Menopause and longevity: what the cohorts actually show

Two meta-analyses of the same question, on the same definition, reach different conclusions about mortality — one finds a raised risk of dying earlier, the other finds nothing. Both are well built. The interesting part is what separates them.

mixedStudies disagree, or the mechanism is contested.Revised 05 Oct 2026

A smooth unglazed clay hourglass standing on a pale wooden surface against a plain cream backdrop, with three small clay pebbles resting beside its base

Two reviews, same question, different answers on mortality. The cardiovascular signal repeats; the death signal does not.

The question people actually have about menopause is not medical. It is arithmetic: does the timing of it change how long you live? Everything else — the flushes, the sleep, the weight, the mood — is a quality-of-life question, and those are answered differently. The longevity question has a real literature behind it, and the interesting thing about that literature is that it does not give one answer.

Two meta-analyses, both rigorous, both examining essentially the same exposure, reach opposite conclusions about the thing longevity writing cares about most: death.

The two answers

The first is from 2016, in JAMA Cardiology (PMID 27627190). It pooled 32 studies covering 310,329 women, comparing those whose menopause began before 45 with those at 45 or later. The cardiovascular signal is strong and clean: overall coronary heart disease carried a relative risk of 1.50 (95% CI 1.28–1.76), cardiovascular disease mortality 1.19 (1.08–1.31), and all-cause mortality 1.12 (1.03–1.21). Stroke was not among the outcomes that carried the signal: the review reports a stroke mortality figure of 1.11 without a confidence interval in the abstract, and nothing in its conclusions rests on it.

So in 2016, the answer to the longevity question was: a small but detectable increase, about twelve per cent.

The second is from 2026, in Frontiers in Public Health (PMID 42694168). Seven cohorts, more than 1.77 million postmenopausal women — nearly six times the sample of the earlier review, and drawn from the era in which early menopause is both more common and better recorded. Early menopause was defined the same way, before 45. And on all-cause mortality the result was null in both strata: 1.20 (95% CI 0.83–1.74) in women with type 2 diabetes, and 1.18 (0.94–1.47) in women without. Both intervals straddle 1. The review’s own summary says the association was not statistically significant.

The cardiovascular events, meanwhile, are still there in the 2026 data. Among non-smoking women, early menopause went with a higher risk of heart failure (HR 1.23, 95% CI 1.02–1.49) and coronary heart disease (RR 1.26, 1.18–1.35). In women with type 2 diabetes, heart failure again (HR 1.29, 1.11–1.50) — but not stroke, which came out at HR 1.11 with a confidence interval of 0.99–1.25 and did not clear significance. The clean stroke signal in the 2026 data belongs to the women without diabetes (HR 1.11, 1.09–1.13).

Read those two reviews side by side and a consistent shape emerges. The cardiovascular signal repeats across decades of data and tens of millions of person-years. The death signal does not.

What the two disagree about, and what that means

A twenty per cent swing in a mortality estimate between two reviews of the same exposure is not a rounding error, and it is worth being precise about what moved.

The 2016 review’s own detail on one sub-question is instructive. When it compared women whose menopause began between 50 and 54 with those beginning before 50, it found lower fatal CHD (RR 0.87, 95% CI 0.80–0.96) and no effect on stroke. And for time elapsed since onset — arguably the more longevity-relevant exposure, since it maps onto how many years of post-menopausal life someone has — only four observational studies reported anything, and their results were described as inconsistent.

That is the weak point, and it is where the 2026 review differs most. It stratified rather than pooled blindly: in women who smoke, the early-menopause associations with heart failure (1.15, 0.88–1.52) and coronary heart disease (1.63, 0.90–2.96) both vanished, and the coronary estimate carried a heterogeneity of 97.2% — a number that means the underlying studies are close to disagreeing with each other. A pooled figure drawn across such studies is being asked to average things that are not the same thing.

There is a third pattern that should make anyone cautious about the mortality question, and it comes from an older cohort that pointed the opposite way to intuition. The Iowa EPESE study followed 1,684 women aged 65 and over for up to 24 years, of whom 1,477 died (87.7% of respondents) (J Womens Health 2012, PMID 21970557). It tested the hypothesis that later menopause would mean lower mortality. The result was the reverse: women whose natural menopause came at 55 or later had higher all-cause and cardiovascular mortality than those who reached it younger, and the pattern survived adjustment for confounders. The type of menopause, and the age at a surgical menopause, were not related to mortality at all.

Read that carefully, because it is not a contradiction of the pooled reviews. It is a demonstration that the relationship is not a straight line running from early to late. A cohort finding at the late end of the range and a pooled estimate across the whole span can both be true at the same time and can both fail to point the same way. The 2016 meta-analysis noticed the same shape too, in its 50–54 versus under-50 comparison.

The definition problem underneath all of it

Before reading any of these numbers, it is worth knowing that the exposure is usually self-reported — and the meta-analyses say so.

The 2016 mortality review (Climacteric, PMID 26576012), which looked specifically at primary ovarian insufficiency and early natural menopause, drew on nine articles from seven prospective cohorts, and states plainly that in all of them the age at menopause was self-reported. Recall of a date is exactly the kind of variable where misclassification runs in both directions, and misclassification in the exposure blurs the estimate towards null — a technical route to a null finding that has nothing to do with biology.

That review’s own results show how much the definition does. Splitting “early” into two categories gave two different answers. For primary ovarian insufficiency, spontaneous menopause before 40, the pooled relative risk of death from all causes was 1.39 (95% CI 1.10–1.77) and from ischaemic heart disease 1.48 (1.02–2.16). For early natural menopause at 40–44, the ischaemic heart disease estimate was 1.09 (1.00–1.18) — a sliver of an effect with a confidence interval that only just excludes 1.

A “premature menopause” that means before 40 and an “early menopause” that means 40 to 44 are not the same exposure, and the difference between 1.39 and 1.09 is almost entirely down to where the line was drawn. The headline number in any news story about this is a function of a definitional choice, which is a good reason to read the definition before the figure.

What the data do not show

They do not show that early menopause shortens life. The largest and most recent synthesis, on more than 1.77 million women, found no statistically significant association with all-cause mortality. An earlier review found a small one. The honest statement is that this is unresolved, and that anyone telling you it is settled is choosing the review that suits them.

They do not show that cardiovascular events after early menopause are caused by it. The associations are consistent and they are large in number, but every cohort here is observational. Women whose menopause comes earlier differ in many ways beforehand, and cohorts adjust for what they measured rather than for everything that might matter.

They do not show anything about the timing question. The 2016 review found only four studies reporting outcomes against time since onset, with inconsistent results. How many years of post-menopausal life you have is the most longevity-relevant version of the question and it is the least studied.

They do not transfer across populations. The 2026 review’s smoking stratification is the clearest example: among women who smoke, the early-menopause associations with heart failure and coronary heart disease were not statistically significant, with heterogeneity at 73.7% and 97.2% respectively. A pooled estimate across such groups describes an average that no individual woman resembles.

They do not show anything about an individual. A relative risk of 1.12 is a statement about a distribution of outcomes across many people, not a prediction. It does not say what will happen to one particular woman, at any particular age, and it cannot be inverted into a personal plan.

The part that is actually settled

It is worth ending on the one thing the evidence does support without much dispute, because it is the part that has nothing to do with mortality.

The cardiovascular association is the most robust finding across both reviews and across the intervening decade. Whether it is framed as a twelve per cent rise in mortality or a twenty-six per cent higher coronary risk in non-smokers, the shape is the same: the earlier the transition, the more cardiovascular signal in the cohorts that follow. Two independent meta-analyses, different samples, different countries, different decades, same direction.

Everything beyond that — whether it translates into years of life, whether the effect is causal, whether the curve is monotonic, whether it holds in smokers — is open. That is a narrower and more honest version of the claim than the one usually made, and it is a genuinely useful thing to know.