Somebody will quote you a figure for how common polycystic ovary syndrome is among Indian women. Somebody else will quote you a different one. Line up enough of them and you notice the spread is too wide for all of them to be measuring the same thing, which is the first genuinely useful piece of information in the whole exercise.
The number is not being suppressed. It does not exist in the form people want it in, and the reasons it does not exist turn out to be more interesting than the number would have been.
Every count starts with a choice of definition
Before anyone can be counted, somebody has to decide what counts. The syndrome has been defined more than once and by more than one body. An early set of criteria drew the boundary narrowly, requiring a tighter combination of features. The Rotterdam criteria published in 2003 drew it more widely. A specialist society issued its own set afterwards, sitting somewhere between the two. Each of those definitions puts the line in a different place, and a wider line admits women a narrower one leaves out.
Run the same group of women past two definitions and you will get two prevalence figures out of one population. Neither result is wrong. They are answers to different questions, and the question is usually buried in a methods section almost nobody opens. Any account of PCOS prevalence in India that does not state which criteria were applied is describing something without saying what the something is.
There is a second-order effect worth noticing here. Because the wider definition catches more women, and because the wider definition is also the one most widely adopted, the figures produced after it came into use are not directly comparable with the figures produced before. So a graph showing the condition becoming more common may be showing a genuine change, or a change in who qualifies, or some mixture of the two that nobody has separated out. Both readings get published confidently.
Where the women were found changes what gets found
The second variable is the sampling frame, and it moves the answer at least as much as the definition does.
A study recruiting from a gynaecology outpatient department is counting women who already had a reason to be sitting there. Irregular cycles. Difficulty conceiving. Something that worried them enough to take a morning off work and wait. Prevalence measured in that room will always run higher than prevalence in the surrounding population, because the room selects for the condition before the researcher arrives.
A community study is a different instrument entirely. Going house to house, or screening an enumerated population, catches women who would never have presented anywhere. It is slower, costlier and considerably harder to run, which is exactly why there is less of it. Both designs get published under the heading of prevalence research, and they are not comparable in any meaningful way, which does not stop them being averaged together in summary articles and news coverage.
College samples, city samples, and everyone missing from both
A large share of Indian research on this runs through colleges and urban clinics, for reasons that are practical rather than sinister. The population is accessible, consent is manageable, follow-up is feasible, and the research has to happen somewhere.
What that produces is a reasonably detailed picture of young urban women who have some access to healthcare. It tells you considerably less about women in their thirties and forties, less again about districts where the nearest gynaecologist involves a bus and most of a day, and close to nothing about women who have never been screened for anything in their lives.
The urban and rural difference here is not purely biological. It is about who reaches a clinic, who has a reason to go, who can afford an ultrasound and the travel to get to one, and who gets told that irregular cycles run in the family and sent home with that.
The diagnosis leans on tools that are not uniform
Even with the criteria fixed and the sample chosen, the measurement itself moves around.
- Ovarian imaging depends on the machine, the route used and the person interpreting the image. Agreement between operators is a known and openly discussed limitation.
- Hormone assays differ between laboratories. A result that crosses a threshold at one lab may not cross it at another, which matters a great deal when a definition depends on that crossing.
- Cycle history is usually self-reported, and recall is imperfect, especially for something a woman stopped paying attention to years earlier because it never seemed to be a medical matter.
- Adolescence complicates all of it. Irregular cycles are common in the years after menarche for reasons unrelated to this syndrome, so any study including teenagers has to decide how to handle that, and different studies decide differently.
Stack those together and you have measurement variation sitting on top of definitional variation sitting on top of sampling variation. The wonder is not that the estimates disagree.
The silence is part of the measurement
There is a further layer that no methods section captures properly, and in India it may be the largest one.
In a great many families, periods are not a topic. A girl learns early that the subject is managed privately, discussed with her mother if at all, and never raised where her brother or father might hear. Irregularity gets absorbed into that silence and reframed as something else along the way: it runs in our family, it will settle after marriage, it will settle after a child. Unwanted hair growth becomes a cosmetic matter to be dealt with quietly and at some expense. Weight becomes a question of discipline, and discipline is a moral category, so now there is shame attached and a further reason not to mention any of it.
A woman who never names the symptom never presents. A woman who never presents never enters a count. Under-diagnosis is not a caveat at the end of the prevalence question. It is one of that question’s main terms, and it runs in a single direction: every count is missing women, and nobody can say how many, because the missing are missing.
You hear the same silence from the other side when a woman describes years of irregular cycles and then says, almost in passing, that she had always assumed everyone’s were like that. She had no comparison available. Nobody around her was talking either.
Why the variance is itself the finding
A wide range of estimates usually gets read as noise, something to be resolved by better studies later. Read properly it is a description of the system that generated it.
The honest answer to how common this is among Indian women is that nobody knows, and saying so is not a failure of research literacy. It is the accurate summary of what the available work supports. A researcher will tell you the same thing in more careful language.
The spread tells you that the definitions have moved over time, that most sampling has been convenient rather than representative, that diagnostic tools are unevenly distributed and unevenly interpreted across the country, and that a substantial population of affected women sits outside the measurement apparatus altogether. Every one of those is a finding about Indian healthcare in its own right, and not one of them needs a precise prevalence figure to be worth acting on.
It also tells you what to treat sceptically. A single confident number, quoted with no criteria attached and no sampling frame mentioned, is usually a number that has been copied from a secondary source several times over since anyone last checked where it came from. (You can sometimes trace the same figure through article after article and never arrive at a study.)
What that leaves you with if you are wondering whether you are normal
Most people looking up prevalence are not doing epidemiology. They are trying to settle a personal question: is what my body does unusual enough to take seriously, or am I making something out of nothing.
A population figure cannot answer that. In India it particularly cannot, because the comparison group you are implicitly measuring yourself against is a group that does not discuss the thing you are trying to compare. The thought that nobody you know has this is not evidence of anything. It may only mean nobody you know has said so out loud.
What does answer the question is an individual assessment. A history, an examination, the relevant bloodwork and imaging, other causes excluded, carried out by a clinician who can then say what pattern is present and what it calls for. That is the point at which the differences between PCOS phenotypes stop being a taxonomy and start being a plan, because the pattern determines what gets treated first and in what order. No article can do that part, and no prevalence figure would help if it could.
The number nobody can hand you was never the useful thing. Your own results, read by somebody qualified to read them, always were.
