The association of infertility, miscarriage, stillbirth, and parity with osteoporosis: a pooled analysis of five cohort studies

Overall, 141,222 women were included ( Fig. 1 ). They were at a median (quartiles) age of 59.0 years (54.0, 63.0) at cohort baseline and were followed up to the age of 72.0 years (68.0, 76.0). By the end of follow-up, 16,725 (11.8%) women developed osteoporosis. Women who were older, less educated, had lower BMI and lower physical activity level at baseline, were more likely to develop osteoporosis ( Table 2 ). Women with infertility were at 16% higher risk of osteoporosis compared with women without infertility (HR = 1.16, 95% CI: 1.13–1.19; Table 3 ). The association for infertility was not attenuated by additional adjustment for age at natural menopause (HR = 1.16, 95% CI: 1.13–1.19). Among women who had ever been pregnant, a history of miscarriage (HR = 1.06, 95% CI: 1.04–1.08), especially recurrent miscarriages (≥3; HR = 1.17, 95% CI: 1.05–1.30), was associated with a higher risk of osteoporosis. The effect of recurrent miscarriages was attenuated slightly after additional adjustment for age at natural menopause (HR = 1.15, 95% CI: 1.03–1.29; Table 3 ). Among women who had ever been pregnant, women with stillbirth were at 14% higher risk of osteoporosis, compared with women without stillbirth (HR = 1.14, 95% CI: 1.10–1.17), and this changed very little after adjustment for age at natural menopause (HR = 1.13, 95% CI: 1.10–1.16; Table 3 ). Low parity (no live births: HR = 1.20, 95% CI: 1.15–1.25; one live birth: HR = 1.15, 95% CI: 1.14–1.16) was associated with a higher risk of osteoporosis compared with women with two live births ( Table 3 ). Additional adjustment for age at natural menopause had little influence on the observed association for low parity (no live births: HR = 1.16, 95% CI: 1.12–1.20; one live birth: HR = 1.13, 95% CI: 1.12–1.13). Adding ICD code M82 to identify osteoporosis revealed similar results with the main analysis ( Supplemental Table 2 ). After stratifying analyses by HRT use, the results remained consistent with the main analysis, although the 95% CIs became wider because of a smaller sample size ( Supplemental Table 3 ). Excluding women who had diabetes, asthma, or rheumatoid arthritis before the age of 40 years did not change the results ( Supplemental Table 4 – 6 ). Excluding women with breast, ovarian, endometrial, or cervical cancer showed similar results ( Supplemental Table 7 ). E-values for infertility, recurrent miscarriages, stillbirth, and low parity ranged from 1.53 to 1.69 for the point estimates. These values suggested that to bring the observed association to the null, an unmeasured confounder would need to be associated with both the reproductive factor and osteoporosis with an HR of at least this magnitude ( Supplemental Table 8 ). Estimates from the imputed datasets had almost no changes from the main analysis ( Supplemental Table 9 ). In the causal mediation analysis, there was little evidence that age at natural menopause mediated the association between the fertility factors of interest and osteoporosis. ( Supplemental Table 10 ) Two-stage meta-analysis with random effects revealed similar results with the main analysis, although the 95% CIs for recurrent miscarriage and no live births became much wider ( Supplemental Fig. 1 – 4 , available online). In supplementary analysis, similar results were observed among women with natural or surgical menopause, although the associations with infertility, miscarriage, stillbirth, and low parity (no or one live birth) were stronger among women with surgical menopause ( Supplemental Table 11 – 12 ). Their associations were attenuated slightly after additional adjustment for age at menopause. Among women with hysterectomy or unilateral oophorectomy, the risk of osteoporosis was slightly elevated among women with low parity, and possibly increased among women with recurrent miscarriages ( Supplemental Table 13 ).

The InterLACE has pooled individual-level data from 27 observational studies. The present study included five cohorts from InterLACE, which were the Australian Longitudinal Study on Women’s Health 1946–51 cohort (ALSWH-mid) ( 26 ), the Australian Healthy Aging of Women’s study (HOW) ( 27 ), the Study of Women’s Health Across the Nation (SWAN) ( 28 ), UK Biobank ( 29 ), and the Dutch Prospect-EPIC Utrecht study in the European Prospective Investigation into Cancer and Nutrition (Prospect-EPIC) ( 30 ) Cohort characteristics are presented in Table 1 . The study sample included women who completed the first survey with osteoporosis data and reported natural menopause ( Fig. 1 ). Inclusion required no missing data on at least one of the relevant fertility factors, osteoporosis, and covariates. In the analyses of miscarriage and stillbirth, women who had never been pregnant or had missing data on their pregnancy history were excluded. Data on infertility, miscarriage, stillbirth, and parity were collected at baseline or follow-up surveys ( Supplemental Table 1 , available online). Women were considered to have infertility if they reported the experience of failing to become pregnant for 12 months or more, had a diagnosis of infertility, fertility treatment, or a medical consultation for infertility. The history of miscarriage and stillbirth (ever or never) was self-reported. The number of miscarriages was categorized as 0, 1, 2, and ≥3, with recurrent miscarriages defined as three or more miscarriages. Parity was categorized according to the number of live births (0, 1, 2, 3, and ≥4). Menopausal status was self-reported in multiple surveys. Natural menopause was deemed to have occurred after 12 consecutive months of amenorrhea without an obvious physiological or pathological cause ( 31 ). Age at natural menopause was calculated as the age when a woman reached natural menopause according to this definition. Among women with natural menopause, their menopause status was classified as before or after menopause. Premature and early menopause (menopause age <40 years and 40–44 years) are well-established risk factors of osteoporosis, and less is known about the risk among women experiencing menopause at a later age ( 22 ). To investigate the effect of age at natural menopause, the category of after menopause was further divided into seven groups (i.e., <40, 40–44, 45–47, 48–49, 50–51, 52–53, and ≥54). Osteoporosis was identified through survey, hospital, death registry, primary care, or pharmaceutical data ( Supplemental Table 1 ). All included cohorts had survey data on physician-diagnosed osteoporosis. Three cohorts (i. e., ALSWH-mid, UK Biobank, and Prospect-EPIC) provided death registry and hospital data, and primary care data were available only for a subset of UK Biobank participants. Death registry, hospital, and primary care data were coded according to the International Classification of Diseases-9 (ICD-9) or ICD-10 codes. International Classification of Diseases-9 codes A733.0 (osteoporosis), A733.1 (pathologic fracture), and ICD-10 codes M80 and M81 (osteoporosis with and without pathologic fracture) were used to identify osteoporosis. Cohort ALSWH-mid additionally had data on medications, and the selected items for osteoporosis were presented in Supplemental Table 1 . Women were classified as having osteoporosis if a diagnosis was recorded in any of the above data sources. Age at the earliest record of osteoporosis was taken as age at osteoporosis. Information on race (White, Asian, and others), education level (no formal qualification, year 10, year 12, diploma, and university), smoking status (never, past, and current smoker), alcohol intake (nondrinker, monthly, weekly, and daily drinker), body mass index (BMI; <18.5, 18.5–22.9, 23.0–27.4, and ≥27.5 kg/m 2 for Asian women; <18.5, 18.5–24.9, 25.0–29.9, and ≥30.0 kg/m 2 for White and other women), physical activity level (sedentary, low, moderate, and vigorous), and age at menarche (≤11, 12, 13, 14, and ≥15 years) were collected at cohort entrance. Physical activity level was categorized on the basis of metabolic equivalent of task minutes per day (<40, 40–<600, 600–<1200, and ≥1200 metabolic equivalent of task minutes per day) in all cohorts except HOW, where physical activity levels were self-rated into four levels. Categorical variables were presented as numbers and percentages, and continuous variables as medians and quartiles. Cox regression models were used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) for the associations of infertility, miscarriage, stillbirth, and parity with osteoporosis. Survival time was age at osteoporosis or age at last update of data, whichever came first. All models were adjusted for fixed covariates, including race, education level, smoking status, alcohol intake, BMI, physical activity level, and age at menarche. Cohort difference was taken into account by including cohort as a fixed covariate, and robust variance estimates were used to account for within-cohort correlation. To explore whether the age at natural menopause played a role in the associations between fertility factors and osteoporosis, all models were further adjusted for age at natural menopause as a time-varying covariate. If the effects of fertility factors were attenuated, this could imply that part of their association with osteoporosis was mediated by age at menopause. Several sensitivity analyses were conducted. First, ICD-10 code M82 (osteoporosis in diseases classified elsewhere) was additionally used to identify osteoporosis. Second, analyses were stratified by hormone replacement therapy (HRT) use at baseline, because HRT is often prescribed for women with early menopause (<45 years) and may reduce bone loss in postmenopausal women ( 32 ). Third, analyses were restricted to women who did not have diabetes, asthma, or rheumatoid arthritis before the age of 40 years. Women with these conditions were at a higher risk of osteoporosis, and the conditions were also associated with fertility issues ( 33 – 39 ). Similarly, the analyses were restricted to women who had not experienced breast cancer, ovarian cancer, endometrial cancer, or cervical cancer, given their influence on BMD through hormonal imbalance and treatment ( 40 , 41 ). Fourth, E-values were calculated to assess the potential influence of unmeasured confounding ( 42 , 43 ). Fifth, multiple imputation (10 times) using fully conditional specification was employed to impute missing data on covariates. Imputed datasets were used to assess the robustness of the findings to missing data. Sixth, a causal mediation analysis was performed. A linear regression model was used to estimate the association between fertility factor (exposure) and age at natural menopause (mediator), and a Weibull parametric survival model was used to estimate the association between exposure and outcome (osteoporosis), adjusting for age at natural menopause. All models included the fixed effects for cohort and were adjusted for the same covariates as the main analysis. Mediation effects were estimated using the mediate function in R version 4.2.2 (R Core Team, Vienna, Austria), under the assumption of no unmeasured confounding between fertility factor, age at natural menopause, and osteoporosis. Finally, a two-stage meta-analysis with random effects was conducted. Supplementary analyses were conducted among women with hysterectomy or unilateral oophorectomy; among women with surgical menopause (bilateral oophorectomy); and among women with natural or surgical menopause. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC), and mediation analysis was conducted using R version 4.2.2. Computer codes are available from the corresponding author on request. InterLACE used nonidentifiable data from existing studies and received ethics exemption from the University of Queensland (2024/ HE000390 ). Each study in the InterLACE received ethics approval from the relevant National Health Service research ethics committee, the Human Research Ethics Committee or the Institutional Review Board at each study institution.

Among women with natural menopause, a history of infertility, recurrent miscarriages, stillbirth, or low parity was associated with a higher risk of osteoporosis. Age at natural menopause had almost no or little influence on the observed associations. The present study extends current knowledge on female-specific risk factors of osteoporosis, which could be incorporated into the risk assessment and prevention of osteoporosis.

Among naturally postmenopausal women, a history of infertility, recurrent miscarriages (≥3), stillbirth, or low parity (≤1) was associated with the occurrence of osteoporosis. Additional adjustment for age at natural menopause had almost no or little influence on the observed associations. Two previous studies assessed the relationship between fertility issues (e.g., infertility, miscarriage, or stillbirth) and osteoporosis or BMD. In a cross-sectional study with 392 British women (aged 50–54 years), there was no clear association between the number of miscarriages and BMD at lumbar spine (linear regression correlation coefficient r = 0.03) or femoral neck (r = 0.01) ( 8 ). Another cross-sectional study, enrolling 3,476 Dutch women (aged 46–57 years), found an association between the number of miscarriages and lumbar spine BMD among premenopausal women (linear regression coefficient b = 0.0076, SE = 0.0036, P = .033), but not among postmenopausal women (b = 0.0023, SE = 0.0056, P = .679) ( 9 ). In the present study, 141,222 naturally postmenopausal women were enrolled and followed up to the age of 72.0 years (68.0, 76.0). With a large sample size and sufficient follow-up period, the present study indicated an association between miscarriage, particularly recurrent miscarriages, and osteoporosis among women with natural menopause. Moreover, the present study clarified the relationships of stillbirth and infertility with osteoporosis among women with natural menopause. Fertility issues might be linked to osteoporosis through estrogen and progestogen deficiency, which could contribute to ovulation and implantation issues, and an imbalance between osteogenesis and bone resorption ( 5 , 7 , 44 – 47 ). Additionally, among women with fertility issues, pre-existing and subsequent conditions (e.g., hyperthyroidism, hyperprolactinemia, unstable diabetes, and vitamin D insufficiency) could lead to bone loss ( 48 – 51 ). Despite a number of studies on parity and osteoporosis, results are inconclusive. Three cross-sectional studies enrolling postmenopausal women from the United States, Iran, and Jordan found that parity ≥6 was associated with higher odds of osteoporosis (≥6 vs. 1–2: odds ratio [OR] = 3.88, 95% CI: 1.64–9.18; ≥6 vs. ≤2: OR = 4.45, 95% CI: 1.69–11.74; and ≥6 vs. ≤2: OR = 1.6, P = .027) ( 10 , 11 , 14 ). In another Korean study, parity = 2 (OR = 2.46, 95% CI: 1.10–5.50) and parity≥3 (OR = 3.08, 95% CI: 1.30–7.30) were associated with the occurrence of osteoporosis, compared with women with one live birth ( 12 ). In support of these findings, studies by Yang et al. ( 11 ) and Panahi N et al. ( 15 ) revealed a relationship between high parity and low lumbar spine BMD (≥6 vs. 1–2: b = −0.072, 95% CI: −0.125, −0.018; >4 vs. ≤4: b = −0.0223, P = .012). However, two other cross-sectional studies did not detect any association between parity and osteoporosis (≥3 vs. ≤2: OR = 0.91, 95% CI: 0.51–1.92; per additional child: OR = 0.78, 95% CI: 0.29–2.10) ( 13 , 18 ). Another study by Mori et al. ( 19 ) suggested no relationship between parity and lumbar spine BMD (b = −0.026, 95% CI: −0.058, 0.006) or femoral neck BMD (b = 0.0002, 95% CI: −0.025, 0.025). Previous studies treated parity differently (continuous or categorical), and classified the number of live births inconsistently, which might contribute to the divergent results. In the present study, parity was classified as (0, 1, 2 [ref.], 3, and ≥4) to assess the effects of low and high parity. Low parity (0 or 1 live birth) was associated with a higher risk of osteoporosis among women with natural menopause. Women with low parity were more likely to have fertility issues and earlier natural menopause ( 20 ). Among them, estrogen deficiency in the early reproductive stage and estrogen reduction during menopause could lead to lower peak bone mass and earlier accelerated bone loss ( 45 ). In clinical practice, health care providers should be aware of the potentially increased risk of osteoporosis among women with these reproductive experiences. This information might inform earlier detection and prevention strategies. Future studies with data on BMD and osteoporotic fractures would provide more comprehensive information about the relationships between female reproductive factors and bone health. Besides, studies are also needed to assess whether taking into account these female reproductive factors could improve risk assessment for osteoporosis. The large sample size provided sufficient power to detect the associations between uncommon reproductive factors (e.g., recurrent miscarriages) and osteoporosis, which broadens the knowledge on female bone health. Nevertheless, there were several limitations. First, data on infertility, miscarriage, stillbirth, and parity were self-reported, which may introduce recall bias. Previous studies have shown that self-reported infertility (sensitivity = 72.0%, specificity = 70.0%), pregnancy loss (sensitivity = 73.5%, specificity = 99.4%), and parity (weighted kappa = 0.95, 95% CI: 0.92–0.97) are generally reliable, whereas stillbirth (sensitivity = 100.0%, specificity = 30.0%) may be overreported ( 52 – 55 ). Misclassification due to recall is likely to be nondifferential with respect to osteoporosis, which might bias observed associations toward null ( 56 ). Future studies combining self-reported data with other sources (e.g., patient registries and administrative data) may provide more accurate reproductive histories. In addition, the underlying causes of infertility, miscarriage, and stillbirth were not consistently collected across the pooled dataset and could not be accounted for in analysis. For instance, polycystic ovary syndrome (PCOS) has been linked to lower BMD among women with low BMI ( 57 , 58 ). Therefore, the relationship between infertility and osteoporosis may vary depending on the presence of PCOS. However, data on PCOS were only available for 38.1% of the women with infertility data. Among them, only 1.84% reported PCOS (possibly because of underreporting). Additionally, this study could not distinguish between female and male infertility. Misclassifying women who were not biologically infertile as infertile would dilute the observed association between infertility and osteoporosis. Nevertheless, males are solely responsible for 20%–30% of infertility cases, with the majority involving female or male-female factors ( 34 ). Our findings are likely to reflect the predominant contribution of female or combined female-male factors. Second, this study lacked detailed data on HRT (e.g., type, dose, and duration). Therefore, the effect of HRT on osteoporosis could not be fully explored, which is particularly for women who experienced surgical menopause. Hormone replacement therapy is commonly prescribed to women with surgical menopause to mitigate sudden hormone loss. Combining data from women with natural or surgical menopause may bias the association with osteoporosis because of insufficient adjustment for HRT use ( 23 , 59 ). Third, although multiple data sources were used to identify women with osteoporosis, some data sources (e.g., pharmaceutical data and primary care data) were only available for part of the participants ( Supplemental Table 14 ). In addition, participants from some cohorts were followed up to a relatively young age, which was not enough for the occurrence of osteoporosis ( Supplemental Table 14 ). These factors may have led to a lower prevalence of osteoporosis in the pooled data than reported elsewhere ( 1 , 2 ). Analysis by osteoporosis severity was not conducted, as relevant data (e.g. BMD measurements and fragility fracture history) were unavailable. The outcome was based on self-report or physician diagnosis, which does not allow differentiation between mild and severe osteoporosis. Future studies with such data may provide deeper insight into the association. Fourth, lifestyle factors were not consistently or repeatedly collected across all cohorts. Survey intervals varied from 1 to 10 years among pooled cohorts, and loss to follow-up led to a high proportion of missing data. Therefore, lifestyle changes were not included in the analysis. Fifth, there might be unmeasured confounder(s) (e.g., dietary patterns). E-values indicated that the observed associations might be moderately robust to unmeasured confounding. For example, the western dietary pattern has been modestly associated with osteoporosis (relative risk = 1.46, 95% CI: 1.02–2.10), but there is no clear evidence on the association with infertility or pregnancy loss ( 60 – 62 ). It is unlikely that dietary pattern alone could account for the observed effects, given the E-values. However, residual confounding remains a possibility, and future studies with more comprehensive lifestyle data are needed. Finally, selection bias might exist. There were 20,619 (12.7%) women with natural menopause who were excluded because of missing data. Excluded women were more likely to be less educated, physically inactive, and obese than those included ( Supplemental Table 15 ). However, multiple imputations suggested the findings would be robust to missing data ( Supplemental Table 9 ).

Supplemental data for this article can be found online at https://doi.org/10.1016/j.fertnstert.2025.07.382 .