If one doctor has told you that you have PCOS and another has told you that you do not, the disagreement may not be about your body. India's first nationally representative study of polycystic ovary syndrome screened 8,993 women and produced two different prevalence figures from the same population: 7.2% under one diagnostic rule set and 19.6% under another.

Both are correct. They are answers to different questions. What follows is where those national numbers come from, how the women were found and examined, and why the definitions disagree with each other.

• In the ICMR PCOS Task Force community study, national prevalence was 7.2% by NIH 1990 criteria and 19.6% by Rotterdam 2003 criteria, in the same 8,993 women.
• A third rule set, the AE-PCOS 2006 criteria, put the same population at 13.6%.
• Recruitment was house-to-house from voter rolls across five zones and eight states, not from clinics, which makes it the first Indian PCOS figure that describes a population rather than a waiting room.
• Every ultrasound in the study was transabdominal, a measurement choice likely to over-count polycystic ovarian morphology against current diagnostic thresholds.
• The widely circulated "one in five Indian women" line does not originate in this study.

The study is "Prevalence, Phenotypes, and Comorbidities of Polycystic Ovary Syndrome Among Indian Women" by Ganie and colleagues, published in JAMA Network Open in October 2024. It was commissioned under the Indian Council of Medical Research PCOS Task Force and designed as a nationwide, multicentre, community-based cross-sectional study.

Recruitment ran from November 2018 to July 2022. Sites were drawn using a multistage cluster design built on 2011 Census voter identification lists. At each participating institution, one rural and one urban Vidhan Sabha constituency was selected, then 8 to 10 polling booths were randomly sampled, and every woman aged 18 to 40 who had lived there at least a year and appeared on those rolls was invited.

Across five zones and eight states, 12,100 women were approached. Of those, 831 refused and 2,276 were ineligible, leaving 8,993 who completed the screening questionnaire, a participation rate of roughly 92% among eligible women. Mean age was 29.5 years.

The recruitment method is the part most worth holding on to. Trained staff conducted house-to-house, face-to-face interviews using a validated questionnaire. Nobody had to have a symptom, a referral or a reason to see a doctor in order to be counted. Most Indian PCOS figures in circulation come from clinics, colleges or single cities, where the women being counted are already there for a reason, and a study of women who sought care cannot describe women who did not.

Screening sorted women into three groups: 196 already carried a documented PCOS diagnosis, 2,251 screened positive on the questionnaire, and 6,546 screened negative. Screen-positive meant reporting oligomenorrhoea, secondary amenorrhoea or clinical signs of androgen excess. All questionnaire-positive women, plus a 20% random sample of questionnaire-negative women, were then invited for full evaluation with hirsutism, acne and alopecia scoring, biochemical androgen assays and pelvic ultrasound.

The study defined each of the four diagnostic components as follows:

Clinical androgen excess

Clinical androgen excess was defined as a modified Ferriman-Gallwey score above 8, and/or grade 3 acne on the Leeds scale, and/or grade 2 to 3 androgenic alopecia on the Ludwig scale.

Biochemical androgen excess

Biochemical androgen excess was defined as total testosterone at or above 0.88 ng/mL and/or DHEA-S at or above 246 ng/mL, assayed centrally by electrochemiluminescence immunoassay.

Ovulatory dysfunction

Ovulatory dysfunction was defined as fewer than eight cycles a year or an interval longer than 35 days. Secondary amenorrhoea was defined as a cycle longer than 90 days.

Polycystic ovarian morphology

Polycystic ovarian morphology was defined as 12 or more peripheral follicles of 2 to 8 mm and/or an ovarian volume above 10 cm³ in at least one ovary, measured using transabdominal ultrasound.

Two details in these diagnostic definitions carry more weight than they might appear to:

• First, the ultrasound was transabdominal throughout, with no transvaginal scanning. This is likely to inflate the count of polycystic ovaries, because current diagnostic guidance sets a considerably higher follicle threshold when high-resolution transvaginal imaging is used.
• Second, women who had only one of the three components (isolated androgen excess, isolated irregular cycles, or isolated polycystic ovaries) were not counted as PCOS at all. The study labelled them "pre-PCOS", a category with no treatment implication attached to it here.

None of these thresholds is a self-test. They are the operating definitions a research team applied to a population under standardised conditions, and a diagnosis for any individual woman remains an act performed by a registered medical practitioner after examination.

Three published rule sets exist for diagnosing PCOS, and they combine the same three findings differently. Applied to the same 8,993 women, they gave three different national estimates:

NIH 1990 criteria

The NIH 1990 criteria require both androgen excess and irregular ovulation for a PCOS diagnosis. Using this stricter definition, the national study found a PCOS prevalence of 7.2%, with a 95% confidence interval of 4.8% to 10.8%.

AE-PCOS 2006 criteria

The AE-PCOS 2006 criteria require androgen excess plus either irregular ovulation or polycystic ovaries. When these criteria were applied to the same population, PCOS prevalence increased to 13.6%, with a 95% confidence interval of 8.4% to 21.6%.

Rotterdam 2003 criteria

The Rotterdam 2003 criteria require any two of three findings: androgen excess, irregular ovulation, or polycystic ovaries. Because this definition allows more combinations of findings to qualify, it produced the highest prevalence in the study at 19.6%, with a 95% confidence interval of 12.7% to 29.2%.

The women who sit in the gap between 7.2% and 19.6% are identifiable. They are the ones with polycystic ovaries and irregular cycles but no measurable androgen excess, and some with androgen excess and polycystic ovaries but broadly regular cycles. Rotterdam counts both groups. NIH counts neither, because it requires androgen excess and irregular ovulation together.

So a woman can genuinely meet the definition of PCOS in one consulting room and fail it in another, without either clinician being careless. The confidence intervals are worth noticing too. The Rotterdam estimate spans 12.7% to 29.2%, which is wide, and reflects how much prevalence varied between the sampled clusters.

Published commentary does not settle this, and the disagreement is real rather than a matter of one camp being out of date. Clinical endocrinology reviews argue that the NIH-type presentations carry the highest metabolic risk, and that admitting polycystic ovarian morphology alone as a qualifying feature risks labelling women with milder metabolic profiles in a clinical setting.

Guideline groups and global prevalence reviews take the other position, endorsing modified Rotterdam criteria as the standard for public health surveillance precisely because they capture a wider spectrum of women with reproductive or metabolic vulnerability.

Both positions are defensible because they answer different questions. For national planning, meaning how many gynaecology and endocrinology services a state needs to fund, 19.6% is the more useful figure. For individual diagnosis and metabolic risk counselling, some clinicians prefer the stricter NIH threshold. Neither number tells any individual woman what she has, and this piece does not resolve what the literature has not.

Before the ICMR study, India had regional and clinic-derived estimates with wide variation in design and criteria. A community-based study of 778 girls aged 15 to 24 in one urban block of Mumbai found 22.5% prevalence by Rotterdam criteria and 10.7% by AES criteria, using ultrasound and clinical assessment. College-based and city-based cohorts in Andhra Pradesh, Bhopal, Lucknow, Haryana and Kashmir reported figures ranging from roughly 3.7% to 35%, often with limited biochemical testing or convenience sampling.

A 2022 systematic review and meta-analysis pooling 27 Indian studies estimated national prevalence at about 11.3% by Rotterdam criteria and 5.8% by NIH criteria, showing the same criteria gap at a smaller scale. Recent international meta-analyses give pooled adult prevalence of roughly 11% to 12% by Rotterdam and 7% to 8% by NIH, which places India at the higher end of the international range rather than outside it.

The line traces to two sources, neither of which is a national dataset of all Indian women. A 2021 systematic review of Indian adolescents aged 14 to 19 reported pooled prevalence of 17.74 per 100 by Rotterdam criteria and described that as approximately one in five, for adolescent girls in Indian settings.

Opinion pieces and news explainers then generalised the phrase to young Indian women and, more recently, to all Indian women. A second route runs through the 2024 national study itself, whose 19.6% Rotterdam figure in women aged 18 to 40 gets rounded in coverage to "nearly one in five."

None of the secondary sources generates new data. They re-quote either the adolescent meta-analysis or the ICMR national study. Where the phrase appears with no primary study linked behind it, as of 2026 it is best read as a rounded headline, and worth asking which criteria and which setting produced it.

Every study has an edge, and this one is clear about its own. The paper names five zones and eight states without listing each state in the main text, and reports zone-level estimates in a table rather than in prose, so site-level sampling fractions and non-response by site cannot be traced from the article alone.

On measurement, five specific gaps cannot be resolved from the published article:

• Machine-level variation between sites and inter-observer reliability for follicle counts are not reported.
• It is not stated whether scans were re-read centrally, which could move the polycystic-morphology numbers.
• Androgen thresholds rest on total testosterone and DHEA-S by immunoassay, with no free testosterone or sex hormone binding globulin reported.
• No mass-spectrometry re-measurement was reported, so the degree of biochemical misclassification cannot be quantified.
• The study was not designed to estimate prevalence outside the 18 to 40 age band, to stratify by caste, income or occupation, or to describe treatment patterns.

The practical value of this study for an individual reader is not the percentage. It is that a PCOS diagnosis depends on which of three published definitions a clinician applies, on whether all three components were assessed, and on how the ultrasound was performed.

If you have been told different things by different doctors, the useful question is not which one was right, but which criteria each was using and which findings were measured in your case. A registered medical practitioner can answer that from your records.

What the diagnosis then means for metabolic risk in your twenties is a separate question, and so is the belief that it makes weight loss impossible rather than harder. Where PCOS sits alongside weight, waist or glucose concerns, those findings are read together rather than as three unrelated reports, and the glucose half of that picture is easy to miss on the test most often ordered.

Been told different things by different doctors about your PCOS? A Voy clinician can tell you which criteria were applied and which findings were actually measured in your case. Book a Voy 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.