then I do not need it. And if a woman leaves the room more anxious after a printout than she was before, then the measurement did harm and not good.

Where I explicitly follow the guideline

Three things I do not derive from an insulin value

First: no diagnosis. Insulin resistance in PCOS is not a diagnosis that a fasting value could make. The reference method is the clamp, and that happens in studies.

Second: no cut-off drama. I do not tell any woman that she is ill from 2.5 upwards and healthy below. That line does not exist in the data.

Third: no substitute action. A fasting insulin never replaces the glucose tolerance test when it comes to glycaemic status. Both guidelines agree on this point, and that is the point that counts for your long-term health.

Reframe

If your doctor does not go along with the HOMA index, she is not arguing with you. She is quoting correctly. Knowing that changes the conversation. You do not have to defend your perception against a guideline.

The better question in the consulting room is therefore not: can we please do the HOMA? It is: when was my last glucose tolerance test, and when is the next one due? That question is covered by both guidelines, and it answers what really counts.

Two practical points go with that, so the question lands well in the consulting room. The test needs a normal, carbohydrate rich diet in the days beforehand, otherwise the result can come out falsely abnormal. And it is not meant for every situation: with already known diabetes, during an acute infection and after certain stomach operations, your doctor will decide differently.

And now you know why two professional worlds speak differently on this point without either of them being unserious.

What lowers insulin levels in studies

Now to the part you were probably looking for most. And I do it deliberately differently from most pages here: with effect sizes instead of adjectives. Because a measure that moves something moves a certain amount. Knowing that amount protects you from disappointment and from false promises alike.

One note up front that matters to me. All numbers in this section are study results from groups. They are not a target for you, not a plan and not a yardstick. If the topic of weight is hard for you, or if eating is bound up with pressure and control, please read this section at a distance and talk about it. More on that further down. And one more point that matters especially to me in this topic: if you want a child, nothing in this section is a reason to push a gynaecological or reproductive medicine workup back. Time is a real factor in that question. What is written here can accompany such a workup, but it cannot replace it and it must not delay it.

Weight change: real and modest

Human intervention study Eight weeks, small group, clear movement

Jakubowicz and Nestler followed 12 obese women with PCOS and 11 obese women with regular menstrual cycles through eight weeks of a hypocaloric diet and measured steroids and the response to a stimulation test before and afterwards.

In the PCOS group, free testosterone fell from 9.03 plus minus 1.39 to 5.95 plus minus 0.50 picomoles per litre and total testosterone from 2.47 to 1.56 nanomoles per litre. The stimulated peak value of 17-hydroxyprogesterone fell from 14.9 plus minus 2.6 to 8.9 plus minus 0.8 nanomoles per litre. SHBG rose 4.5-fold in the PCOS group. In the control women the androgens did not change.

For you this means: the second lever from the first section can be seen in action here, SHBG came up markedly. And the necessary limitation: 12 women, eight weeks, an obese starting point. That is a signal, not a rule.

Jakubowicz DJ, Nestler JE, J Clin Endocrinol Metab 1997. DOI: 10.1210/jcem.82.2.3753 · PMID: 9024253 [Human intervention, n=23]

Meta-analysis, Cochrane The most honest number in this field

Moran and colleagues analysed six randomised studies with 164 participants for Cochrane, in which lifestyle treatment was tested against minimal or no treatment.

What they found: weight minus 3.47 kilograms, fasting insulin minus 2.02 microunits per millilitre, waist circumference minus 1.95 centimetres, total testosterone minus 0.27 nanomoles per litre and the hirsutism score minus 1.19. And explicitly no evidence of an effect on BMI, free androgen index, SHBG, glucose or cholesterol. On pregnancy, live birth and miscarriage there were no data at that time.

For you this means: the effect is real, and it is modest. Anyone promising you miracles here does not know this table. And for the questions that occupy many women most, there were simply no robust numbers.

Moran LJ et al., Cochrane Database Syst Rev 2011. DOI: 10.1002/14651858.CD007506.pub3 · PMID: 21735412 [Meta-analysis, k=6, n=164]

Why a calorie deficit alone is often not the whole picture is covered in Calorie deficit: why it is not always enough. And who else has a hand on the weight, meaning leptin and the satiety signals, is covered in Leptin and insulin.

Movement: three meta-analyses in favour, one against

Here it gets interesting, because the data do not agree. And I think you should know that instead of only reading the convenient numbers.

Meta-analysis Exercise against control

Kite and colleagues analysed 27 publications from 18 studies on exercise or exercise plus diet in PCOS.

Compared with control, fasting insulin fell by 2.44 microunits per millilitre with an interval from minus 4.24 to minus 0.64, HOMA-IR by 0.57, waist circumference by 2.62 centimetres, and VO2max rose by 3.84 millilitres per kilogram per minute. The authors explicitly rated the evidence quality for insulin and HOMA as very low and urged caution because the confidence intervals are wide.

For you this means: exercise moves something on insulin. And the authors themselves say how uncertain that estimate is.

Kite C et al., Syst Rev 2019. DOI: 10.1186/s13643-019-0962-3 · PMID: 30755271 [Meta-analysis, k=18]

Meta-analysis, 777 women Intensity appears to count more than volume

Patten and colleagues analysed 19 studies with 777 women, with intervention durations between 6 and 26 weeks.

After training at vigorous intensity, VO2max rose by 24.2 percent, HOMA-IR fell by 36.2 percent with 90 percent confidence limits from minus 55.3 to minus 9.0, and waist circumference fell by 4.2 percent. The authors conclude that intensity appears to matter more than volume and that at least 120 minutes of vigorous intensity per week appear to be needed.

For you this means: the magnitude is considerable, and the threshold sits at around two hours of demanding movement per week. The 90 percent confidence limits from minus 55.3 to minus 9.0 also show how wide the range is.

Patten RK et al., Front Physiol 2020. DOI: 10.3389/fphys.2020.00606 · PMID: 32733258 [Meta-analysis, k=19, n=777]

The counter-finding I do not leave out

Two analyses find an effect, one finds none

A meta-analysis of seven randomised studies with 423 participants found, for high-intensity interval training, a fall in HOMA-IR of 0.57 with an interval from minus 0.98 to minus 0.16 and moderate certainty of evidence, plus a BMI reduction of 1.90 with high certainty of evidence.

Another meta-analysis of 16 studies compared high-intensity interval training directly against moderate continuous training. Result: neither interval training with a change of minus 0.257 and a P of 0.374 nor moderate continuous training with minus 0.341 and a P of 0.078 improved HOMA-IR statistically significantly. Moderate continuous training, by contrast, improved VO2max markedly.

My reading: the meta-analyses do not agree here, and that is down to small sample sizes and very different training protocols. What improved consistently across all analyses is fitness. That is a value in itself, independent of the HOMA index.

Santos IK et al., PLoS One 2021, DOI: 10.1371/journal.pone.0245023 · PMID: 33465123 [Meta-analysis, k=7, n=423] · Richards CT et al., Front Physiol 2021, DOI: 10.3389/fphys.2021.715881 · PMID: 34483969 [Meta-analysis, k=16]

And strength work? The German S2k guideline 2025 recommends in recommendation 3.4.2 endurance and strength training at individually adapted intensity, with a consensus strength of 12 of 12. Mechanistically that is well founded, because in healthy people skeletal muscle takes up the largest share of insulin mediated glucose uptake. Studies testing strength against endurance in PCOS directly on the HOMA index are, however, rare and small. This is an area where the mechanism reaches further than the data.

Dietary patterns: a ranking from small groups

Network meta-analysis Which pattern led on glucose values

A Hungarian research group around Juhász analysed 19 randomised studies with 727 patients and compared ten dietary approaches plus metformin against each other in a network, evaluated with Bayesian ranking probabilities.

The DASH pattern reached the highest ranking values for HOMA-IR with a SUCRA of 92.33 percent, for fasting glucose with 85.92 percent, for fasting insulin with 79.73 percent and for triglycerides with 82.07 percent. For BMI, the calorie-restricted diet led with 84.59 percent, and for weight loss the combination of calorie-restricted diet plus metformin with 74.38 percent.

For you this means: there is no magic PCOS eating pattern in the data. And the limitation is considerable: 727 women spread across ten interventions gives very small groups, and SUCRA values are ranking probabilities, not effect sizes.

Juhász AE et al., Reprod Health 2024. DOI: 10.1186/s12978-024-01758-5 · PMID: 38388374 [Meta-analysis, k=19, n=727]

The DASH pattern, by the way, is nothing exotic. Plenty of vegetables, legumes, nuts, little sugary food, few highly processed products. How nutrition, blood sugar and the female hormones touch each other is covered in more depth in Nutrition, hormones and blood sugar. And why the liver plays a special role with fructose, which matters for SHBG, is covered in Sugar, fructose and liver metabolism.

Sleep: the lever almost nobody has on the list

Human intervention study One single short night

Donga and colleagues studied nine healthy people twice, once after normal sleep duration and once after a night with four hours of sleep. Insulin action was measured with the clamp, sleep was monitored by polysomnography.

Sleep duration was 226 plus minus 11 versus 454 plus minus 9 minutes. After the short night, endogenous glucose production was higher, glucose disposal fell from 40.7 to 32.5, and the glucose infusion rate fell by about 25 percent with a P of 0.001. Free fatty acids rose.

For you this means: one single shortened night lowered insulin action measurably. Important limitation: nine metabolically healthy people without PCOS. What transfers is the direction, not the exact number.

Donga E et al., J Clin Endocrinol Metab 2010. DOI: 10.1210/jc.2009-2430 · PMID: 20371664 [Human intervention, n=9]

Human study, sleep laboratory Breathing during sleep and insulin

Vgontzas and colleagues studied 53 women with PCOS and 452 premenopausal control women from a random sample for one night each in the sleep laboratory.

The risk of sleep-disordered breathing was 30-fold higher in the women with PCOS, with an odds ratio of 30.6 and a very wide interval from 7.2 to 139.4. Treatment was recommended in 9 of 53 compared with 3 of 452 controls. Daytime sleepiness was reported by 80.4 versus 27.0 percent. The affected women with PCOS had higher fasting insulin levels, while testosterone and fasting glucose did not differ between the PCOS subgroups.

For you this means: insulin was a stronger risk factor for the breathing disorder here than BMI or testosterone. The wide confidence interval also shows how small the sample was.

Vgontzas AN et al., J Clin Endocrinol Metab 2001. DOI: 10.1210/jcem.86.2.7185 · PMID: 11158002 [Cohort, n=505]

If you snore, wake up with a headache or never really feel awake during the day, that belongs in a proper workup. How that works and which signs point to it is covered in Recognising sleep apnoea. And if you regularly lie awake at three in the morning, Waking up at night is the better entry point.

Reframe

The biggest numbers in this section do not stand next to food. They stand next to training intensity and next to sleep. That is not what most women with PCOS hear first.

I find that a relief. Because sleep and movement are areas where it is not about going without, but about adding. And if your head tightens at the word nutrition, then it is wise to start at another corner of the axis. The axis is round. You can get on anywhere.

And now you know which orders of magnitude sit behind the advice you meet everywhere.

Metformin, honestly weighed up

Hardly any medicine gets discussed as often in PCOS. And hardly any is experienced so differently. For some it is the turning point, for others two weeks of stomach ache and then done. I will try to let both stand side by side here.

First the most important sentence, and it stands at the beginning on purpose and not at the end: metformin is a prescription medicine. In PCOS it is used in Germany in several situations outside its licence, meaning off label. Whether it is an option for you, whether it fits, and what it has to do with your other findings is decided solely by your doctor together with you. An existing course is not changed, reduced or stopped on your own, and a new one is not started on your own. Nothing in this section is an instruction to act.

Why it works on this axis at all

Randomised trial The chain runs in both directions

Nestler and Jakubowicz gave 24 obese women with PCOS either metformin or placebo at random over four to eight weeks and measured steroids, LH and the insulin response in a glucose tolerance test before and afterwards.

In the eleven women in the metformin arm the area under the insulin curve fell from 9303 plus minus 1603 to 4982 plus minus 911 microunits per millilitre and minute. Basal 17-hydroxyprogesterone fell from 135 plus minus 21 to 66 plus minus 7 nanograms per decilitre, free testosterone from 0.34 to 0.19 nanograms per decilitre, basal LH from 8.5 to 2.8, and SHBG rose from 0.8 to 2.3 micrograms per decilitre. In the placebo group nothing changed significantly.

For you this means: less insulin went together with fewer androgens and more SHBG in this study. So the chain from the first section can be rolled up from the other end too. Important limitation: 24 women in total, only eleven of them in the metformin arm, a very short duration, an obese starting point.

Nestler JE, Jakubowicz DJ, N Engl J Med 1996. DOI: 10.1056/NEJM199608293350902 · PMID: 8687515 [RCT, n=24, active arm n=11]

What the German guideline says about it, in full

S2k guideline 2025, recommendations on metformin

Benefit and limitation in the same document

Fields of use: Named among others are support of weight reduction from a BMI of 25, diabetes prevention in off-label use, hyperandrogenaemia and hyperandrogenism, cycle regulation and the lipid profile. Use is possible regardless of BMI.
The limitation that stands right next to it: For hyperandrogenism and for cycle regulation, the lower efficacy compared with alternatives must explicitly be pointed out.
Recommendation 4.2.8: Women with PCOS should be informed that basic measures can achieve results comparable to treatment with metformin. Consensus strength 10 of 10.
Recommendation 4.2.9: Dosing is titrated slowly according to the product information. That is the reason why many women experience tolerability so differently.
Recommendation 4.3.1: Pioglitazone cannot be recommended. Consensus strength 11 of 11.

Source: DGE, S2k guideline PCOS, AWMF 089-004, version 1.1, June 2025 [Guideline]

What is checked before any prescription

The most common side effects are gastrointestinal, which is why the guideline provides for a slow dose titration according to the product information. Beyond that, there are points that are often missing from articles written for the public, although they are checked before any prescription:

Metformin is not licensed in markedly reduced kidney function, because in rare cases lactic acidosis can develop. Kidney values are therefore checked before starting and during treatment.
It is usually paused before examinations with iodinated contrast media and before surgery.
Particular caution applies in liver disease, with regular alcohol consumption and in acute illness involving oxygen deficiency or severe fluid loss.
With longer use vitamin B12 levels can fall, which is why they are usually monitored.

This list does not replace the product information and is not complete. Which of these points apply to you is clarified solely by your doctor together with you.

I find recommendation 4.2.8 remarkable. A guideline that discusses a medicine and in the same breath records that basic measures can achieve something comparable is not a guideline that wants to sell something. That is fairness towards the patient.

Meta-analysis, 13 randomised trials Metformin compared with other insulin sensitisers

Melin and colleagues compared metformin, rosiglitazone and pioglitazone in PCOS on behalf of the international guideline work and included 13 randomised trials from 1660 screened publications.

Metformin did better than rosiglitazone on weight with minus 4.39 kilograms, on BMI with minus 0.95 and on testosterone with minus 0.10 nanomoles per litre. Against pioglitazone no difference appeared. Adding a glitazone to metformin did not improve metabolic outcomes. One qualifying finding from the same paper belongs here: rosiglitazone appeared superior to metformin in lowering lipid concentrations.

Two sentences of context belong here, because these two names would otherwise land wrongly. Rosiglitazone has not been on the market in the European Union since 2010, its marketing authorisation is suspended over concerns about cardiovascular risk. Pioglitazone is a prescription medicine and is judged with restraint because of possible fluid retention, heart failure and bone fractures, and a bladder cancer signal is under discussion. Both appear here only as comparator substances from a study and are not an option to ask for.

For you this means: if an insulin sensitiser is used, then as a rule it is metformin. According to this analysis the glitazones add little overall, and the German guideline advises against pioglitazone.

Melin JM et al., Clin Endocrinol (Oxf) 2023. DOI: 10.1111/cen.14983 · PMID: 37933831 [Meta-analysis, k=13]

Two side notes for completeness. Anti-androgens are also prescription medicines. One point belongs here without exception: they can disturb the development of a male fetus and are therefore only used together with reliable contraception. When you are trying to conceive they are not an option. Depending on the substance, further points come in, for example a rise in potassium or effects on the liver, which is why blood monitoring is usual. For cyproterone acetate a dose dependent risk of meningioma is also described, which is why its use has been restricted by the regulators.

On the evidence: a systematic analysis of 20 studies, 13 of which were pooled, found no significant difference for hirsutism and SHBG compared with placebo plus lifestyle. The conclusion of the authors has two halves, and both belong here. First, current evidence does not support anti-androgens as a preferred alternative to the combined pill for hyperandrogenism. Second, in the view of the same authors they could be considered for hirsutism where the pill is contraindicated, poorly tolerated, or gives an insufficient response after at least six months. Compared with metformin plus lifestyle, anti-androgens plus lifestyle did better in the same paper on hirsutism, SHBG and fasting insulin.

And because this could otherwise land wrongly: none of it is a reason to stop, reduce or pause a prescribed anti-androgen. Starting, continuing and ending such a treatment is decided solely by your doctor together with you. And on inositol there is a systematic analysis for the 2023 guideline work whose conclusion is that the evidence is limited and not conclusive. More on that belongs in its own text, here it stays with this one neutral sentence and no recommendation.

Reframe

A medicine is not a defeat, and going without one is not a victory. I meet both camps, and both argue from a stance rather than from a weighing up.

The sober version reads: there is an option with a documented, moderate benefit, with a common gastrointestinal side effect, with contraindications and monitoring points that belong in medical hands, and with a guideline that adds that basic measures can achieve something comparable. From that combination a good, individual decision can be built. Just not alone and not on the internet.

Why I also look at environment, sleep and inflammation

If you have read this far, you know a chain that explains a lot. And you know its limit. Insulin is an amplifier of an ovarian dysregulation, not its sole cause. So one question stays open: what else lifts insulin levels and androgens along with them?

At this point I look more broadly in my consultations than is usual, and I would like to explain why. Not so that you become more afraid. But so that you understand why I ask questions you may not have heard elsewhere.

Four players pulling at the same axis

Chronic stress and cortisol. Cortisol raises blood glucose in the short term, that is its job. Under lasting strain the body can stay in a mode in which more insulin is needed to accommodate the same glucose. This rests mainly on short stress experiments and observational studies in humans, not on long intervention studies. How that lands in the female cycle is covered in Cortisol, stress and female hormones.

Sleep and breathing during sleep. The numbers are further up. A night with four hours of sleep lowered insulin action by about a quarter in a small study, and women with PCOS had markedly more breathing disorders in a sleep laboratory study. That belongs in the questions asked.

Silent inflammation. Inflammatory messengers can disturb the insulin signalling chain at the cell. This mechanism is described mainly in cell experiments and animal models, in humans it is supported by observational studies. It is one of the reasons why weight, gut and insulin action can hang together. More on that is covered in Silent inflammation and weight.

The thyroid. An underactive thyroid can change cycle, weight and metabolism and thereby create a picture that resembles PCOS or overlays it. That is why it belongs in every initial workup. The connection is covered in Thyroid and female hormones.

And the point I sometimes get sideways looks for

There are environmental substances that can dock onto hormone receptors. That is not a fringe opinion, that is toxicology. And so it is clear what this rests on: the evidence for that docking comes from receptor and cell experiments and from animal models, in humans observational studies are added. Plasticisers and certain industrial chemicals have been the subject of regulatory assessments for decades. And there are mould toxins with an oestrogen-like structure, above all zearalenone, which is well documented in animal models and in livestock farming. What that looks like in everyday life is covered in Xenoestrogens in everyday life and in Zearalenone as a mycoestrogen. How the liver metabolises oestrogens and what the gut has to do with it is covered in Oestrogen, liver and relief.

And now the sentence that belongs in every text of this kind, including mine: from the fact that individual environmental substances can be hormonally active, it does not follow that every hormonal disorder has an environmental cause. The evidence in this field consists mostly of cell experiments, animal models and observational studies. Intervention studies in humans showing that reducing exposure changes a PCOS do not exist in a quality that would carry such a claim. Anyone presenting it differently goes beyond the data, and that damages the topic more than it helps it.

Why I still ask: because the question costs nothing and because in individual cases it finds something. Water damage behind the bedroom wardrobe, a workplace with solvents, a striking cluster within a household. Those are not diagnoses. They are puzzle pieces that often find no room in a short appointment, quite regardless of who is holding that appointment.

The paragraph against compulsive control

I have seen how this topic can tip over. Curiosity becomes a list. The list becomes a rulebook. And at some point every meal is a test, every packaging a risk and every restaurant visit a burden. That is not health any more. That is a second stressor on top, and stress works on the same axis as insulin.

So I say it explicitly: perfection is not a goal here and not an advantage either. Exposure is lowered in large, simple steps, not in a hundred small rules. And if you notice that eating and control are closely bound together for you, that thoughts about food take up a lot of space, or that going without no longer feels like a decision, then that is a topic of its own with help of its own. In Germany you can reach the eating disorder helpline of the Federal Centre for Health Education free and anonymously on 0221 892031. If you are feeling very bad, or if thoughts of not wanting to live come up: Telefonseelsorge is available around the clock on 0800 111 0 111, and in an emergency call 112. If you also want to read up on it, Understanding eating disorders is a starting point. Raising this point is not a weakness, it is good medicine.

And on the relationship with gynaecology

I do not want to set up a front here, because there is none. The gynaecological workup remains the foundation. Ultrasound, hormone profile, classification according to the recognised criteria and ruling out other causes of raised androgens. That is the ground on which everything else stands, and without that ground the rest is speculation.

Two of those other causes I want to name, because they often get lost. Non-classical congenital adrenal hyperplasia, that is a partial 21-hydroxylase deficiency, can produce a picture that looks very much like PCOS. It is distinguished through 17-hydroxyprogesterone in the early follicular phase, and it has different consequences, among others when you are trying to conceive and in genetic counselling. Cushing syndrome too can cause cycle disturbance, weight gain and insulin resistance at the same time and should be checked when the corresponding signs are there. Both are clarified by the gynaecological or endocrinological workup, not by this article.

That environmental factors rarely come up in this has, in my view, nothing to do with indifference. It has to do with training priorities and with time budgets that are tight in routine care. A conversation about living situation, workplace and sleep needs time, and time is a scarce resource in every consultation. I take that time for these questions because they are my area of focus. That is a question of format and not a statement about the professional quality of other practices.

Reframe

The environmental perspective comes on top, not instead. It replaces no diagnosis, no workup and no treatment that is offered to you. It adds questions that in individual cases explain something.

And it has an honest evidence base that I will not paint prettier than it is. Strong mechanisms, good animal data, thin human data. That is exactly why I handle it in measured doses instead of building a worldview out of it.

And now you know why, after the lab printout, questions come up with me that at first glance have nothing to do with your menstrual cycle.

Frequently asked questions

What does insulin have to do with my male hormones?

Insulin acts in two places at once. In cell cultures of theca cells from women with PCOS, insulin increased testosterone production, and blocking the insulin receptor abolished this effect. In parallel, insulin can throttle the production of SHBG in liver cell cultures, the transport protein for sex hormones. Less transport protein means a higher free fraction. Two levers, one cause. Important: both of these findings come from cell culture, not from the living human body.

Why is my SHBG low, and what does that say about me?

SHBG is made in the liver, and the liver adjusts production to the metabolic situation. Sugar and fructose can drive new fat production, a transcription factor called HNF-4alpha can drop along with it, and SHBG production can drop too. This has been described in liver cell cultures. A low SHBG is therefore rarely an isolated hormone problem. It is usually a metabolic signal. In the Women's Health Study, the risk of type 2 diabetes in the highest SHBG quartile carried an odds ratio of 0.09 compared with the lowest. Those women, however, were postmenopausal and did not have PCOS.

Can I have PCOS and insulin resistance even though I am lean?

Yes. Using the euglycaemic clamp, the demanding reference method, an Australian research group found insulin resistance in 75 percent of lean and 95 percent of overweight women with PCOS. Back in 1989, a study using the same method showed that the reduced insulin action existed independently of body weight. Weight amplifies something that can be there without it.

Which test shows insulin resistance in PCOS best?

Strictly speaking, none of the routine tests can. The reference method is the clamp, and it only happens in studies. What the 2023 international guideline recommends for glycaemic status is the 75 gram oral glucose tolerance test, regardless of body weight. Fasting glucose and HbA1c are explicitly described there as substitutes with markedly reduced accuracy. Fasting insulin and the HOMA index are not a diagnostic tool, at most they offer orientation.

At what HOMA index does insulin resistance start?

There is no cut-off that holds across all laboratories. In the Gutenberg Health Study, the median in the strictly selected reference group was 1.09 and the 97.5th percentile was 2.35. In the same project, three insulin immunoassays were compared, they correlated unsatisfactorily and produced systematically different HOMA values. The authors therefore call for assay-specific reference ranges. A HOMA of 2.6 from lab A and one from lab B are not the same thing.

Why does my gynaecologist say the HOMA index is not needed?

Because she is quoting the German S2k guideline of 2025 correctly. Recommendation 1.3.2 states, with full consensus strength, that parameters of insulin resistance should not be measured outside of studies. Three understandable reasons sit behind this: measurement methods that are not harmonised, no validated cut-off, and no compelling therapeutic consequence, because the same guideline permits metformin even without prior insulin testing. That is not indifference, it is consistency.

Is a fasting blood glucose enough to rule out diabetes in PCOS?

No. In a prospective study of 254 women with PCOS, criteria based on fasting glucose found 3.2 percent with diabetes, while criteria using the glucose tolerance test found 7.5 percent. In a 2024 meta-analysis with 2628 women, a fasting value from 7.0 mmol per litre reached a pooled sensitivity of 58.14 percent against the glucose tolerance test. So an unremarkable fasting value rules out little.

How often should I have my glucose metabolism checked if I have PCOS?

Both guidelines give the same interval: at first diagnosis and then every one to three years depending on the individual risk profile. The German S2k guideline 2025 allows the glucose tolerance test, HbA1c and fasting glucose side by side, while the 2023 international guideline puts the glucose tolerance test clearly first. When you are trying to conceive and during pregnancy, separate and tighter rules apply. Your doctor sets the rhythm that fits you.

What is the point of measuring insulin during an OGTT as well?

It shows how much insulin your body has to mobilise to keep glucose in range. That explains a constellation many women know: glucose values fine, symptoms still there. The limitation is the same as for fasting insulin. The assays are not harmonised, there are no validated cut-offs for the curve, and the German guideline explicitly counts this among the parameters that should not be measured outside of studies. For me it is orientation, not proof.

Is the triglyceride to HDL ratio a useful pointer?

As a rough pointer yes, and it sits in almost every standard lipid panel. In a cross-sectional study of 71 women with PCOS, the ratio of triglycerides to HDL cholesterol reached an AUROC above 0.80 against a true measure of insulin action, with an optimal cut-point of 1.3. In the women without PCOS, the best cut-point was 1.9. For context: that comparison group consisted exclusively of premenopausal women with overweight or obesity, so the cut-point of 1.9 cannot be transferred to every woman. Both cut-points are an orientation and not a diagnostic value. The big advantage: both values are standardised between laboratories, unlike insulin.

Is a glucose sensor useful in PCOS?

A sensor shows you something. It treats nothing. Robust intervention data showing that a sensor changes hormone values or the menstrual cycle in PCOS barely exist. What it can do is make patterns visible. A helpful frame of reference comes from a study of 153 metabolically healthy people: they spent a median of 96 percent of the time between 70 and 140 mg per decilitre and roughly half an hour a day above that. So a single spike is not a finding.

How much weight loss does it take for something to move in the lab?

The 2011 Cochrane review summarised six randomised studies with 164 participants. Under lifestyle treatment, weight fell by 3.47 kilograms, fasting insulin by 2.02 microunits per millilitre, waist circumference by 1.95 centimetres and total testosterone by 0.27 nanomoles per litre. For BMI, free androgen index, SHBG, glucose and cholesterol, no evidence of an effect was found. So the effect is real and modest at the same time. Important: these numbers are study results, not a target you have to reach.

What helps more, strength training or endurance?

The German S2k guideline 2025 recommends both, endurance and strength training at an individually adapted intensity. The meta-analyses on the effect on the HOMA index do not agree. Two analyses found reductions of 0.57 and 36.2 percent respectively, a third rated its own evidence quality as very low, and a fourth found no significant effect for either interval or moderate continuous training. What probably counts most is what you enjoy doing regularly.

What role does metformin play in PCOS according to the guidelines?

Some context first: metformin is a prescription medicine, in PCOS it is used off label in several situations, and it is therefore not something to decide about on the internet. What follows only describes how the guidelines discuss the topic. Whether it is an option for you is a medical decision after examination.

The German S2k guideline 2025 names fields of use such as supporting weight reduction from a BMI of 25, diabetes prevention, hyperandrogenaemia and cycle regulation, while explicitly pointing out the lower efficacy compared with alternatives. And it records that basic measures can achieve comparable results. The most common side effects are gastrointestinal, which is why the guideline provides for a slow dose titration according to the product information.

Further points are checked before any prescription. Metformin is not licensed in markedly reduced kidney function, because in rare cases lactic acidosis can develop. It is usually paused before examinations with iodinated contrast media and before surgery. Particular caution applies in liver disease, with regular alcohol consumption and in acute illness involving oxygen deficiency. With longer use vitamin B12 levels can fall, which is why they are usually monitored. Which of these points apply to you is clarified solely by your doctor together with you.

Where to read on

This article covers one section. Depending on where you stand right now, one of these places is a sensible next step.

If you have only recently had the diagnosis

Understanding PCOS: causes and symptoms

The basics of the condition, the criteria and the differential diagnosis. This article picks up exactly there.

If you want to understand the axis in general

Insulin resistance and hormones in women

The connection between blood sugar and hormones, independent of a particular diagnosis. The broader frame around this text.

If you want to know which test makes sense when

Testing hormones: which test, which timing

Cycle day, preparation and informative value of the common hormone tests. Practical if you are planning a blood draw.

If you want to change something at the table tomorrow

Avoiding blood sugar spikes

Order, combination and timing of meals. The concrete everyday part that this article deliberately left out.

If you want to see your own curves

14 days with a glucose sensor

How a sensor phase runs, what it can show and where its limit lies. Observation instead of diagnosis.

If the kilos stay despite everything

Losing weight with insulin resistance

Why weight responds differently when insulin action is disturbed, and which levers have been studied.

If the skin is the loudest symptom

Hormonal acne from the inside

Free testosterone, sebaceous gland and inflammation. The most visible endpoint of the chain described in this article.

If chronic stress is the second engine

Cortisol, stress and female hormones

How lasting strain moves the glucose and hormone balance along with it. The player most often overlooked in PCOS.

If you snore or never feel awake during the day

Recognising sleep apnoea

Signs, diagnostics and treatment routes. Markedly more common in PCOS than in the general population.

If the thyroid is on the table in parallel

Thyroid and female hormones

Why an underactive thyroid can create a similar picture and why it belongs in every initial workup.

If you want to know who else moves the weight

Leptin and insulin

The satiety and storage signals and their interplay. Adds the second half to the insulin perspective.

If eating and control have become exhausting

Understanding eating disorders

When mindfulness turns into pressure and where help begins. Important before nutrition topics create additional stress.

If the diagnosis is not settled yet

Diagnosing PCOS: Rotterdam and the phenotypes

The Rotterdam criteria, the four phenotypes and the exclusion work that often gets skipped.

If inositol is on the table

Inositol for PCOS

What meta-analyses and the 2023 guideline say about myo and D chiro inositol, with numbers instead of promises.

If the values look normal

SHBG: the overlooked value

The transport protein that explains why two women with the same testosterone can experience very different things.

If shakiness and cravings come with it

Blood sugar and cortisol

The chain from blood sugar to cortisol to free hormones, and what reactive hypoglycaemia really means.

Shukri Jarmoukli

Physician · Area of focus by his own account: integrative medicine, not a specialist qualification · ViveCura Berlin

I work in Berlin with people whose standard values look unremarkable while their experience still does not match. With hormonal topics I therefore look at the axis of blood sugar, stress, sleep, inflammation and environment, and not only at the hormone profile. What I quote from studies is marked as a study result. What I observe in consultations is marked as an observation. That separation matters more to me than a smooth story.

ViveCura · Skalitzer Straße 137 · 10999 Berlin

One closing note, so the context is right

This article reflects the state of 25 August 2026. It describes relationships and study results and does not replace medical advice, diagnosis or treatment. If you are unsure whether any of it applies to you, discuss it with your doctor.

Sources

31 sources, each checked via PubMed and, where a DOI exists, cross-checked via Crossref. The German guideline was read in full text from the AWMF registry. The type of study is given in square brackets in each case, so that you can judge for yourself how strongly a statement is supported.

Teede HJ, Tay CT, Laven JJE et al. Recommendations From the 2023 International Evidence-based Guideline for the Assessment and Management of Polycystic Ovary Syndrome. J Clin Endocrinol Metab. 2023;108(10):2447-2469. PMID: 37580314 · DOI: 10.1210/clinem/dgad463 [Guideline]
German Society of Endocrinology. S2k guideline on the diagnosis and treatment of polycystic ovary syndrome (PCOS). AWMF registry number 089-004, short version 1.1, full version June 2025. [Guideline]
Nestler JE, Jakubowicz DJ, de Vargas AF, Brik C, Quintero N, Medina F. Insulin stimulates testosterone biosynthesis by human thecal cells from women with polycystic ovary syndrome by activating its own receptor and using inositolglycan mediators as the signal transduction system. J Clin Endocrinol Metab. 1998;83(6):2001-2005. PMID: 9626131 · DOI: 10.1210/jcem.83.6.4886 [In vitro]
Loukovaara M, Carson M, Adlercreutz H. Regulation of sex hormone-binding globulin secretion and gene expression by cycloheximide in vitro. J Steroid Biochem Mol Biol. 1995;54(3-4):141-146. PMID: 7662587 · DOI: 10.1016/0960-0760(95)00141-l [In vitro]
Pugeat M, Nader N, Hogeveen K, Raverot G, Dechaud H, Grenot C. Sex hormone-binding globulin gene expression in the liver: drugs and the metabolic syndrome. Mol Cell Endocrinol. 2010;316(1):53-59. PMID: 19786070 · DOI: 10.1016/j.mce.2009.09.020 [Mechanism Review]
Rosenfield RL, Ehrmann DA. The Pathogenesis of Polycystic Ovary Syndrome (PCOS): The Hypothesis of PCOS as Functional Ovarian Hyperandrogenism Revisited. Endocr Rev. 2016;37(5):467-520. PMID: 27459230 · DOI: 10.1210/er.2015-1104 [Mechanism Review]
Barnes RB. The pathogenesis of polycystic ovary syndrome: lessons from ovarian stimulation studies. J Endocrinol Invest. 1998;21(9):567-579. PMID: 9856411 · DOI: 10.1007/BF03350782 [Mechanism Review]
Dunaif A, Segal KR, Futterweit W, Dobrjansky A. Profound peripheral insulin resistance, independent of obesity, in polycystic ovary syndrome. Diabetes. 1989;38(9):1165-1174. PMID: 2670645 · DOI: 10.2337/diab.38.9.1165 [Cohort, n=48]
Stepto NK, Cassar S, Joham AE et al. Women with polycystic ovary syndrome have intrinsic insulin resistance on euglycaemic-hyperinsulaemic clamp. Hum Reprod. 2013;28(3):777-784. PMID: 23315061 · DOI: 10.1093/humrep/des463 [Cohort, n=73]
Dunaif A, Xia J, Book CB, Schenker E, Tang Z. Excessive insulin receptor serine phosphorylation in cultured fibroblasts and in skeletal muscle. J Clin Invest. 1995;96(2):801-810. PMID: 7635975 · DOI: 10.1172/JCI118126 [Cohort, human tissue]
Zhu S, Li Z, Hu C et al. Imaging-Based Body Fat Distribution in Polycystic Ovary Syndrome: A Systematic Review and Meta-Analysis. Front Endocrinol. 2021;12:697223. PMID: 34566888 · DOI: 10.3389/fendo.2021.697223 [Systematic Review]
Legro RS, Kunselman AR, Dodson WC, Dunaif A. Prevalence and predictors of risk for type 2 diabetes mellitus and impaired glucose tolerance in polycystic ovary syndrome. J Clin Endocrinol Metab. 1999;84(1):165-169. PMID: 9920077 · DOI: 10.1210/jcem.84.1.5393 [Cohort, n=254]
Celik C, Abali R, Bastu E, Tasdemir N, Tasdemir UG, Gul A. Assessment of impaired glucose tolerance prevalence with hemoglobin A1c and oral glucose tolerance test in 252 Turkish women with polycystic ovary syndrome. Hum Reprod. 2013;28(4):1062-1068. PMID: 23335611 · DOI: 10.1093/humrep/det002 [Cohort, n=252]
Belsti Y, Enticott J, Azumah R et al. Diagnostic accuracy of oral glucose tolerance tests, fasting plasma glucose and haemoglobin A1c for type 2 diabetes in women with polycystic ovary syndrome. Diabetes Metab Syndr. 2024;18(3):102970. PMID: 38442646 · DOI: 10.1016/j.dsx.2024.102970 [Systematic Review]
Matli B, Schulz A, Koeck T et al. Distribution of HOMA-IR in a population-based cohort and proposal for reference intervals. Clin Chem Lab Med. 2021;59(11):1844-1851. PMID: 34380182 · DOI: 10.1515/cclm-2021-0643 [Cohort, n=10,340]
Blum MR, Popat RA, Nagy A, Cataldo NA, McLaughlin TL. Using metabolic markers to identify insulin resistance in premenopausal women with and without polycystic ovary syndrome. J Endocrinol Invest. 2021;44(10):2123-2130. PMID: 33687700 · DOI: 10.1007/s40618-020-01430-2 [Cohort, n=242]
Ding EL, Song Y, Manson JE et al. Sex hormone-binding globulin and risk of type 2 diabetes in women and men. N Engl J Med. 2009;361(12):1152-1163. PMID: 19657112 · DOI: 10.1056/NEJMoa0804381 [Cohort, n=718]
Shah VN, DuBose SN, Li Z et al. Continuous Glucose Monitoring Profiles in Healthy Nondiabetic Participants: A Multicenter Prospective Study. J Clin Endocrinol Metab. 2019;104(10):4356-4364. PMID: 31127824 · DOI: 10.1210/jc.2018-02763 [Cohort, n=153]
Jakubowicz DJ, Nestler JE. 17 alpha-Hydroxyprogesterone responses to leuprolide and serum androgens in obese women with and without polycystic ovary syndrome after dietary weight loss. J Clin Endocrinol Metab. 1997;82(2):556-560. PMID: 9024253 · DOI: 10.1210/jcem.82.2.3753 [Human intervention, n=23]
Moran LJ, Hutchison SK, Norman RJ, Teede HJ. Lifestyle changes in women with polycystic ovary syndrome. Cochrane Database Syst Rev. 2011;(7):CD007506. PMID: 21735412 · DOI: 10.1002/14651858.CD007506.pub3 [Systematic Review]
Kite C, Lahart IM, Afzal I et al. Exercise, or exercise and diet for the management of polycystic ovary syndrome: a systematic review and meta-analysis. Syst Rev. 2019;8(1):51. PMID: 30755271 · DOI: 10.1186/s13643-019-0962-3 [Systematic Review]
Patten RK, Boyle RA, Moholdt T et al. Exercise Interventions in Polycystic Ovary Syndrome: A Systematic Review and Meta-Analysis. Front Physiol. 2020;11:606. PMID: 32733258 · DOI: 10.3389/fphys.2020.00606 [Systematic Review]
Santos IK, Nunes FASS, Queiros VS et al. Effect of high-intensity interval training on metabolic parameters in women with polycystic ovary syndrome. PLoS One. 2021;16(1):e0245023. PMID: 33465123 · DOI: 10.1371/journal.pone.0245023 [Systematic Review]
Richards CT, Meah VL, James PE, Rees DA, Lord RN. HIIT'ing or MISS'ing the Optimal Management of Polycystic Ovary Syndrome: A Systematic Review and Meta-Analysis of High- Versus Moderate-Intensity Exercise Prescription. Front Physiol. 2021;12:715881. PMID: 34483969 · DOI: 10.3389/fphys.2021.715881 [Systematic Review]
Juhász AE, Stubnya MP, Teutsch B et al. Ranking the dietary interventions by their effectiveness in the management of polycystic ovary syndrome: a systematic review and network meta-analysis. Reprod Health. 2024;21(1):28. PMID: 38388374 · DOI: 10.1186/s12978-024-01758-5 [Meta-analysis, k=19, n=727]
Donga E, van Dijk M, van Dijk JG et al. A single night of partial sleep deprivation induces insulin resistance in multiple metabolic pathways in healthy subjects. J Clin Endocrinol Metab. 2010;95(6):2963-2968. PMID: 20371664 · DOI: 10.1210/jc.2009-2430 [Human intervention, n=9]
Vgontzas AN, Legro RS, Bixler EO, Grayev A, Kales A, Chrousos GP. Polycystic ovary syndrome is associated with obstructive sleep apnea and daytime sleepiness: role of insulin resistance. J Clin Endocrinol Metab. 2001;86(2):517-520. PMID: 11158002 · DOI: 10.1210/jcem.86.2.7185 [Cohort, n=505]
Melin JM, Forslund M, Alesi SJ et al. Effects of different insulin sensitisers in the management of polycystic ovary syndrome: A systematic review and meta-analysis. Clin Endocrinol (Oxf). 2023;100(2):149-163. PMID: 37933831 · DOI: 10.1111/cen.14983 [Systematic Review]
Nestler JE, Jakubowicz DJ. Decreases in ovarian cytochrome P450c17 alpha activity and serum free testosterone after reduction of insulin secretion in polycystic ovary syndrome. N Engl J Med. 1996;335(9):617-623. PMID: 8687515 · DOI: 10.1056/NEJM199608293350902 [RCT, n=24]
Alesi S, Forslund M, Melin J et al. Efficacy and safety of anti-androgens in the management of polycystic ovary syndrome: a systematic review and meta-analysis of randomised controlled trials. EClinicalMedicine. 2023;63:102162. PMID: 37583655 · DOI: 10.1016/j.eclinm.2023.102162 [Systematic Review]
Fitz V, Graca S, Mahalingaiah S et al. Inositol for Polycystic Ovary Syndrome: A Systematic Review and Meta-analysis to Inform the 2023 Update of the International Evidence-based PCOS Guidelines. J Clin Endocrinol Metab. 2024;109(6):1630-1655. PMID: 38163998 · DOI: 10.1210/clinem/dgad762 [Systematic Review]

Transparency on the evidence: where the data are thin

The chain itself is shown at cell level, not in the living human body. The effect of insulin on the theca cell and on SHBG production comes from cell cultures. In humans the connection appears indirectly, for example through intervention studies in which falling insulin went together with falling androgens. That is a strong indication and not direct proof.
The HOMA index and fasting insulin have no solid measurement basis. Three insulin immunoassays correlated unsatisfactorily in a German cohort and produced systematically different HOMA values. There is therefore no cut-off that holds across all laboratories. Every number in this field is assay dependent.
The meta-analyses on training do not agree. Three analyses find an effect on the HOMA index, one finds none. One of them explicitly rates its own evidence quality as very low. Anyone quoting only the convenient numbers tells half the story.
The ranking of dietary patterns comes from small groups. 727 participants spread across ten interventions gives very small cells. SUCRA values are ranking probabilities and not effect sizes. A proven advantage of one pattern cannot be derived from this.
The glucose sensor is an observation tool without intervention evidence in PCOS. There are essentially no robust data showing that a sensor changes hormone values or the menstrual cycle in PCOS. The frame of reference from the study in metabolically healthy people also describes a group without PCOS.
The claim of hidden visceral fat in lean women affected is not covered. In the subgroup using MRI and CT as the most accurate methods, the difference from BMI-matched controls disappeared. The connection stays plausible, but with the best methods it is not confirmed.
Serine phosphorylation explains only one part. The pattern was found in about half of the women studied, not in all, and the triggering cause lay outside the receptor. It is a possible mechanism in a subgroup.
Transferability of the sleep and SHBG data. The sleep study included nine metabolically healthy people without PCOS. The large SHBG cohort consisted of postmenopausal women. Neither is the same starting point as a 28 year old with PCOS.
The metformin numbers from the 1990s are small and short. They show the direction of the chain well and the size of the everyday benefit poorly. The German guideline 2025 records alongside them that basic measures can achieve comparable results.
The environmental perspective rests on mechanisms, animal data and observation. Intervention studies in humans showing a change in PCOS through reduced exposure are missing. That is why no sentence in this text claims such a connection for the individual case.
What deliberately does not appear here. No dosing for you, no treatment protocol, no product recommendation and no advice to start, change or end an existing medication. That applies equally to the contraceptive pill, hormone preparations, anti-androgens, thyroid hormones and metformin. Nothing in any section implies that a gynaecological workup or a recommended treatment should be postponed. When you are trying to conceive, time is a real factor, and reproductive medicine advice is replaced by nothing in this text. What I describe from my consultations is marked as an observation and is not a study result.

PCOS and insulin resistance: the engine behind the androgens

What this article covers

The chain this is all about

The first lever sits in the ovary

The second lever sits in the liver

One signal, two directions, one result

And now the context that often gets lost

Why this affects lean women too

Where does it come from, if not from weight?

And what about visceral fat?

Visceral fat in PCOS: probable, but not as clear-cut as claimed

Why fasting glucose often does not see the problem

HbA1c is convenient and still coarse

What can be measured how, honestly sorted

The costliest blind spot: the HOMA index has no universal cut-off

Two values that are already on your printout

SHBG as a warning light

The sensor: a window, not a diagnosis

What the guidelines say, and why they differ

The international guideline 2023

On glycaemic status in PCOS

The German S2k guideline 2025

Three recommendations that belong together

The three reasons behind the no, written out

And now my view, clearly separated

Why I still look at a markedly raised fasting insulin

Three things I do not derive from an insulin value

What lowers insulin levels in studies

Weight change: real and modest

Movement: three meta-analyses in favour, one against

Two analyses find an effect, one finds none

Dietary patterns: a ranking from small groups

Sleep: the lever almost nobody has on the list

Metformin, honestly weighed up

Why it works on this axis at all

What the German guideline says about it, in full

Benefit and limitation in the same document

Why I also look at environment, sleep and inflammation

Four players pulling at the same axis

And the point I sometimes get sideways looks for

And on the relationship with gynaecology

Frequently asked questions

Where to read on

Understanding PCOS: causes and symptoms

Insulin resistance and hormones in women

Testing hormones: which test, which timing

Avoiding blood sugar spikes

14 days with a glucose sensor

Losing weight with insulin resistance

Hormonal acne from the inside

Cortisol, stress and female hormones

Recognising sleep apnoea

Thyroid and female hormones

Leptin and insulin

Understanding eating disorders

Diagnosing PCOS: Rotterdam and the phenotypes

Inositol for PCOS

SHBG: the overlooked value

Blood sugar and cortisol

Shukri Jarmoukli

Sources

Have questions or want to book an appointment?