PCOS Diagnostic Criteria: The Rules Your Doctor Is Actually Using

First, the name

In May 2026 a global consensus process involving 56 academic, clinical and patient organisations renamed the condition polyendocrine metabolic ovarian syndrome, or PMOS, in The Lancet. The biology, the criteria and the treatment are all unchanged — the label moved, and nothing under it did. If you want that story in full, read why PCOS is now PMOS. This page uses “PCOS” because that is still what appears on most referral letters.

Why there are three sets of criteria and not one

Most conditions have one definition. PCOS has three, written 13 years apart, and all three are still cited in clinics today.

NIH 1990 required both androgen excess and chronic anovulation. Two features, both mandatory.

Rotterdam 2003 widened it to any two of three, adding ovarian appearance on ultrasound as a third qualifying feature and explicitly requiring the exclusion of other causes. That single change created two new patient groups: women who ovulate but have androgen excess and polycystic ovaries, and women with irregular cycles and polycystic ovaries but no androgen excess.

AE-PCOS 2006 pushed back. The Androgen Excess Society task force argued that PCOS should first be considered a disorder of androgen excess, so androgen excess became mandatory again, plus one of the other two features.

The disagreement is not academic. A 2016 systematic review and meta-analysis of 24 eligible studies measured how much the definition changes the answer: 6% of women met NIH criteria, 10% met Rotterdam, and 10% met AE-PCOS.

Criteria set	What it requires	Prevalence found (95% CI)	Who it leaves out
NIH 1990	Androgen excess and ovulatory dysfunction — both	6% (5–8%)	Anyone with regular cycles; anyone without androgen excess
Rotterdam 2003	Any two of three: androgen excess, ovulatory dysfunction, polycystic ovarian morphology	10% (8–13%)	Widest net of the three
AE-PCOS 2006	Androgen excess plus one of the other two	10% (7–13%)	The non-androgenic pattern — irregular cycles plus polycystic ovaries only
2023 international guideline (current)	Rotterdam, with AMH permitted instead of ultrasound in adults	10–13% globally	Adolescents are held to a stricter bar (below)

What the current criteria say

The 2023 international guideline — the largest evidence review in this area — kept the Rotterdam structure and made one substantive change. Two of three features are still needed. Other causes must still be excluded. But anti-Müllerian hormone (AMH) can now replace the ultrasound in adults.

It also wrote down a shortcut that saves a lot of appointments: if you have irregular cycles and androgen excess, you do not need an ultrasound or an AMH test. Those two features alone satisfy the two-of-three rule. Any scan after that adds cost, not information.

Feature	What is measured	Threshold that counts
Irregular cycles — 3+ years post-menarche	Cycle length	Shorter than 21 days, longer than 35 days, or fewer than 8 cycles a year
Irregular cycles — 1 to 3 years post-menarche	Cycle length	Shorter than 21 days or longer than 45 days
Irregular cycles — any single cycle, 1+ year post-menarche	Cycle length	Longer than 90 days
Primary amenorrhoea	No period ever	By age 15, or more than 3 years after breast development began
Clinical androgen excess	Hirsutism, scored on the modified Ferriman-Gallwey scale	mFG of 4–6 or above, varying by ethnicity
Biochemical androgen excess	Total and free testosterone; free testosterone by calculated free androgen index	Above the laboratory reference range, measured by LC-MS/MS rather than direct immunoassay
Polycystic ovarian morphology — ultrasound	Follicle number per ovary	20 or more in at least one ovary; or ovarian volume of 10 mL or more where equipment or image quality limits counting
Polycystic ovarian morphology — bloods	Serum AMH	No international cut-off exists; laboratories must use population- and assay-specific values

Note the first row of that table carefully. “Irregular” has a numeric definition, and a 33-day cycle does not meet it. Plenty of people are told their cycles are irregular when they are not, and plenty are told 40-day cycles are fine when they are not.

What has to be ruled out before the label goes on

Every version of the criteria — 1990, 2003, 2006 and 2023 — requires exclusion of other causes. This is the step most often skipped, and skipping it is how a thyroid problem gets managed as PCOS for three years.

The guideline’s diagnostic algorithm names four baseline tests: TSH, prolactin, 17-OH progesterone and FSH. Further testing is added where the clinical picture calls for it — Cushing’s syndrome, adrenal or ovarian androgen-secreting tumours, and hypogonadotrophic hypogonadism from low body fat or heavy training, assessed with LH and FSH.

The 2023 change that matters most: AMH instead of ultrasound

This is the one genuinely new thing in the current criteria, and it came from a meta-analysis of 82 studies run for the guideline itself.

Three practical consequences. AMH is never a standalone PCOS test. You should have either an ultrasound or an AMH — not both, because doing both increases over-diagnosis. And a blood draw is cheaper and less invasive than a transvaginal scan, which for many people is the point.

If you are under 20, the bar is deliberately higher

Adolescent criteria are stricter, because normal puberty looks like PCOS. Multi-follicular ovaries and irregular cycles are both ordinary features of the years after menarche.

So for adolescents, both androgen excess and ovulatory dysfunction are required — two of three is not enough. Pelvic ultrasound is not recommended for diagnosis within 8 years of menarche, and AMH is not recommended at all. Where only one feature is present, the guideline suggests an “at risk” label with symptomatic treatment and re-evaluation, rather than a diagnosis that may not hold.

Which criteria set finds which phenotype

Rotterdam’s two-of-three rule produces four combinations, usually labelled A to D. They are not equally recognised by the three criteria sets, which is the practical reason the definitions matter.

• A — androgen excess + ovulatory dysfunction + polycystic ovaries. Recognised by all three sets.
• B — androgen excess + ovulatory dysfunction, normal ovaries. Recognised by all three.
• C — androgen excess + polycystic ovaries, cycles regular. Rotterdam and AE-PCOS only.
• D — ovulatory dysfunction + polycystic ovaries, no androgen excess. Rotterdam only.

If you sit in group D, an NIH-trained or AE-PCOS-trained clinician will not diagnose you. If you sit in group C, NIH criteria will not either. These are the two groups most likely to spend years being told nothing is wrong — and in an observational survey of 1,385 women with a reported PCOS diagnosis, 33.6% said it took more than two years and 47.1% saw three or more health professionals before the diagnosis was made.

Phenotype is not the same thing as the driver of your symptoms. For that, see the four types of PCOS. For the criteria under the new name, see PMOS diagnosis criteria, and for the history of the 2003 consensus itself, the Rotterdam criteria. The rest of the diagnosis section picks up from here.

What to do this week

Start a cycle log with actual dates, not impressions — first day of bleeding, every time, for three months. Table 2 turns that log into a yes or no on one of the three criteria, and it is the single piece of evidence no clinician can generate for you.

Then, at your next appointment, ask three questions: which criteria set was used, whether TSH, prolactin, 17-OH progesterone and FSH were checked, and whether an ultrasound is actually needed given what is already documented.