What actually changes in your hormones from 35 onwards

What made me curious

In my early forties I’d been feeling different for quite a while, and I couldn’t really say how. Then my periods started changing too, and that’s when I began to wonder whether all of it might be hormonal.

What I wanted to know

Once I started reading, what surprised me most was how far it reached. Nobody had told me that hormonal change, and perimenopause in particular, could touch so much: how clearly I thought, my mood, my sleep, how easily I learned new things, even how much I wanted to be around other people. I genuinely had no idea.

So I wanted a map. Which hormones change from the mid-thirties onwards, in what order, and by how much? Is it the slow, steady decline people describe, or something messier? And why would a change in the ovaries show up in the brain, the bones and sleep at all? This page is my attempt at that map, and it sits at the centre of the Hormones section.

The big picture first

A lot of the best long-term data comes from SWAN, the Study of Women’s Health Across the Nation. It’s a large US study that has followed women through midlife. One limitation runs through everything below: SWAN enrolled women aged 42 to 52, so the late thirties and early forties are mostly pieced together from smaller studies and models. I found that slightly frustrating, since it’s the stretch I was most curious about.

A 2021 Endocrine Society review describes the transition as a gradual loss of eggs, a change in how the ovaries and brain respond to each other’s signals, wide hormonal fluctuations and irregular cycles, over a period that can last more than a decade.

What changes first, usually in your late thirties

The earliest changes are ones you can’t feel, and they start years before anything shows up in your cycle. I hadn’t realised that at all.

AMH (anti-Müllerian hormone) works as a rough marker of the ovaries’ remaining egg supply. A model built from data on 3,260 women found that it peaks at around 24.5 years and declines from then until menopause. Age explained only about a third of the variation. So two women of the same age can have very different levels, which is worth remembering before you compare notes with a friend.

Inhibin B falls too. A 2007 study sampled 76 women three times a week across a full cycle and grouped them by stage. Inhibin B fell steadily from one stage to the next, and AMH dropped ten- to fifteen-fold.

FSH (follicle-stimulating hormone) is the brain’s signal to the ovaries. As the feedback between them changes, FSH rises. In SWAN, it began rising about six years before the final period and sped up about two years before. That helped me make sense of feeling different for so long before anything obvious happened, though I can’t know what my own levels were doing.

Estradiol: swings first, a decline later

Estradiol is the estrogen most of this research measures. I’d assumed it simply drifted down over the years, and the research describes a rather different shape.

On average, estradiol didn't change until about two years before the final period. It then fell fastest around the final period itself and levelled off about two years afterwards. FSH had started rising roughly six years before.

That’s the average, and one woman’s cycles can look quite different. A small 1996 study collected daily urine samples from 11 women aged 43 and over, and 11 younger women. The older group was excreting more estrogen on average, with higher FSH and less progesterone in the second half of the cycle. The 2007 study above also found estradiol in ovulatory cycles rising across the stages. This was the finding that surprised me most, but both studies are small, so I’d treat “higher than before” as promising rather than settled.

So how can both be true? If one month runs high and the next runs low, and those swings land at different times for different women, the average can stay flat while each woman is on something of a rollercoaster. As far as I can tell, that’s the most useful way to hold it: swings first, and the sustained decline later.

There’s no single typical curve, either. A 2012 SWAN analysis found four estradiol patterns: a slow decline (about 27%), a flat line (about 29%), a rise followed by a slow decline (about 13%) and a rise followed by a steep decline (about 32%). Body size shifted the odds, and in a different direction depending on stage. In SWAN’s 2004 analysis of 3,257 women, a higher BMI went with lower estradiol early in the transition and higher estradiol later. Levels also differed by ethnicity.

Doctors who work with a lot of women at this stage often say perimenopause can start in the late thirties and last seven to ten years, which fits the research above. That view comes from years of seeing patients, and it’s worth taking seriously, since plenty of things were noticed in practice long before anyone measured them. Here the cohort data happen to agree.

Progesterone: often the first to slip

Progesterone rises after ovulation, so as ovulation becomes less reliable, progesterone falls with it. SWAN’s Daily Hormone Study collected daily samples from 848 women aged 43 to 53. Over three years, the share of cycles that looked ovulatory fell from 80.9% to 64.7%, while cycles with neither ovulation nor bleeding rose from 8.4% to 24%. The progesterone marker fell by about 6.6% a year. One detail I found striking is that both very high and very low estradiol were linked to cycles without ovulation, which fits the picture of swings in both directions.

Testosterone: an earlier, quieter story

I’d assumed testosterone drops at menopause along with everything else. The research suggests a different timeline, though the studies don’t fully agree on the details.

A 2005 study of 1,423 women aged 18 to 75 found testosterone and related hormones declining with age, most steeply in the earlier decades, with no separate effect of natural menopause. A study following 172 women through menopause found testosterone unchanged around the final period, while the protein that binds it (SHBG) fell by 43%, so the amount of testosterone free to act rose by about 80% on one measure. And SWAN found DHEAS, an androgen made by the adrenal glands, rising in the late transition in most women.

The disagreement seems to come partly from study design, since snapshots of different women tell a different story from following the same women over time. What does seem consistent is that there’s no menopause cliff for testosterone. Whether the earlier decline causes anything women notice is a separate question, and one I haven’t seen answered well.

The stages, in plain language

Researchers describe all this with a staging system called STRAW+10, published in 2012. It’s built mainly on what your periods are doing, with hormones as supporting evidence, and applies regardless of age, ethnicity or body size. Roughly:

• Late reproductive (stage −3b). Periods still regular. FSH unchanged, but AMH and the number of visible follicles already low.
• Late reproductive (stage −3a). Subtle cycle changes, usually shorter cycles. FSH early in the cycle starts rising and becoming more variable.
• Early transition (stage −2). The length of consecutive cycles differs by seven days or more, and that keeps happening within ten cycles. FSH raised but variable.
• Late transition (stage −1). Sixty days or more without a period. Hormones fluctuate most here, and cycles without ovulation become common. Usually lasts one to three years, and hot flashes are likely.
• Early postmenopause (+1a, +1b, +1c). FSH keeps rising and estradiol keeps falling for about two years after the final period, then both settle. The last part of this stage lasts an estimated three to six years.

In this system, perimenopause begins at stage −2 and ends twelve months after the final period. That means the final period itself can only be identified looking back, once a year has passed without another.

Why a single blood test rarely settles it

Given the swings, a single blood draw tells you about one day. NICE, the UK guideline body, recommends identifying perimenopause without laboratory tests in otherwise healthy women aged 45 or over who have recently started having hot flashes or cycle changes. It also advises against FSH tests in women taking combined hormonal contraception or high-dose progestogen. Testing deserves its own piece, but once I understood the swings, it made sense to me that the guideline leans on age and symptoms rather than a one-off panel trying to pin down something that moves from week to week.

Why estrogen matters far beyond reproduction

This was the part that finally explained my surprise. What follows is mostly mechanism: it’s about where estrogen can act, and it doesn’t on its own prove what estrogen does in any one woman.

Estrogen acts through at least three receptors, known as ERα, ERβ and GPER. Reviews describe them in bone, the brain, blood vessel walls and the heart, the lining of the gut (ERβ is the main estrogen receptor in the colon), immune cells and fat. A 2013 review describes estrogen acting on several skin cell types and hair follicles, with thinner, drier, less elastic skin when estrogen is low. Each receptor has its own pattern, and much of the detail comes from animal work.

Some clinicians put it even more broadly, saying estrogen receptors are in almost every cell, which helps explain why symptoms turn up in such unexpected places. The receptor reviews support the spirit of that, though distribution varies by tissue and cell type.

What the large cohort studies add is timing, which is a different thing from cause. In SWAN:

• Bone loss began about a year before the final period and was fastest until two years after. Of roughly 10.6% lost from the lower spine over ten years, about 7.4% came in that three-year window.
• Body composition shifted at the start of the transition: the rate of fat gain rose from about 1% to 1.7% a year and lean mass began to fall, even though weight rose at a steady pace throughout.
• Sleep got harder through the transition. Falling estradiol was associated with trouble falling and staying asleep, and hot flashes with all the sleep problems measured.
• Learning dipped. In 2,362 women, those in perimenopause didn’t show the usual improvement from repeating a test in processing speed and verbal memory, and it came back after menopause.

That last one stopped me for a moment, because the ability to learn was one of the things I’d had no idea hormones could touch. The American Heart Association also describes the transition as a time when body composition, cholesterol and blood vessel health worsen, and calls it a window for early prevention.

The brain research is the newest and, to me, the most intriguing. A research group at Weill Cornell ran a 2024 imaging study that measured estrogen receptors directly in living brains.

Estrogen-receptor density in the brain was progressively higher from pre- to peri- to postmenopause, regardless of age or blood estradiol. Higher density went with poorer memory and predicted mood and cognitive symptoms after menopause.

One possible explanation is that the brain is compensating for less estrogen, but that’s an interpretation the study can’t show. It’s one group’s work in 54 women and hasn’t been replicated. An earlier imaging study from the same group, in 161 women, found brain structure, connectivity and energy use differing by menopause stage, then largely stabilising afterwards. With SWAN’s learning results, that suggests some brain changes may belong to the transition rather than lasting indefinitely. I’ll admit that’s the reading I’d hope for, so I’ve tried to hold it loosely.

What the evidence actually says

• Established: AMH and inhibin B fall and FSH rises years before the final period. On average, estradiol holds until about two years before and then falls. Progesterone falls as ovulation becomes less regular. Testosterone has no menopause cliff.
• Established as timing, not cause: bone loss, the shift in body composition, sleep problems and the learning dip cluster around the transition.
• Promising: that estrogen is often higher than before for stretches of perimenopause (studies of 11 and 76 women), and that brain estrogen receptors increase (one study of 54).
• Conflicting: exactly how testosterone and related hormones behave, depending on study design.
• Marketing: that you can “balance your hormones”, or that one blood panel tells you where you are in the transition.
• Unknown: which pattern any individual woman is on, and whether symptoms follow hormone levels or the swings.

You can read how I weigh studies like these on how we research.

What this means in real life

If you’re in your late thirties or forties and something feels different, the hormones may already be moving, even if your periods look normal. A run of cycles that vary by a week or more is the clearest outward sign. Symptoms that come and go, or look nothing like the hot flashes you were told to expect, fit a picture of swings better than a slow fade.

Bone, muscle and heart health shift over the same years. That makes this a reasonable time to think about strength training, about which supplements have evidence behind them, and about what to ask your doctor. If you’re new here, Start here explains how the site fits together.

What I still don’t know

• How common the high-estrogen phases are, and in whom. The detailed daily studies are tiny, and no large study gives a prevalence figure.
• Whether symptoms track hormone levels or the size of the swings. The cohort studies link symptoms to falling estradiol, and the “it’s the volatility” idea rests mostly on mechanism and clinical observation. I couldn’t find a study that settles it.
• What rising brain estrogen receptors mean. Compensation is one reading. Replication would help.
• What happens between 35 and 42. SWAN started at 42, so this stretch is under-studied directly.
• Which pattern I’m on. The trajectory research shows the groups exist, but it can’t tell any one woman which group she’s in.
• Whether testosterone’s earlier decline matters for how women feel. That’s a separate evidence question, and I’d like to look at it properly.

If any of that changes, I’ll come back and update this.