REF-1404 · Greendale et al., 2012 · Journal of Bone and Mineral Research · Paper · PubMed · Free full text
When does the menopause start costing bone?
In 862 midlife women (not athletes), spine and hip density began falling about a year before the last period, fastest for 3 years (spine ~2.5% a year), then slower; spine −10.6% in 10 years.
Main caveat: Density, not fractures. DXA areal density at two sites; no fracture outcome, no bone geometry or microarchitecture (the authors' concern about turnover is not…
Greendale and colleagues asked when menopausal bone loss starts and stops relative to the final menstrual period (FMP), how fast and how much density falls in the 5 years either side of it, and whether age at the FMP, body mass index (BMI) and race or ethnicity change the rate.
From the SWAN bone cohort they took 862 women (242 African-American, 384 Caucasian, 117 Chinese, 119 Japanese) who were premenopausal or early perimenopausal at baseline, not on hormone therapy or other bone-active drugs, and who later had a natural FMP whose date could be fixed from annual interviews (the last bleeding before 12 months without menses).
Lumbar-spine and femoral-neck density were measured yearly by DXA. Smoothed curves showed no measurable loss until 1 year before the FMP; loss then began and decelerated, but did not cease, 2 years after it, at both sites.
Piece-wise models with breaks at FMP minus 1 year and plus 2 years gave, for the Caucasian reference woman, a spine loss of 2.46% a year (95% CI 2.31 to 2.61) and a femoral-neck loss of 1.76% a year (1.61 to 1.92) over the 3-year "transmenopause", and about 1.1% a year at both sites from 2 to 5 years after the FMP.
Cumulative 10-year loss was 10.6% at the spine (7.38% of it in the transmenopause) and 9.1% at the femoral neck. Each additional BMI unit slowed loss; African-American women lost less and Chinese and Japanese women more, differences of about 1-2% over 10 years. A later FMP meant faster loss in the transmenopause but no difference over the 10 years.
Menstrually defined stages were a poor signal of the start: in the year before the FMP, 68% of observations were in women still classed as early perimenopausal (cycles irregular but no gap of 3 months), and only 28% had reached late perimenopause when rapid loss began.
The authors judge the average transmenopausal loss too small on its own to reach treatment thresholds, flag the fast turnover behind it as a possible threat to bone microarchitecture, and say the data are not sufficient to recommend treatment.
Our reading: this gives the base the timing and size of menopausal bone loss it lacked (PHYS-37 carries the loading principle, not the life stage); for an athlete it says when bone is most at stake (the years around the last period, which she cannot date in advance from her cycles), not what to do about it — the tested loading prescription is in REF-1405, in older women with low bone mass.
Protocol
Design: SWAN, a multi-site, community-based longitudinal cohort of the menopause transition; this analysis covers baseline to follow-up visit 10 in the bone cohort. Written informed consent; IRB approval at each site.
Eligibility (SWAN): age 42-52, intact uterus and at least one ovary, not using hormone therapy, at least one menstrual period in the 3 months before screening, self-identified as one of the eligible ethnic groups. Bone density was measured at 5 of 7 sites (Boston, Detroit, Pittsburgh, Los Angeles, Oakland); 2,335 were in the bone cohort at baseline.
This sample: n = 862 with a determinable natural FMP; hormone therapy and bone-active drugs (tamoxifen, raloxifene, GnRH agonists, corticosteroids, osteoporosis treatments) excluded at baseline; data censored when a woman started a bone-active medicine. At baseline: age 46.7 (SD 2.6) years, 58% premenopausal and 41% early perimenopausal, BMI 27.4 (SD 7) kg/m², 16% current smokers; mean age at FMP 51.6 (SD 2.4) years. Mean 9 scans per woman (median 10 of a possible 10).
Measurements: lumbar-spine and femoral-neck BMD by Hologic DXA (4500A; two sites changed from 2000 to 4500A at visit 8 and cross-calibrated on 40 women), with central quality control; short-term in vivo precision 1.4% (spine) and 2.2% (femoral neck).
Exposure: months before or after the FMP at each scan. Menopause stages from annual bleeding reports: premenopausal (regular), early perimenopausal (menses in the prior 3 months but less predictable), late perimenopausal (3 to under 12 months of amenorrhoea), postmenopausal (12 months or more).
Analysis: loess curves of baseline-normalised BMD, then mixed-effects piece-wise linear regression with knots at FMP −1 year and +2 years (chosen by explained within-woman variance; moving the first knot worsened fit, moving the second by 6 months did not), random intercept and slopes; adjusted for baseline BMD and clinical site, with age at FMP, ethnicity and baseline BMI as fixed effects on each slope.Segments: pre-transmenopause (5 to 1 years before the FMP), transmenopause (1 year before to 2 years after), postmenopause (2 to 5 years after). Slopes are percentage change from baseline per year; reference = Caucasian women of average BMI, baseline BMD and age at FMP.
Funding and interests: SWAN grants from the US National Institutes of Health through the National Institute on Aging, the National Institute of Nursing Research and the NIH Office of Research on Women's Health (NR004061; AG012505, AG012535, AG012531, AG012539, AG012546, AG012553, AG012554, AG012495)."All authors state that they have no conflicts of interest." No device maker, federation or data platform is named; the Hologic densitometers and Synarc quality-control programme are described as methods.
Key Results
Annual BMD change (95% CI) in the Caucasian reference woman, and cumulative change over the 10 years from 5 years before to 5 years after the FMP (Tables 1 and 2):
| Site | Pre-transmenopause (−5 to −1 y) | Transmenopause (−1 to +2 y) | Postmenopause (+2 to +5 y) | 10-year cumulative |
|---|---|---|---|---|
| Lumbar spine | −0.02% (−0.08% to +0.05%) | −2.46% (−2.61% to −2.31%) | −1.06% (−1.21% to −0.91%) | −10.6% (−11.2% to −10.0%) |
| Femoral neck | −0.06% (−0.13% to +0.01%) | −1.76% (−1.92% to −1.61%) | −1.12% (−1.32% to −1.04%) | −9.1% (−9.7% to −8.5%) |
Transmenopausal share: 7.38% of the spine's 10.6% was lost in the 3-year transmenopause; for the femoral neck the abstract and the first Discussion paragraph give 5.8%, a later paragraph 5.3% (see Limitations).
BMI: each kg/m² higher at baseline slowed spine loss in every segment (+0.008%, +0.063%, +0.018% a year) and 10-year loss by 0.28% (0.21% to 0.34%); at the femoral neck only in the transmenopause (+0.025% a year), 10-year +0.08% (0.01% to 0.16%). A Caucasian woman 7.5 kg/m² above average still lost 8.50% at the spine over 10 years. Gaining BMI during follow-up went with a higher final density (spine +0.10%, femoral neck +0.35% per BMI unit).
Ethnicity (10-year, vs Caucasian): spine — Chinese −2.0% (−3.1% to −0.7%; 12.6% total loss), African-American +0.9% (−0.1% to +1.9%; 9.6%, borderline), Japanese +0.5% (−0.7% to +1.8%; no evidence of a difference); femoral neck — Japanese −1.7% (−2.9% to −0.38%), Chinese −1.2% (−2.4% to +0.0%), African-American +1.1% (0.1% to 2.1%).Transmenopausal femoral-neck loss: 1.42% a year in African-American, 2.13% in Japanese and 2.17% in Chinese women. The differences sat almost entirely in the transmenopause.
Age at FMP: each later year meant slightly faster transmenopausal loss (spine −0.050% a year, femoral neck −0.055%) and no difference in 10-year loss (spine −0.04%, −0.19% to +0.12%).
Bleeding pattern as a signal (Table 3): in the 12 months before the FMP, 503 of 742 scans (68.0%) were in women still classed as early perimenopausal and 205 (27.7%) late perimenopausal; in the 12 months after it, 266 of 871 (30.5%) were still early and 535 (61.4%) late perimenopausal. Crude mean spine BMD was 1.07 g/cm² from 5 to 1 years before the FMP and 0.97 g/cm² from 3 to 5 years after.
Clinical size (authors' example): a Caucasian woman at the cohort's 5th percentile of femoral-neck density, 0.69 g/cm² (T-score −1.4), would be at about 0.64 g/cm² (T-score −1.8) two years after the FMP.
Mechanisms
Measured: density held steady until about a year before the final period, then fell fastest for three years and more at the spine than at the femoral neck; afterwards both sites kept losing at similar, slower rates.
Authors' interpretation: the larger spinal loss fits the spine's higher share of trabecular bone; the early fast phase has been ascribed to losing oestrogen's tonic restraint on bone turnover and the later slow phase to oestrogen-deficiency secondary hyperparathyroidism (a model they cite, not tested here). They suggest the rapid turnover may matter more than the amount lost, through thinning and disconnecting trabeculae and eroding the inner cortex.
Authors' interpretation: the menopause transition, with its change in oestradiol and other factors, appears to start cortical loss and amplify trabecular loss; DXA is too insensitive to see the earlier, small trabecular losses that CT studies report.
Inference: the window when bone loading would matter most is the years around the final period, which a woman cannot date while she is in them. Whether loading in this window slows the loss is not tested in this paper; the base's loading principle is PHYS-37 and its tested programme is REF-1405, in women about 15 years older with low bone mass.
Practical Application
When does the menopause start costing bone? In 862 midlife women (not athletes), spine and hip density began falling about a year before the last period, fastest for 3 years (spine ~2.5% a year), then slower; spine −10.6% in 10 years.
Irregular cycles are already the window: most women were still classed as early perimenopausal (cycles irregular, no 3-month gap) in the year their bone loss began, so a woman in her mid- to late 40s whose cycles are changing may already be in it (B+, cohort, non-athletes).Not from this paper; the base's own practice: add impact or heavy lifting for bone (PHYS-37; a tested programme is REF-1405); that the transition, not the time after periods stop, is when to have it in the week is our inference from this cohort's timing.
The loss does not stop: about 1.1% a year at both sites from 2 to 5 years after the last period (Caucasian reference), so the case for loading carries on into the postmenopause.
How much it matters for one woman: on average the transmenopausal loss is not enough by itself to reach a treatment threshold (the authors' 5th-percentile example moves from a T-score of −1.4 to about −1.8), and the authors did not find the data sufficient to recommend treatment.A woman who enters the transition with low density — cyclists and swimmers often do (REF-896, REF-522) — starts closer to it (inference). Bone density testing and any drug, including hormone therapy, are a clinician's decision; what the menopause society says about hormone therapy and bone is in REF-1406.
A lengthening cycle at 45 is not only a fuelling question (inference): in this cohort irregular cycles were the normal run-up to the final period.The base's referral line in SYNTH-37 and the REDs screen in SYNTH-54 now send a missing period to a doctor's work-up and do not read the irregular cycles of the menopause transition as REDs by default (2026-10-28, the base's inference; SYNTH-96); in a woman in her 40s a changing cycle may be the transition, and either way the answer is a doctor, not a label.
What it is good for as a source: the timing and size of menopausal bone loss at the spine and hip in a large, multi-ethnic cohort, anchored to the final period. What it does not show: fracture risk, rates in trained or loaded women, whether training or any treatment changes the loss, or anything before 5 years ahead of the final period.
Limitations
Density, not fractures. DXA areal density at two sites; no fracture outcome, no bone geometry or microarchitecture (the authors' concern about turnover is not measured here).
Who. Community-dwelling women recruited by ethnicity at US sites, natural menopause only, not using hormone therapy or bone-active drugs at entry (censored if they started one), average BMI 27.4; physical activity not modelled. No athletes; the rates are not known for trained women or for women already loading bone.
Timing uncertainty. The deceleration point at 2 years after the FMP was much less distinct than the onset (moving it 6 months either way did not change model fit); the authors computed 10-year totals to work round this. SWAN enrolled women in their mid-forties, so nothing earlier than 5 years before the FMP is captured, and follow-up stops 5 years after.
Numbers that do not add up in the manuscript: the Results text gives spine postmenopausal change as −1.04% a year and femoral-neck as 1.15% a year, while Tables 1 and 2 give −1.06% and −1.12% (the femoral-neck interval, −1.32% to −1.04%, is not centred on its estimate); femoral-neck transmenopausal loss is 5.8% in the abstract and first Discussion paragraph but 5.3% later in the Discussion, and 3 years at 1.76% a year is 5.28% (our arithmetic), so 5.3% is the consistent figure; mean age at FMP is 51.6 in the text and 51.7 in the table footnotes.The tables are used above.
Subgroups. Non-Caucasian groups were large enough to detect ethnic differences but not to test interactions within groups; ethnic differences were about 1-2% over 10 years and, in the authors' words, unlikely to explain ethnic differences in fracture rates.
Version read. The NIH author manuscript in PMC, not the typeset version of record; the supplementary flow diagram was not read.
Funding and interests: NIH (NIA, NINR, ORWH) grants to SWAN; all authors state no conflicts of interest. No device maker, federation or data platform is named.
How this entry was checked and changed
Why this level
large observational cohort (n = 862, above the rubric's 200), up to 10 yearly scans per woman (median 10) on cross-calibrated Hologic machines; read from the NIH author manuscript in PMC (the accepted version, not the publisher's typeset text). Held at B+ because it describes change in density, not fractures, in women who were not using hormone therapy at entry, and gives no intervention
Full text (checked 2026-10-28)
The analytic sample consisted of 242 African-American, 384 Caucasian, 117 Chinese and 119 Japanese women; at baseline, mean value of age was 46.7 years and mean age at FMP was 51.6 (full text)
BMD loss began 1 year before the FMP and decelerated (but did not cease) 2 years after the FMP, at both the lumbar spine (LS) and femoral neck (FN) sites (full text)
Caucasian transmenopausal change in LS BMD was −2.46% per year and postmenopausal change was −1.04% annually; summed 10-year change was −10.6% (full text)
Cumulative, 10-year LS BMD loss was 10.6%; 7.38% was lost during the transmenopause. Cumulative FN loss was 9.1%; 5.8% was lost during the transmenopause (full text)
Caucasian women with sample-average BMI, age at FMP and baseline FN BMD (0.832 gm/cm^2) lost 1.76% annually during the transmenopausal interval and 1.15% of FN BMD annually during the postmenopausal segment (full text)
During the year before the FMP, when bone loss accelerated, 68% of participants who were observed were still classified as early perimenopausal based on bleeding patterns; only 28% of women had reached late perimenopause when rapid BMD loss began (full text)
A Caucasian woman with a baseline FN BMD at the 5^th percentile for SWAN Caucasians, 0.69 g/cm2 (a T score −1.4), would have a femoral neck BMD of ~0.64 g/cm^2 two years after the FMP (a T score of −1.8) (full text)
We do not believe that the currently available data are sufficient to recommend treatment (full text)
SWAN has grant support from the NIH through the NIA, the NINR and the ORWH; all authors state that they have no conflicts of interest (full text)