Summary
An analysis of more than 300 million laboratory tests reconstructed group-level changes around the final menstrual period and found step-like shifts across many body systems. The study also found hormone changes beginning more than a decade before that point and an observational association between menopausal hormone therapy and smaller shifts.
A study published in Nature Medicine on 12 October 2026 used more than 300 million laboratory tests from two large health datasets to reconstruct how measurements change around the final menstrual period (FMP). Its model found step-like shifts in many physiological systems, alongside accelerated changes in several sex hormones more than a decade before the FMP.
The analysis covered 90 types of laboratory tests and more than one million female individuals in US and Israeli datasets. The researchers describe the modeled pattern as a “physiological cliff”, but it is a population-level reconstruction: it estimates average trajectories aligned to menopause rather than tracking abrupt changes in each individual.
A model aligns tests to the final menstrual period
The FMP is the last menstrual period before menopause, which is identified after 12 months without a period. In large medical datasets, the timing of the FMP is often missing, and comparing test results only by age can blur changes that occur at different times for different people.
To address this, the researchers developed a computational method called YoGlen. It uses the distribution of menopause ages to adjust for that variation, in a process the authors compare to removing blur from an image. The method estimates what average test trajectories might look like when aligned by time relative to the FMP. It does not predict the FMP date for each person.
The team applied the method to US National Health and Nutrition Examination Survey data collected from 1999 to 2022 and to health records from Clalit, an Israeli health provider. Although Clalit’s records span 2004–2024, laboratory results in both datasets were analysed cross-sectionally. The researchers compared the reconstructed patterns with those for male individuals, using age 50 as the equivalent reference point in that comparison.
Many test results shifted; hormone changes began earlier
The researchers found a statistically significant menopause-associated jump in 52 of the 90 tests among female individuals, with estimated jump sizes ranging from 0.1 to 1.4 standard deviations. These included tests relating to hormones, bone, liver, kidney, blood chemistry and lipids. The corresponding abrupt shifts were not seen in the male comparison group. The sizes of the jumps across tests were strongly correlated between the US and Israeli datasets (r = 0.89).
The pattern was not uniformly adverse: iron-, haemoglobin- and red-blood-cell-related measurements generally shifted in a direction associated with less anaemia, consistent with menstrual bleeding stopping. The researchers also found that the prevalence of several health deficits, including osteoporosis and cardiovascular outcomes, rose more steeply after menopause. These are changes in population-level prevalence patterns, rather than evidence that the lab shifts themselves caused those outcomes.
The reconstructed trajectories placed earlier changes in three hormones. Anti-Müllerian hormone began an accelerated decline around 15 years before the FMP; follicle-stimulating hormone began rising faster around 14 ± 1 years before it; and estradiol began declining faster around 12 ± 1 years before it. These estimates extend the timeline captured by the study, but they are inferred from population data rather than measurements that followed every participant throughout the transition.
The researchers also found that many menopause-associated lab shifts were smaller among participants using menopausal hormone therapy (MHT). The finding appeared in the US analysis and was moderately correlated with results from the Israeli dataset. The pattern was not uniform: C-reactive protein was among the exceptions.
That comparison was observational, not a randomised treatment trial. MHT users may differ from non-users in ways the analysis could not fully account for, so the results describe an association and do not establish that MHT caused the smaller shifts. The study’s broader contribution is a method for reconstructing group-level biological timelines from large datasets when the timing of a transition is not recorded for everyone.