A large prospective cohort study evaluating 119,969 female participants from the UK Biobank found that giving birth to two or three children was associated with a significantly reduced long-term risk of developing osteoporosis compared with having no live births. The investigation, published in Archives of Osteoporosis, clarified that while pregnancy and lactation involve periods of intense maternal calcium expenditure and transient bone loss, higher parity did not elevate the risk of postmenopausal osteoporosis later in life. Instead, the analysis revealed a distinct non-linear relationship where long-term skeletal protection emerged most strongly among women reporting two or three live births, whereas women reporting four or more live births did not exhibit statistically significant protective differences after controlling for clinical, genetic, and sociodemographic variables.
To evaluate these long-term associations, researchers drew upon data from the multicenter, population-based UK Biobank cohort, which enrolled 502,236 adults aged 40 to 69 years across 22 assessment centers located throughout England, Scotland, and Wales between 2006 and 2010. The study cohort was refined through rigorous inclusion and exclusion criteria designed to minimize potential confounding and reverse causation. Investigators excluded individuals with missing records for parity or key covariates, as well as participants who reported a history of malignant cancer, hyperthyroidism, hypothyroidism, osteoarthritis, or inflammatory bowel disease at baseline. In addition, women who had received prior prescriptions for osteoporosis medications or had already been diagnosed with osteoporosis before baseline assessment were excluded, yielding a final analytical sample of 119,969 women.
At the baseline assessment visit, participants recorded their reproductive history by answering a touch-screen question asking how many live births they had delivered. The study followed participants prospectively from baseline until their first diagnosis of osteoporosis using International Classification of Diseases Tenth Revision codes M80 through M82, death, or the end of the follow-up period on November 30, 2022. During the observation window, 5,390 participants, representing approximately 4.5 percent of the analyzed cohort, received a confirmed diagnosis of osteoporosis through linked national electronic inpatient records. Secondary endpoints tracked by the investigators included incident osteoporotic fractures and dual-energy X-ray absorptiometry measurements of areal bone mineral density at the lumbar spine and femoral neck, which were obtained in an imaging substudy of 12,579 participants beginning in 2014.
A comparison of baseline characteristics between participants who developed osteoporosis and those who remained disease-free revealed several prominent demographic and clinical differences. Those diagnosed with osteoporosis were significantly older on average, with a mean age of 60.45 years compared to 54.67 years in the non-osteoporosis group. Women who developed osteoporosis also reported a shorter mean reproductive duration of 8.27 years compared with 13.61 years among unaffected participants, and a higher proportion of them had already reached postmenopausal status at baseline, recorded at 91.6 percent versus 64.2 percent. Furthermore, participants who developed osteoporosis exhibited lower mean body mass index values, with greater proportions categorized as normal weight or underweight, alongside a higher baseline prevalence of weak grip strength, prior hormone therapy use, and cardiovascular comorbidities such as hypertension, coronary heart disease, and stroke.
To address potential confounding across parity categories, the investigators constructed weighted pseudo-populations using inverse probability of treatment weighting based on propensity scores. Propensity scores were estimated through logistic regression models incorporating age, body mass index, ethnicity, educational attainment, smoking status, alcohol consumption patterns, diet quality, physical activity levels, residential area classification, employment status, the Townsend deprivation index, age at menarche, measured grip strength, and an osteoporosis polygenic risk score. In unweighted multivariable Cox proportional hazards regression models adjusted for this full suite of covariates, women with a parity of 2 demonstrated a hazard ratio of 0.904, and those with a parity of 3 exhibited a hazard ratio of 0.894, with both estimates achieving statistical significance compared to nulliparous women.
The application of inverse probability of treatment weighting confirmed this persistent inverse association across the primary outcome models. In the weighted, fully adjusted Model 3, parity of 2 was associated with a hazard ratio of 0.894, with a 95 percent confidence interval spanning 0.825 to 0.970, and parity of 3 yielded a hazard ratio of 0.892, with a 95 percent confidence interval ranging from 0.811 to 0.981. Although parity of 1 and parity of 4 exhibited point estimates below 1.00 relative to nulliparity, neither achieved statistical significance. Women reporting five or more live births displayed point estimates slightly above 1.00, but this trend was likewise not statistically distinguishable from the null reference group. Sensitivity analyses excluding participants who were diagnosed with osteoporosis during the first two years of follow-up reaffirmed that a history of two or three live births was associated with decreased disease risk.
When examining secondary skeletal endpoints, researchers observed a parallel reduction in the risk of incident osteoporotic fractures among women who reported three live births. In unweighted fully adjusted Cox regression, parity of 3 was associated with an osteoporotic fracture hazard ratio of 0.667, with a 95 percent confidence interval of 0.472 to 0.942, while the weighted model produced a hazard ratio of 0.645 with a 95 percent confidence interval between 0.449 and 0.925. A sensitivity analysis corroborating these findings yielded a hazard ratio of 0.673. However, when evaluating areal bone mineral density among the 12,579 women with dual-energy X-ray absorptiometry scans, the protective association did not mirror differences in measured mineral density. Although crude models showed slightly lower femoral neck density in parous women, the association diminished and became non-significant after adjusting for clinical and demographic covariates in all weighted and unweighted multivariable models.
Subgroup analyses stratified by demographic and metabolic criteria demonstrated that the observed protection was concentrated within specific patient subsets. Stratification using the World Health Organization threshold of 60 years revealed no statistically significant association between parity and osteoporosis among participants under 60 years of age. Conversely, among women aged 60 years and older, parities of 2, 3, and 4 were each associated with a statistically significant reduction in osteoporosis risk across both unweighted and weighted models. Furthermore, stratification by body mass index categories revealed that the statistically significant protective association for a parity of 2 was confined to women within the normal weight range of 18.5 to 24.9 kilograms per square meter, whereas no significant protective relationship was observed in participants categorized as overweight or obese.
The researchers noted that the dissociation between clinical fracture protection and unchanged areal bone mineral density underscores the diagnostic constraints of dual-energy X-ray absorptiometry, which captures total mineral mass but does not measure structural microarchitecture or separate cortical adaptations from trabecular compartments. The authors contextualized these findings within prior inconsistent literature, noting that earlier studies by Seo, Egund, Gildee, Long, Liang, and Wang had reported conflicting outcomes due to cross-sectional designs, smaller samples, heterogeneous grouping methods, or residual confounding. To explain the potential biological mechanisms, the study authors hypothesized that pregnancy and lactation expose maternal bones to repeated cycles of hormonal shifts involving estrogen, progesterone, relaxin, parathyroid hormone-related protein, and prolactin. They suggested these remodeling cycles might induce epigenetic adaptations in mesenchymal and hematopoietic stem cells or promote permanent compensatory enhancements in maternal renal and intestinal calcium regulation.
Despite its large prospective cohort and extensive covariate adjustment, the investigation noted important methodological limitations. The UK Biobank dataset lacked direct measurements of bone microarchitecture, bone turnover markers, and maternal age at last childbirth, the latter of which is critical because births occurring closer to the menopausal transition may exert a more profound and long-term influence on maternal bone. Additionally, the registry contained no information regarding infant birth weight or breastfeeding practices, including lactation duration and intensity, precluding adjustment for these known influences on bone remodeling.
Subject of Research: Association between parity and long-term osteoporosis risk in women
Article Title: Non-linear associations between parity and long-term risk of osteoporosis
Article References: Yuan, W., Liao, J., Tao, L., Song, H., Wang, H., Huang, J., He, J., Liu, H., Wu, T., Zhou, H., Zhang, Y., & Song, C. (2026). Non-linear associations between parity and long-term risk of osteoporosis. Archives of Osteoporosis, 21(1), Article 159. https://doi.org/10.1007/s11657-026-01766-x
Image Credits: AI Generated
DOI: 10.1007/s11657-026-01766-x
Keywords: osteoporosis, parity, bone mineral density, UK Biobank, fracture risk, women's health, bone remodeling, epidemiology, Non-linear, associations, between, long-term
News Source: Ophelia Keating. (November 6, 2026). Childbirth History Linked to Reduced Long-Term Osteoporosis Risk in UK Women. Scienmag.
