Diagnosing PCOS properly: the Rotterdam criteria and the four phenotypes

You are lying on the couch, the ultrasound probe glides across your lower abdomen, and then comes the sentence. Your ovaries look polycystic. Probably PCOS.

You walk out, type the word into your phone and within five minutes you are reading about infertility, diabetes, facial hair and lifelong illness. At home you dig out your old reports. Somewhere it said that your testosterone was completely fine.

And that is exactly where the confusion begins.

Because PCOS is not an ultrasound diagnosis. It is a combination and exclusion diagnosis. A picture of your ovaries can be one building block, but it can neither make nor rule out this diagnosis. And depending on which two of the three features apply to you, you land in one of four groups that differ clearly in metabolic risk, fertility and the focus of care.

This article walks through the diagnostic process. Not through treatment, not through every cause, but through the question: what has to be measured, asked and excluded for this diagnosis to hold? And what makes it unusable even though everything was measured?

One word about this article up front: it runs almost entirely on human data. With this topic that is a stroke of luck and not a given. Where I talk about environmental factors at the end, the level of evidence is named explicitly, because it is weaker there.

PCOS is a combination diagnosis, not an ultrasound diagnosis

The name is the problem. Polycystic ovary syndrome sounds as if it were about cysts on the ovaries. As if you could count them and then know.

They are not cysts. They are immature follicles, so-called antral follicles, that have stalled at an early stage. Picture a bakery where twenty pieces of dough are meant to go into the oven every morning. In PCOS they are all lying on the tray, but the oven does not fire up. The dough is not ill. It is waiting.

This is why the international consensus process in Rotterdam recorded something back in 2003 that still holds and is still constantly overlooked: no single criterion is enough for the clinical diagnosis. Neither the androgen excess nor the ultrasound picture.

How big the gap between picture and diagnosis really is becomes clear in a meta-analysis of prevalence. And that number is the reason I put this section at the beginning.

Before we go deeper, one box belongs at the front. There are situations in which the calm, step-by-step work-up is not the right one. They are rare. But when they apply, time counts.

Why PCOS develops, what insulin, LH and fat tissue are doing in the background, and which symptoms belong to it, is a topic of its own. I have described it in detail elsewhere, in PCO syndrome: causes and symptoms. Here we stay with the question of how the diagnosis comes about.

And now you know why the first question after the ultrasound appointment is not how many follicles there were, but which two criteria were actually assessed.

The three Rotterdam criteria in detail

Three features, two have to apply. Sounds simple. The devil sits in every single one of them.

Criterion 1: infrequent or absent ovulation

The technical term is oligo-ovulation or anovulation. Translated: your ovary releases an egg rarely or not at all.

From the outside this usually shows up in the menstrual cycle. Very long gaps between periods, irregular gaps, or the period stays away for months. The guideline uses cycle lengths of more than 35 days and fewer than eight periods a year as the usual reference points in adults. Very short cycles under 21 days also count as abnormal.

What often gets lost: a regular period does not prove ovulation. There are cycles that bleed on time and still run without ovulation. This is why, with a regular cycle and a persisting suspicion, a progesterone value can be measured in the second half of the cycle to see whether a corpus luteum worked at all.

And the other way round: an absent period does not automatically mean PCOS. That is so important that I give it its own section further down.

Criterion 2: androgen excess, clinical or biochemical

Here sits the most common thinking error. Many women believe this criterion is a laboratory question. It is not. One of the two is enough.

Clinical means: you can see it. Increased body hair in a male distribution pattern, on the upper lip, chin, chest, midline of the abdomen, back or inner thighs. Along with persistent acne beyond the usual degree. And androgenetic hair loss, the thinning crown.

Biochemical means: it is measured. Raised free testosterone or a raised free androgen index. Sometimes androstenedione or DHEAS as well, although those two rather add supplementary information.

Why a normal testosterone rules nothing out

I get this question a lot. My testosterone is normal, can it still be PCOS?

Yes, for three reasons.

First, the clinical picture counts just as much. If you have visible hirsutism, the criterion is met, even if the laboratory stays unremarkable.

Second, two out of three are enough. There is a phenotype in which androgens stand out neither visibly nor measurably, phenotype D. More on that in a moment.

Third, total testosterone measures something other than you think. Most of the testosterone in the blood is bound to a transport protein, sex hormone binding globulin, SHBG for short. Picture SHBG as a car pool. Whoever sits in the car does not arrive at any cell. Only the free fraction gets out and can dock. If SHBG falls, the freely available fraction rises even though the total value stays the same. That is exactly what can happen in insulin resistance, because high insulin levels can dampen SHBG formation in the liver.

This is why free testosterone and the free androgen index, which sets testosterone in relation to SHBG, as a rule say more than the total value alone. How hormone testing in women is sensibly planned, I have described in Testing hormones in women.

Criterion 3: polycystic ovarian morphology

The term that spoiled the whole name. What is meant is an ultrasound picture with a great many small antral follicles, mostly between 2 and 9 millimetres, and often an enlarged ovarian volume.

These are not cysts in the medical sense. A cyst is a fluid-filled cavity that has to be treated or observed. These follicles are normal structures in unusual numbers. As a rule they do not cause pain, and normally they do not have to be removed. The distinction matters: alongside these follicles a woman can additionally have a true cyst, and that can indeed cause symptoms. If you have sudden, severe or one-sided lower abdominal pain, that is a reason to have it looked at promptly, independently of the PCOS finding.

And the threshold has shifted. It used to be at least 12 antral follicles per ovary. Today it is at least 20, when modern equipment is used. That is the question that occupies many women holding an old and a new report, which is why it comes up in detail in the next section.

One more thing matters: a transvaginal ultrasound is the method that makes this counting possible at all. In young women who have not yet had intercourse, or when a transvaginal examination is not an option for other reasons, counting follicles through the abdominal wall is not reliable enough. In adults, the laboratory route via AMH then moves to the foreground.

And now you know why a clean report should contain three things, not one: the cycle history, the androgen part clinically and in the laboratory, and the ultrasound picture or, as a substitute, AMH.

What the 2023 international guideline changed

In 2023 the authoritative international guideline on PCOS was published anew. It is the anchor of this article, and it shifted four things.

Change 1: AMH as an alternative to ultrasound, in adults

AMH stands for anti-Müllerian hormone. It is produced by the small, waiting follicles. Many waiting follicles mean a lot of AMH. In that sense AMH measures roughly the same thing as counting follicles on ultrasound, only through the blood.

This is exactly why the 2023 guideline allows AMH in adults as an alternative to the ultrasound definition of polycystic ovarian morphology. That is practical: no transvaginal examination needed, no device-dependent counting.

But the upgrade has limits, and they are precisely worded.

Why could no cut-off be recommended? There are two very concrete reasons, and both are independently documented.

Change 2: 12 follicles became 20

This change unsettles many women because it sounds as if the condition had become stricter. The opposite is the case.

The threshold was raised because the machines have got better. A modern endovaginal ultrasound device with a frequency band that includes 8 MHz simply displays more follicles than a device from the early 2000s. With the same old threshold, more and more healthy women would have been classed as abnormal. Raising it is a safeguard against overdiagnosis.

The current threshold is also described in the background section of a paper by a Chinese working group that examined women by the new criteria: the traditional Rotterdam definition with at least 12 antral follicles per ovary was questioned because of more sensitive equipment and the resulting risk of overdiagnosis, and the new international guidelines set the threshold at at least 20 antral follicles per ovary when endovaginal transducers with a frequency band including 8 MHz are used (Wang Y et al., Hum Reprod 2023;38(Suppl 2):ii69-ii79. PMID: 37982419 · DOI: 10.1093/humrep/dead191). The actual results of that paper concern DNA methylation in 70 women from one centre. That is basic research and not a diagnostic tool, and I quote here only its framing of the threshold question.

Change 3: stricter rules apply in adolescents

To me this is the most important point of the whole section, and it is rarely said out loud.

In the first years after the first period, irregular cycles are common and often completely normal. The axis between the brain and the ovary has to settle in. Ovaries that look polycystic are also common at this age without any illness behind them.

Change 4: associated risks have been upgraded

The fourth shift concerns not the diagnostic criteria but what happens afterwards. The 2023 guideline emphasises metabolic risk factors, cardiovascular disease, sleep apnoea and the very high prevalence of psychological features considerably more strongly than the preceding version from 2018.

That is why a good PCOS work-up does not end with the finding. It opens the question of blood sugar, blood pressure, blood lipids, sleep and mood. The link with glucose metabolism is a topic of its own, and I have set it out in Insulin resistance and female hormones.

And now you know why a report from 2019 and one from 2026 can sound different even though nothing about your body has changed.

The four phenotypes A to D and why they differ

Three criteria, of which two are enough. Arithmetically that gives four possible combinations. Those are exactly the phenotypes A to D, as they were named at an NIH conference in 2012.

And this is the point at which a diagnosis stops being a label and starts becoming a map.

Why this is more than sorting letters is shown by two independent review articles.

One detail from an Austrian analysis belongs here too, because it explains why the androgen part steers the diagnostic process so strongly: in 750 women with and without PCOS at the Medical University of Graz, hyperandrogenaemic women per se had a higher risk of insulin resistance than women with phenotypes without androgen excess and than healthy controls. The association held even when no further Rotterdam criterion was met (Borzan V et al., J Clin Med 2021;10(4):829. PMID: 33670546 · DOI: 10.3390/jcm10040829). The authors therefore recommend testing women with isolated hyperandrogenaemia for insulin resistance as well. [Cohort, retrospective, n=750]

And now you know why the question about the letter is not academic. It decides whether metabolism, the cycle or the skin is in the foreground.

The exclusion part that often gets skipped

PCOS is also an exclusion diagnosis. That stands in every consensus document and is nonetheless the part most likely to fall by the wayside in a busy consulting room.

The reason is understandable: the three criteria are visible and quickly checked. The exclusion part costs extra values, extra time and sometimes a second appointment. And yet it is exactly what decides whether the diagnosis is right.

Because a wrong allocation is not neutral. It sends a woman for years into a form of care that was never designed for her actual cause.

One more number on the thyroid, because it shows that this is not only about exclusion: in a meta-analysis of 40 observational studies with 6,045 women with PCOS and 4,527 controls, the odds ratio for the frequency of TPO antibodies was 2.03 (1.35 to 3.04) and for thyroglobulin antibodies 1.92 (1.23 to 3.01). In the subgroups matched for TSH and BMI the prevalence differences were no longer significant, but the anti-TPO levels remained raised, and for anti-TG the difference in the TSH-matched subgroup was borderline significant (Kwiatkowski J et al., Int J Mol Sci 2025;26(15):7525. PMID: 40806651 · DOI: 10.3390/ijms26157525). These are observational data, not causality. But they explain why the thyroid is not simply ticked off once when PCOS is suspected, but kept in mind. [Systematic Review and Meta-analysis, k=40]

And now you know why, with a report brought to me, the first thing I look for is whether TSH, prolactin and 17-OH progesterone appear anywhere. Not out of distrust. Because that is where the points are set.

What the pill does to the diagnostic process

This is the section I am asked about most often in my consulting room. And one on which I find little in depth online.

Many women have been on the pill since adolescence. It was often prescribed for exactly that reason: for irregular cycles, for acne, for hair growth. So for the symptoms that would later be the diagnostic criteria.

And then blood is taken while the pill is still being taken and testosterone is measured. The value is normal. Everyone breathes out.

Only that value proves nothing.

What still works in practice, even while you are taking it: the clinical signs remain visible and they count. The cycle history from before starting the pill counts too, so it is worth looking into old records or talking to your mother about the years after your first period. Ultrasound usually shows quieter ovaries on the pill and then contributes little. And AMH tends to be somewhat lower under hormonal contraception, which limits its meaning as well.

If stopping is already planned with your doctor, for example when trying to conceive or because you want to, then that is the natural moment to schedule the diagnostic work-up. What can happen physically has been described in Coming off the pill: what happens afterwards, and what the alternatives are in Hormone-free contraception compared. Neither text replaces a conversation, they prepare it.

And now you know why an old laboratory slip is sometimes worth more than a new one.

From the new name to the environment: why PCOS is more than an ovary problem

Perhaps you have already come across the term PMOS. In a report, in a professional article, in a support group. And you wondered whether you now have something different from before.

You do not. But the story behind it explains rather precisely why this whole article is necessary.

The new name: polyendocrine metabolic ovarian syndrome

What is interesting is where the pressure came from. Not from the ivory tower.

For the care of adolescents, the significance of the renaming was worked through separately. There the worldwide frequency in adolescents is given as around 6.3 percent, the consensus process is described with 56 organisations and more than 22,000 survey responses, and it is noted that the old name was experienced as confusing and stigmatising by adolescents, parents and professionals (Ng NBH et al., J Pediatr Adolesc Gynecol 2026, Epub 29 June 2026. PMID: 42372905 · DOI: 10.1016/j.jpag.2026.06.018). The percentage comes from a professional overview, not from an original survey, and should be read as an order of magnitude. [Overview, professional article]

One editorial draws the parallel I like best: it compares the renaming with the road from non-alcoholic fatty liver disease to metabolic dysfunction-associated steatotic liver disease. There too a misleading name had obscured the view of metabolism for decades (Dalamaga M, Metabol Open 2026;30:100479. PMID: 42327014 · DOI: 10.1016/j.metop.2026.100479). The same editorial states, as the author's own assessment, that the ovary-focused naming has contributed to delayed diagnosis in up to 70 percent of those affected. That is not a collection of primary data but one author's estimate, and I quote it explicitly as such. [Editorial, no primary data]

How long it takes to reach a diagnosis is, incidentally, well studied, and the numbers are sobering: in an online survey of 1,385 women with a reported PCOS diagnosis from North America, Europe and other world regions, 33.6 percent reported more than two years to diagnosis and 47.1 percent reported having consulted at least three professionals. 35.2 percent were satisfied with the diagnostic experience, and only 15.6 percent with the information they received (Gibson-Helm M, Teede H, Dunaif A, Dokras A. J Clin Endocrinol Metab 2017;102(2):604-612. PMID: 27906550 · DOI: 10.1210/jc.2016-2963). Caveat: recruitment ran through support-group websites, and that favours women with a difficult course. [Cohort, cross-sectional survey, n=1,385]

Why I also ask about the surroundings

Now comes the part for which many patients get this link from me.

When the gynaecological work-up is in place, when the criteria have been checked and the counterparts excluded, many women are left with one open question: why me? And why now?

There is no complete answer to that. There is a genetic component, there are influences from the time before birth, there is metabolism, weight, sleep, movement and stress. And there is a group of factors that rarely comes up in routine care: substances from the surroundings that can dock onto hormone receptors or change hormone metabolism.

Before I come to the data, I want to make two things clear, because otherwise this field derails quickly.

First: the fact that environmental factors rarely come up in a gynaecology appointment has understandable reasons. Training priorities are set differently, and the time budget in routine care is tight. The 2023 guideline does not list environmental exposure among its diagnostic recommendations, and that is methodologically grounded: intervention studies showing that reducing exposure improves clinical endpoints are missing. Without such data a GRADE process cannot issue a recommendation. That is restraint with a reason, not an oversight.

Second: I look anyway, because observational data in humans exist and because I can live better with an open question than with an overlooked one. But I say what kind of data these are.

Where chronic stress plays in is a chapter of its own. The axis between brain, adrenal gland and ovary reacts sensitively to sustained strain, and that can change cycle patterns which then feed into the diagnostic process. It is set out in detail in Cortisol, stress and female hormones.

And now you know why I ask about housing, damp, occupation and strain even though it is not in the guideline. Not because I know more. But because I would rather ask one question too many and interpret the answer honestly.

Which doctor, which values, which cycle day, what to bring

Finally the practical part. Because the best diagnostic work-up is the one that is complete on the first attempt.

Who makes the diagnosis

The gynaecology practice is the first address. The cycle history belongs there, along with the physical examination, the ultrasound and the basic laboratory work. This combination is the core, and nothing can replace it.

An endocrinological opinion makes sense when an exclusion value is abnormal, when androgens are very clearly raised, when Cushing's syndrome is suspected or when the picture simply does not add up. If you are trying to conceive, reproductive medicine joins in, and time counts there, which is why an early appointment makes sense and should not be postponed.

What I do in my practice comes after and alongside that, not before it. I look at metabolism, signs of inflammation, micronutrients, sleep, strain and surroundings. That complements the gynaecological work-up, it does not replace it.

Which values are usually measured

Which cycle day

For most values the early follicular phase applies, cycle day 3 to 5, counted from the first day of bleeding. LH, FSH, estradiol and the androgens are most comparable there, because the axis is not yet overwritten by a maturing follicle.

Progesterone is the exception. It is measured in the second half of the cycle, roughly a week before the expected period, because that is where it peaks.

AMH varies only slightly across the cycle and can therefore usually be measured on any cycle day. Under hormonal contraception the value tends to be lower, which then makes it harder to interpret.

And if the period does not come? Then measurement is done at any time after discussion with your doctor and the result is interpreted accordingly. A pregnancy test belongs in that situation before anything else is considered. In pregnancy and while breastfeeding these values have to be judged differently in any case, and the PCOS work-up is then usually postponed.

What you can bring to the appointment

And now you know what should be on your list before you walk into a consulting room next time.

Common questions about the PCOS diagnosis

Where PCOS diagnosis connects to the rest

A diagnosis never stands alone. It hangs on metabolism, on the thyroid, on contraception, on the skin and on the question of what surroundings a body lives in. Here are the roads that lead on from this article.