Two women can walk out of two clinics with the same three letters on the file and almost nothing in common physiologically. One has high androgens, irregular cycles and marked insulin resistance. The other has regular androgens, irregular cycles and polycystic ovaries on a scan.

Under the Rotterdam criteria, both have PCOS. Under the older NIH criteria, only one does. That difference has a name, phenotype, and it changes what is worth measuring and watching. What it does not yet change, on the evidence, is which diet either woman is given, and this piece is as much about that gap as about the classification itself.

• PCOS is diagnosed from at least two of three findings (androgen excess, ovulatory dysfunction, polycystic ovarian morphology), and which two a woman has defines her phenotype.
• The four phenotypes produce very different population distributions depending on where women are counted. Community surveys find phenotype C most common in India (around 40%). Fertility clinic populations find phenotype A dominant at roughly 68%.
• Insulin resistance is present across all four phenotypes, not only in the classic ones. Community surveys find even phenotype D carrying insulin resistance in around 38% of cases.
• No randomised dietary, lifestyle, or exercise trial in PCOS has stratified participants by phenotype and reported outcomes by phenotype. Phenotype-specific diet advice is clinical inference, not protocol.
• Lean PCOS and phenotype D are not the same thing. A lean woman can have any phenotype. A woman with phenotype D can carry obesity.
• For Indian women specifically, the word "lean" carries an extra complication: international BMI cut-offs of 25 kg/m² already represent obesity by South Asian criteria, so insulin resistance should be assessed directly rather than assumed absent from body weight.

PCOS is diagnosed from combinations of three findings, not from any one of them. The 2023 international guideline, endorsed by more than 30 professional societies, uses the revised consensus Rotterdam criteria and requires at least two of the following in adults, after other causes have been excluded:

• Hyperandrogenism: Androgen excess, either clinical (hirsutism, acne, androgenic hair loss) or biochemical on a blood test.
• Ovulatory dysfunction: Irregular cycles or biochemical evidence that ovulation is not happening.
• Polycystic ovarian morphology (PCOM): The ovarian appearance on ultrasound, or a raised anti-Mullerian hormone (AMH) level as an alternative marker.

Three points inside that definition are routinely misunderstood:

• Where a woman has both irregular cycles and hyperandrogenism, the guideline states that ultrasound and AMH are not required at all, because the diagnosis is already made on clinical grounds.
• The ultrasound threshold has moved: on modern high-frequency transducers, PCOM is now defined as 20 or more follicles per ovary in at least one ovary, well above the 12 or more follicles of 2 to 9 mm used when the Rotterdam criteria were written in 2003.
• AMH may substitute for ultrasound but is explicitly not a standalone test for PCOS, with laboratories required to use assay-specific cut-offs.

Adolescents are handled differently and more conservatively. Within eight years of menarche, both irregular cycles and hyperandrogenism are required, and ultrasound and AMH are not recommended for diagnosis because normal pubertal ovaries overlap too heavily with PCOM. A teenager with only one feature is classified as at risk of PCOS rather than given the diagnosis.

Two older rule sets still shape how PCOS is discussed. The NIH 1990 definition required androgen excess and chronic anovulation together and gave ultrasound no diagnostic role. The AE-PCOS 2006 definition made androgen excess mandatory and paired it with ovarian dysfunction, meaning either irregular ovulation or polycystic ovaries.

Because Rotterdam allows any two of the three findings, four combinations are possible. These are the phenotypes, labelled A to D:

Phenotype A includes androgen excess, ovulatory dysfunction and polycystic ovaries, with all three findings present.

Phenotype B includes androgen excess and ovulatory dysfunction, with polycystic ovaries absent.

Phenotype C includes androgen excess and polycystic ovaries, with ovulatory dysfunction absent and cycles broadly regular.

Phenotype D includes ovulatory dysfunction and polycystic ovaries, with androgen excess absent.

Phenotypes A and B are what the NIH definition called PCOS. Phenotypes C and D exist only under Rotterdam. The 2023 guideline does not itself use the A to D labels, though it accepts the architecture that produces them and distinguishes hyperandrogenic from normoandrogenic presentations.

In the Indian national community study published in JAMA Network Open in 2024, phenotype C accounted for 40.8% of PCOS cases, phenotype D for 24.6%, phenotype A for 20.2% and phenotype B for 14.3%. An Indian meta-analysis produced a similar spread, with C around 42% and D around 23%.

Indian clinic populations look nothing like that. In an Indian infertility cohort of 164 women, phenotype A accounted for 67.7% of cases, with B at 11%, C at 17.7% and D at 3.6%. A separate Indian series examining the effect of obesity found phenotype D dominant among lean women at 63.1%, with phenotype A dominant among women with obesity.

Where a woman was counted therefore changes what the average PCOS patient appears to look like. A fertility clinic sees mostly phenotype A and reasonably concludes that PCOS is a severe, hyperandrogenic, insulin-resistant condition. A community survey counting women who never sought care finds the ovulatory phenotype C most common.

Insulin resistance is present across all four phenotypes, which is the first thing to say, because it is the finding most often lost when phenotypes are discussed as though only the "classic" ones are metabolic. One broad review puts insulin resistance prevalence at around 80% in the classic phenotypes A and B, 65% in ovulatory phenotype C, and 38% in normoandrogenic phenotype D.

In a 2025 study of 226 women, mean HOMA-IR ran at 3.59 in phenotype A, 2.73 in D, 2.59 in B and 2.05 in C, against a threshold of about 2.1. The ordering is not stable between studies. A 2023 study of 160 women using a HOMA-IR cut-off of 2.5 found the highest insulin resistance prevalence in phenotype B at 91.9%, ahead of phenotype A at 74.7%, with C and D both around 57%. Androgen indices are more consistent: phenotype A shows the highest total testosterone, free androgen index and AMH across multiple cohorts, and phenotype D the lowest, which follows from how D is defined.

Metabolic outcomes are where the phenotype story weakens. The Indian infertility cohort found phenotype A carrying higher BMI, fasting insulin, HOMA-IR and a more deranged lipid profile than the other three. A separate Indian phenotypic study found metabolic syndrome and insulin resistance prevalence broadly similar across all four phenotypes despite differences in anthropometry and AMH.

A UK Biobank analysis found hyperandrogenic PCOS associated with greater hepatic fat and fibroinflammation, while normoandrogenic PCOS showed higher incident all-cause cardiovascular disease, the opposite of the expected direction. Indian phenotype-stratified data on fatty liver, metabolic syndrome and cardiovascular incidence remain limited or unpublished as of 2026.

Read together, that conflict says phenotype is a weak predictor of cardiometabolic risk on its own, and that adiposity, particularly visceral fat, is the stronger determinant. DXA-based work found visceral adipose tissue correlating with HOMA-IR across all phenotypes, not within one of them.

Psychological burden looks similarly phenotype-independent. A case-control study found depression scores raised in roughly 46% and anxiety in about a third of women with PCOS, at similar rates in NIH and non-NIH phenotypes. The 2023 guideline recommends screening every woman with PCOS for both, with no phenotype distinction.

These two categories are routinely conflated and they are not the same thing. Lean PCOS describes body weight: a woman who meets diagnostic criteria at a normal or low BMI. Phenotype D describes hormones: irregular cycles and polycystic ovaries with no androgen excess. A lean woman can have any of the four phenotypes, and a woman with phenotype D can carry obesity.

The confusion matters because lean PCOS is not metabolically benign. An evidence-based lean PCOS review reported insulin resistance in roughly 83% to 93% of lean women with PCOS using surrogate indices, and significantly greater insulin resistance than in BMI-matched women without PCOS. Clamp studies disagree on the detail, with some finding lean PCOS no more insulin resistant than lean controls but showing higher basal insulin, and systematic reviews estimating insulin sensitivity reduced by about 27% in PCOS overall, independent of BMI.

For Indian women, there is a further problem with the word "lean". Most of the international lean PCOS literature uses a BMI cut-off of 25 kg/m², which is the threshold at which Indian and South Asian criteria already classify a woman as having obesity. Asian-Indian bodies carry more visceral fat and greater cardiometabolic risk at any given BMI, so a woman labelled lean in an international paper may sit above the Indian cut-off.

Here is the finding that most PCOS content will not tell you. No randomised dietary, lifestyle or exercise trial in PCOS has stratified or randomised participants by phenotype and reported outcomes by phenotype. Existing trials compare women with PCOS against controls, or test one diet against another within PCOS as a single pooled group. The same holds for medication: metformin, inositol, combined oral contraceptives, anti-androgens, letrozole and clomiphene have not been tested in trials designed around phenotype arms with prespecified phenotype-based analyses.

That absence has a direct consequence. Any statement of the form "phenotype A needs aggressive weight loss first" or "phenotype D responds better to X" is extrapolated from cross-sectional differences in metabolic severity, not from a trial that tested it. It is clinical inference. Inference is not worthless, and it is what experienced clinicians reason with when trials are missing, but it should be labelled as what it is rather than dressed up as protocol.

The 2023 guideline is consistent with that gap. It recommends lifestyle intervention for all women with PCOS without phenotype-specific instruction, and states explicitly that no single dietary pattern or exercise modality is recommended over another, with emphasis instead on a sustained energy deficit where weight loss is indicated, on weight gain prevention, and on minimising weight stigma. Indian practice documents add a caution against unnecessary calorie restriction in lean PCOS, targeting weight maintenance and metabolic health instead.

If the trials have not stratified, a fair question is what the classification is for. It informs surveillance and expectation-setting more than it selects a treatment, and the treatment decisions that do get made are driven by metabolic risk, symptoms and fertility intent rather than by the letter.

Current information treats PCOS largely as one entity where medicines are concerned. The guidance below reports indication and limits only. None of it is a reason for any reader to start, stop or change a medicine, and all of it is a prescriber's decision after examination.

Metformin

Metformin is recommended primarily for metabolic and anthropometric features in adults with BMI 25 kg/m² or above, with conditional consideration below that and in adolescents. The guideline grades the underlying evidence low to very low certainty and does not differentiate by phenotype. Indian guidance positions it for PCOS with type 2 diabetes or impaired glucose tolerance where lifestyle change has not succeeded.

Inositol

Evidence for inositol is described as inconsistent and of very low certainty. No specific preparation, dose or combination can currently be recommended, and no trial has tested whether any phenotype responds differently.

Combined oral contraceptives

Combined oral contraceptives are a first-line pharmacological option in adults for menstrual irregularity and hyperandrogenism, favouring lower-dose preparations and those carrying lower thrombotic risk. Contraindication screening against standard medical eligibility criteria comes first. No formulation is recommended by phenotype.

Anti-androgens (spironolactone, cyproterone acetate)

Anti-androgens are second-line or adjunctive for hirsutism, where contraceptives and cosmetic measures are insufficient or contraindicated. Teratogenic risk makes contraception a requirement of use, not an optional extra.

Letrozole and clomiphene

Letrozole is first-line ovulation induction for anovulatory infertility in PCOS with no other infertility factor. Clomiphene is preferred over metformin where letrozole is unavailable. Observational cohorts find phenotype A more clomiphene-resistant and more prone to ovarian hyperstimulation, but no trial has randomised ovulation induction by phenotype.

GLP-1 receptor agonists

Evidence for GLP-1 receptor agonists in PCOS is limited. Where discussed at all, it is as a later-line option in high metabolic risk after lifestyle and metformin, with pre-conception cautions. Guidance advises discontinuing before conception and using effective contraception during treatment. Nothing in current guidance selects them by phenotype, and the general restrictions on GLP-1 medication are set out separately.

Where guidance is genuinely uniform is in surveillance, and that uniformity is itself informative. An oral glucose tolerance test is recommended at diagnosis for adults and adolescents with PCOS, with glycaemic assessment repeated every one to three years by individual risk. Lipid profile and blood pressure screening are recommended for every woman with PCOS, with no phenotype-specific algorithm, because PCOS as a whole is treated as a cardiovascular risk-enhancing condition.

Three misuses come up often enough to name.

The first is treating the label as fixed for life. Phenotype describes a snapshot, and expression shifts with age and weight. One reproductive ageing study found that at least half of women with anovulatory PCOS become ovulatory by late reproductive age, which would move them out of phenotypes A, B or D. Longitudinal data tracking A to D over decades are scarce, so how stable the categories are over a lifetime is genuinely unknown.

The second is treating phenotype D as mild PCOS. D has the lowest androgen levels by definition and often the mildest insulin resistance profile, but insulin resistance prevalence in D still runs at roughly half in some cohorts, and the UK Biobank signal for incident cardiovascular disease pointed toward normoandrogenic rather than hyperandrogenic PCOS. Mild on one axis is not mild overall.

The third is applying ethnicity-blind thresholds. The modified Ferriman-Gallwey score used to grade hirsutism is applied at a threshold of 4 to 6 depending on ethnicity, with no consensus on a South Asian-specific cut-off, so clinical hyperandrogenism in Indian women can be scored inconsistently. The follicle-count threshold for polycystic ovaries has been revised upward as ultrasound equipment improved, and remains debated. A phenotype assigned on contested thresholds is a working label, not a fixed fact.

The value of asking which phenotype you have is not that it selects a diet. On current evidence it cannot, and anyone selling a phenotype-specific PCOS diet is selling inference.

Its value is that it tells you which findings were actually assessed in your case, whether androgens were measured or assumed, and what a clinician should be watching over the following years. One woman needs her androgens and cycles followed. Another needs her insulin resistance and visceral fat followed. Both need glucose, lipids and blood pressure checked regardless of which phenotype they carry.

Not sure which phenotype applies to your case, or whether all three components were ever assessed? A Voy clinician can review your existing results and tell you what the findings mean together. Book an assessment.

This article is for general information and education only and is not medical advice, diagnosis, or treatment. GLP-1 and other medications referenced are prescription-only and are appropriate only for certain people under the supervision of a qualified clinician. Do not start, stop, or change any medication based on this article. Please consult a registered medical practitioner about your individual circumstances. Information reflects what was available at the time of review and may change.